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
46 #include <linux/limits.h>
47 #include <linux/kernel.h>
48 #include <linux/string.h>
49 #include <linux/quota.h>
50 #include <linux/types.h>
51 #include <linux/unistd.h>
52 #include <linux/lustre/lustre_fiemap.h>
55 # define __USE_ISOC99 1
57 # include <stdio.h> /* snprintf() */
58 # include <sys/stat.h>
59 # define FILEID_LUSTRE 0x97 /* for name_to_handle_at() (and llapi_fd2fid()) */
60 #endif /* !__KERNEL__ */
62 #if defined(__cplusplus)
66 #ifdef __STRICT_ANSI__
67 #define typeof __typeof__
71 * This is a temporary solution of adding quota type.
72 * Should be removed as soon as system header is updated.
75 #define LL_MAXQUOTAS 3
77 #define INITQFNAMES { \
78 "user", /* USRQUOTA */ \
79 "group", /* GRPQUOTA */ \
80 "project", /* PRJQUOTA */ \
94 * We need to always use 64bit version because the structure
95 * is shared across entire cluster where 32bit and 64bit machines
98 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
99 typedef struct stat64 lstat_t;
100 #define lstat_f lstat64
101 #define fstat_f fstat64
102 #define fstatat_f fstatat64
104 typedef struct stat lstat_t;
105 #define lstat_f lstat
106 #define fstat_f fstat
107 #define fstatat_f fstatat
110 #ifndef STATX_BASIC_STATS
112 * Timestamp structure for the timestamps in struct statx.
114 * tv_sec holds the number of seconds before (negative) or after (positive)
115 * 00:00:00 1st January 1970 UTC.
117 * tv_nsec holds a number of nanoseconds (0..999,999,999) after the tv_sec time.
119 * __reserved is held in case we need a yet finer resolution.
121 struct statx_timestamp {
128 * Structures for the extended file attribute retrieval system call
131 * The caller passes a mask of what they're specifically interested in as a
132 * parameter to statx(). What statx() actually got will be indicated in
133 * st_mask upon return.
135 * For each bit in the mask argument:
137 * - if the datum is not supported:
139 * - the bit will be cleared, and
141 * - the datum will be set to an appropriate fabricated value if one is
142 * available (eg. CIFS can take a default uid and gid), otherwise
144 * - the field will be cleared;
146 * - otherwise, if explicitly requested:
148 * - the datum will be synchronised to the server if AT_STATX_FORCE_SYNC is
149 * set or if the datum is considered out of date, and
151 * - the field will be filled in and the bit will be set;
153 * - otherwise, if not requested, but available in approximate form without any
154 * effort, it will be filled in anyway, and the bit will be set upon return
155 * (it might not be up to date, however, and no attempt will be made to
156 * synchronise the internal state first);
158 * - otherwise the field and the bit will be cleared before returning.
160 * Items in STATX_BASIC_STATS may be marked unavailable on return, but they
161 * will have values installed for compatibility purposes so that stat() and
162 * co. can be emulated in userspace.
166 __u32 stx_mask; /* What results were written [uncond] */
167 __u32 stx_blksize; /* Preferred general I/O size [uncond] */
168 __u64 stx_attributes; /* Flags conveying information about the file [uncond] */
170 __u32 stx_nlink; /* Number of hard links */
171 __u32 stx_uid; /* User ID of owner */
172 __u32 stx_gid; /* Group ID of owner */
173 __u16 stx_mode; /* File mode */
176 __u64 stx_ino; /* Inode number */
177 __u64 stx_size; /* File size */
178 __u64 stx_blocks; /* Number of 512-byte blocks allocated */
179 __u64 stx_attributes_mask; /* Mask to show what's supported in stx_attributes */
181 struct statx_timestamp stx_atime; /* Last access time */
182 struct statx_timestamp stx_btime; /* File creation time */
183 struct statx_timestamp stx_ctime; /* Last attribute change time */
184 struct statx_timestamp stx_mtime; /* Last data modification time */
186 __u32 stx_rdev_major; /* Device ID of special file [if bdev/cdev] */
187 __u32 stx_rdev_minor;
188 __u32 stx_dev_major; /* ID of device containing file [uncond] */
191 __u64 __spare2[14]; /* Spare space for future expansion */
196 * Flags to be stx_mask
198 * Query request/result mask for statx() and struct statx::stx_mask.
200 * These bits should be set in the mask argument of statx() to request
201 * particular items when calling statx().
203 #define STATX_TYPE 0x00000001U /* Want/got stx_mode & S_IFMT */
204 #define STATX_MODE 0x00000002U /* Want/got stx_mode & ~S_IFMT */
205 #define STATX_NLINK 0x00000004U /* Want/got stx_nlink */
206 #define STATX_UID 0x00000008U /* Want/got stx_uid */
207 #define STATX_GID 0x00000010U /* Want/got stx_gid */
208 #define STATX_ATIME 0x00000020U /* Want/got stx_atime */
209 #define STATX_MTIME 0x00000040U /* Want/got stx_mtime */
210 #define STATX_CTIME 0x00000080U /* Want/got stx_ctime */
211 #define STATX_INO 0x00000100U /* Want/got stx_ino */
212 #define STATX_SIZE 0x00000200U /* Want/got stx_size */
213 #define STATX_BLOCKS 0x00000400U /* Want/got stx_blocks */
214 #define STATX_BASIC_STATS 0x000007ffU /* The stuff in the normal stat struct */
215 #define STATX_BTIME 0x00000800U /* Want/got stx_btime */
216 #define STATX_ALL 0x00000fffU /* All currently supported flags */
217 #define STATX__RESERVED 0x80000000U /* Reserved for future struct statx expansion */
220 * Attributes to be found in stx_attributes and masked in stx_attributes_mask.
222 * These give information about the features or the state of a file that might
223 * be of use to ordinary userspace programs such as GUIs or ls rather than
226 * Note that the flags marked [I] correspond to generic FS_IOC_FLAGS
227 * semantically. Where possible, the numerical value is picked to correspond
230 #define STATX_ATTR_COMPRESSED 0x00000004 /* [I] File is compressed by the fs */
231 #define STATX_ATTR_IMMUTABLE 0x00000010 /* [I] File is marked immutable */
232 #define STATX_ATTR_APPEND 0x00000020 /* [I] File is append-only */
233 #define STATX_ATTR_NODUMP 0x00000040 /* [I] File is not to be dumped */
234 #define STATX_ATTR_ENCRYPTED 0x00000800 /* [I] File requires key to decrypt in fs */
236 #define STATX_ATTR_AUTOMOUNT 0x00001000 /* Dir: Automount trigger */
240 typedef struct statx lstatx_t;
242 #define HAVE_LOV_USER_MDS_DATA
244 #define LUSTRE_EOF 0xffffffffffffffffULL
247 #define LL_SUPER_MAGIC 0x0BD00BD0
249 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
251 /* FIEMAP flags supported by Lustre */
252 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
254 enum obd_statfs_state {
255 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
256 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
257 OS_STATE_NOPRECREATE = 0x00000004, /**< no object precreation */
258 OS_STATE_UNUSED1 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
259 OS_STATE_UNUSED2 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
260 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
261 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
262 OS_STATE_SUM = 0x00000100, /**< aggregated for all tagrets */
263 OS_STATE_NONROT = 0x00000200, /**< non-rotational device */
266 /** filesystem statistics/attributes for target device */
268 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
269 __u64 os_blocks; /* total size in #os_bsize blocks */
270 __u64 os_bfree; /* number of unused blocks */
271 __u64 os_bavail; /* blocks available for allocation */
272 __u64 os_files; /* total number of objects */
273 __u64 os_ffree; /* # objects that could be created */
274 __u8 os_fsid[40]; /* identifier for filesystem */
275 __u32 os_bsize; /* block size in bytes for os_blocks */
276 __u32 os_namelen; /* maximum length of filename in bytes*/
277 __u64 os_maxbytes; /* maximum object size in bytes */
278 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
279 __u32 os_fprecreated; /* objs available now to the caller */
280 /* used in QoS code to find preferred
282 __u32 os_granted; /* space granted for MDS */
283 __u32 os_spare3; /* Unused padding fields. Remember */
284 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
292 /** additional filesystem attributes for target device */
293 struct obd_statfs_info {
294 __u32 os_reserved_mb_low; /* reserved mb low */
295 __u32 os_reserved_mb_high; /* reserved mb high */
296 bool os_enable_pre; /* enable pre create logic */
302 * FID is a cluster-wide unique identifier of a file or an object (stripe).
303 * FIDs are never reused.
307 * FID sequence. Sequence is a unit of migration: all files (objects)
308 * with FIDs from a given sequence are stored on the same server.
309 * Lustre should support 2^64 objects, so even if each sequence
310 * has only a single object we can still enumerate 2^64 objects.
313 /* FID number within sequence. */
316 * FID version, used to distinguish different versions (in the sense
317 * of snapshots, etc.) of the same file system object. Not currently
321 } __attribute__((packed));
323 static inline bool fid_is_zero(const struct lu_fid *fid)
325 return fid->f_seq == 0 && fid->f_oid == 0;
328 /* The data name_to_handle_at() places in a struct file_handle (at f_handle) */
329 struct lustre_file_handle {
330 struct lu_fid lfh_child;
331 struct lu_fid lfh_parent;
334 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
335 * MDT-object's FID::f_ver, instead it is the OST-object index in its
336 * parent MDT-object's layout EA. */
337 #define f_stripe_idx f_ver
340 __u32 ol_stripe_size;
341 __u32 ol_stripe_count;
345 } __attribute__((packed));
347 /* The filter_fid structure has changed several times over its lifetime.
348 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
349 * stripe_index and the "self FID" (objid/seq) to be able to recover the
350 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
351 * the OST, the "trusted.lma" xattr was added to the OST objects to store
352 * the "self FID" to be consistent with the MDT on-disk format, and the
353 * filter_fid only stored the MDT inode parent FID and stripe index.
355 * In 2.10, the addition of PFL composite layouts required more information
356 * to be stored into the filter_fid in order to be able to identify which
357 * component the OST object belonged. As well, the stripe size may vary
358 * between components, so it was no longer safe to assume the stripe size
359 * or stripe_count of a file. This is also more robust for plain layouts.
361 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
362 * enough space to store both the filter_fid and LMA in the inode, so they
363 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
364 * an extra seek for every OST object access.
366 * In 2.11, FLR mirror layouts also need to store the layout version and
367 * range so that writes to old versions of the layout are not allowed.
368 * That ensures that mirrored objects are not modified by evicted clients,
369 * and ensures that the components are correctly marked stale on the MDT.
371 struct filter_fid_18_23 {
372 struct lu_fid ff_parent; /* stripe_idx in f_ver */
377 struct filter_fid_24_29 {
378 struct lu_fid ff_parent; /* stripe_idx in f_ver */
381 struct filter_fid_210 {
382 struct lu_fid ff_parent; /* stripe_idx in f_ver */
383 struct ost_layout ff_layout;
387 struct lu_fid ff_parent; /* stripe_idx in f_ver */
388 struct ost_layout ff_layout;
389 __u32 ff_layout_version;
390 __u32 ff_range; /* range of layout version that
391 * write are allowed */
392 } __attribute__((packed));
394 /* Userspace should treat lu_fid as opaque, and only use the following methods
395 * to print or parse them. Other functions (e.g. compare, swab) could be moved
396 * here from lustre_idl.h if needed. */
400 LMAC_HSM = 0x00000001,
401 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
402 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
403 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
404 * under /O/<seq>/d<x>. */
405 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
406 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
407 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
411 * Masks for all features that should be supported by a Lustre version to
412 * access a specific file.
413 * This information is stored in lustre_mdt_attrs::lma_incompat.
416 LMAI_RELEASED = 0x00000001, /* file is released */
417 LMAI_AGENT = 0x00000002, /* agent inode */
418 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
419 is on the remote MDT */
420 LMAI_STRIPED = 0x00000008, /* striped directory inode */
421 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
422 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
423 LMAI_STRIPED | LMAI_ORPHAN)
428 * Following struct for object attributes, that will be kept inode's EA.
429 * Introduced in 2.0 release (please see b15993, for details)
430 * Added to all objects since Lustre 2.4 as contains self FID
432 struct lustre_mdt_attrs {
434 * Bitfield for supported data in this structure. From enum lma_compat.
435 * lma_self_fid and lma_flags are always available.
439 * Per-file incompat feature list. Lustre version should support all
440 * flags set in this field. The supported feature mask is available in
444 /** FID of this inode */
445 struct lu_fid lma_self_fid;
448 struct lustre_ost_attrs {
449 /* Use lustre_mdt_attrs directly for now, need a common header
450 * structure if want to change lustre_mdt_attrs in future. */
451 struct lustre_mdt_attrs loa_lma;
453 /* Below five elements are for OST-object's PFID EA, the
454 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
455 * and the stripe_index (low 16 bits), the size should not exceed
456 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
457 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
458 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
459 * loa_comp_* elements are valid. */
460 struct lu_fid loa_parent_fid;
461 __u32 loa_stripe_size;
463 __u64 loa_comp_start;
468 * Prior to 2.4, the LMA structure also included SOM attributes which has since
469 * been moved to a dedicated xattr
470 * lma_flags was also removed because of lma_compat/incompat fields.
472 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
474 enum lustre_som_flags {
475 /* Unknow or no SoM data, must get size from OSTs. */
476 SOM_FL_UNKNOWN = 0x0000,
477 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
478 SOM_FL_STRICT = 0x0001,
479 /* Known stale - was right at some point in the past, but it is
480 * known (or likely) to be incorrect now (e.g. opened for write). */
481 SOM_FL_STALE = 0x0002,
482 /* Approximate, may never have been strictly correct,
483 * need to sync SOM data to achieve eventual consistency. */
484 SOM_FL_LAZY = 0x0004,
487 struct lustre_som_attrs {
489 __u16 lsa_reserved[3];
495 * OST object IDentifier.
503 struct lu_fid oi_fid;
505 } __attribute__((packed));
507 #define DOSTID "%#llx:%llu"
508 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
509 ((unsigned long long)ostid_id(oi))
511 struct ll_futimes_3 {
513 __u64 lfu_atime_nsec;
515 __u64 lfu_mtime_nsec;
517 __u64 lfu_ctime_nsec;
521 * Maximum number of mirrors currently implemented.
523 #define LUSTRE_MIRROR_COUNT_MAX 16
525 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
527 LL_LEASE_RDLCK = 0x01,
528 LL_LEASE_WRLCK = 0x02,
529 LL_LEASE_UNLCK = 0x04,
532 enum ll_lease_flags {
533 LL_LEASE_RESYNC = 0x1,
534 LL_LEASE_RESYNC_DONE = 0x2,
535 LL_LEASE_LAYOUT_MERGE = 0x4,
536 LL_LEASE_LAYOUT_SPLIT = 0x8,
537 LL_LEASE_PCC_ATTACH = 0x10,
540 #define IOC_IDS_MAX 4096
541 struct ll_ioc_lease {
548 struct ll_ioc_lease_id {
559 * The ioctl naming rules:
560 * LL_* - works on the currently opened filehandle instead of parent dir
561 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
562 * *_MDC_* - gets/sets data related to MDC
563 * *_LOV_* - gets/sets data related to OSC/LOV
564 * *FILE* - called on parent dir and passes in a filename
565 * *STRIPE* - set/get lov_user_md
566 * *INFO - set/get lov_user_mds_data
568 /* lustre_ioctl.h 101-150 */
569 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
570 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
571 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
572 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
573 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
574 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
575 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
576 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
577 /* LL_IOC_RECREATE_OBJ 157 obsolete */
578 /* LL_IOC_RECREATE_FID 157 obsolete */
579 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
580 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
581 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
582 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
583 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
584 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
585 /* IOC_LOV_GETINFO 165 obsolete */
586 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
587 /* LL_IOC_RMTACL 167 obsolete */
588 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
589 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
590 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
591 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
592 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
593 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
594 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
595 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
596 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
597 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
598 /* lustre_ioctl.h 177-210 */
599 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
600 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
601 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
602 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
603 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
604 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
605 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
606 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
607 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
608 struct lustre_swap_layouts)
609 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
610 struct hsm_current_action)
611 /* lustre_ioctl.h 221-232 */
612 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
613 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
614 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
615 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
616 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
617 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
618 #define LL_IOC_GET_LEASE _IO('f', 244)
619 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
620 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
621 #define LL_IOC_MIGRATE _IOR('f', 247, int)
622 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
623 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
624 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
625 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
626 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
627 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
628 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
629 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
631 #ifndef FS_IOC_FSGETXATTR
633 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
636 __u32 fsx_xflags; /* xflags field value (get/set) */
637 __u32 fsx_extsize; /* extsize field value (get/set)*/
638 __u32 fsx_nextents; /* nextents field value (get) */
639 __u32 fsx_projid; /* project identifier (get/set) */
640 unsigned char fsx_pad[12];
642 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
643 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
645 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
646 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
647 #ifndef FS_XFLAG_PROJINHERIT
648 #define FS_XFLAG_PROJINHERIT 0x00000200
652 #define LL_STATFS_LMV 1
653 #define LL_STATFS_LOV 2
654 #define LL_STATFS_NODELAY 4
656 #define IOC_MDC_TYPE 'i'
657 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
658 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
659 #ifdef HAVE_LOV_USER_MDS_DATA
660 #define IOC_MDC_GETFILEINFO_OLD _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
661 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
662 #define LL_IOC_MDC_GETINFO_OLD _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
663 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
666 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
668 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
669 * files, but are unlikely to be used in practice and are not harmful if
670 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
671 * devices and are safe for use on new files. See LU-4209. */
672 /* To be compatible with old statically linked binary we keep the check for
673 * the older 0100000000 flag. This is already removed upstream. LU-812. */
674 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
676 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
678 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
679 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
680 O_LOV_DELAY_CREATE_MASK)
682 #define LL_FILE_IGNORE_LOCK 0x00000001
683 #define LL_FILE_GROUP_LOCKED 0x00000002
684 #define LL_FILE_READAHEA 0x00000004
685 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
686 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
687 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
689 #define LOV_USER_MAGIC_V1 0x0BD10BD0
690 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
691 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
692 #define LOV_USER_MAGIC_V3 0x0BD30BD0
693 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
694 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
695 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
696 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
697 #define LOV_USER_MAGIC_SEL 0x0BD80BD0
699 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
700 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
701 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
703 #define LOV_PATTERN_NONE 0x000
704 #define LOV_PATTERN_RAID0 0x001
705 #define LOV_PATTERN_RAID1 0x002
706 #define LOV_PATTERN_MDT 0x100
707 #define LOV_PATTERN_OVERSTRIPING 0x200
709 #define LOV_PATTERN_F_MASK 0xffff0000
710 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
711 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
712 #define LOV_PATTERN_DEFAULT 0xffffffff
714 #define LOV_OFFSET_DEFAULT ((__u16)-1)
715 #define LMV_OFFSET_DEFAULT ((__u32)-1)
717 static inline bool lov_pattern_supported(__u32 pattern)
719 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
720 (pattern & ~LOV_PATTERN_F_RELEASED) ==
721 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
722 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
725 /* RELEASED and MDT patterns are not valid in many places, so rather than
726 * having many extra checks on lov_pattern_supported, we have this separate
727 * check for non-released, non-DOM components
729 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
731 return pattern == LOV_PATTERN_RAID0 ||
732 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
736 #define LOV_MAXPOOLNAME 15
737 #define LOV_POOLNAMEF "%.15s"
739 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
740 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
741 #define LOV_MAX_STRIPE_COUNT_OLD 160
742 /* This calculation is crafted so that input of 4096 will result in 160
743 * which in turn is equal to old maximal stripe count.
744 * XXX: In fact this is too simpified for now, what it also need is to get
745 * ea_type argument to clearly know how much space each stripe consumes.
747 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
748 * allocation that is sufficient for the current generation of systems.
750 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
751 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
752 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
753 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
755 #define XATTR_LUSTRE_PREFIX "lustre."
756 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
758 /* Please update if XATTR_LUSTRE_LOV".set" groks more flags in the future */
759 #define allowed_lustre_lov(att) (strcmp((att), XATTR_LUSTRE_LOV".add") == 0 || \
760 strcmp((att), XATTR_LUSTRE_LOV".set") == 0 || \
761 strcmp((att), XATTR_LUSTRE_LOV".set.flags") == 0 || \
762 strcmp((att), XATTR_LUSTRE_LOV".del") == 0)
764 #define lov_user_ost_data lov_user_ost_data_v1
765 struct lov_user_ost_data_v1 { /* per-stripe data structure */
766 struct ost_id l_ost_oi; /* OST object ID */
767 __u32 l_ost_gen; /* generation of this OST index */
768 __u32 l_ost_idx; /* OST index in LOV */
769 } __attribute__((packed));
771 #define lov_user_md lov_user_md_v1
772 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
773 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
774 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
775 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
776 __u32 lmm_stripe_size; /* size of stripe in bytes */
777 __u16 lmm_stripe_count; /* num stripes in use for this object */
779 __u16 lmm_stripe_offset; /* starting stripe offset in
780 * lmm_objects, use when writing */
781 __u16 lmm_layout_gen; /* layout generation number
782 * used when reading */
784 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
785 } __attribute__((packed, __may_alias__));
787 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
788 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
789 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
790 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
791 __u32 lmm_stripe_size; /* size of stripe in bytes */
792 __u16 lmm_stripe_count; /* num stripes in use for this object */
794 __u16 lmm_stripe_offset; /* starting stripe offset in
795 * lmm_objects, use when writing */
796 __u16 lmm_layout_gen; /* layout generation number
797 * used when reading */
799 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
800 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
801 } __attribute__((packed, __may_alias__));
803 struct lov_foreign_md {
804 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
805 __u32 lfm_length; /* length of lfm_value */
806 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
807 __u32 lfm_flags; /* flags, type specific */
809 } __attribute__((packed));
811 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
812 offsetof(struct lov_foreign_md, lfm_value))
814 #define foreign_size_le(lfm) \
815 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
816 offsetof(struct lov_foreign_md, lfm_value))
819 * The stripe size fields are shared for the extension size storage, however
820 * the extension size is stored in KB, not bytes.
822 #define SEL_UNIT_SIZE 1024llu
827 } __attribute__((packed));
829 #define DEXT "[%#llx, %#llx)"
830 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
832 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
833 struct lu_extent *e2)
835 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
838 static inline bool lu_extent_is_whole(struct lu_extent *e)
840 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
843 enum lov_comp_md_entry_flags {
844 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
845 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
846 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
847 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
848 LCME_FL_OFFLINE = 0x00000008, /* Not used */
849 LCME_FL_INIT = 0x00000010, /* instantiated */
850 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
851 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
852 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
853 * won't be stored on disk
857 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
858 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
861 /* The component flags can be set by users at creation/modification time. */
862 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
865 /* The mirror flags can be set by users at creation time. */
866 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
868 /* The allowed flags obtained from the client at component creation time. */
869 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
871 /* The mirror flags sent by client */
872 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
874 /* These flags have meaning when set in a default layout and will be inherited
875 * from the default/template layout set on a directory.
877 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
880 /* the highest bit in obdo::o_layout_version is used to mark if the file is
882 #define LU_LAYOUT_RESYNC LCME_FL_NEG
884 /* lcme_id can be specified as certain flags, and the the first
885 * bit of lcme_id is used to indicate that the ID is representing
886 * certain LCME_FL_* but not a real ID. Which implies we can have
887 * at most 31 flags (see LCME_FL_XXX). */
890 LCME_ID_MAX = 0x7FFFFFFF,
891 LCME_ID_ALL = 0xFFFFFFFF,
892 LCME_ID_NOT_ID = LCME_FL_NEG
895 #define LCME_ID_MASK LCME_ID_MAX
897 struct lov_comp_md_entry_v1 {
898 __u32 lcme_id; /* unique id of component */
899 __u32 lcme_flags; /* LCME_FL_XXX */
900 struct lu_extent lcme_extent; /* file extent for component */
901 __u32 lcme_offset; /* offset of component blob,
902 start from lov_comp_md_v1 */
903 __u32 lcme_size; /* size of component blob */
904 __u32 lcme_layout_gen;
905 __u64 lcme_timestamp; /* snapshot time if applicable*/
906 __u32 lcme_padding_1;
907 } __attribute__((packed));
909 #define SEQ_ID_MAX 0x0000FFFF
910 #define SEQ_ID_MASK SEQ_ID_MAX
911 /* bit 30:16 of lcme_id is used to store mirror id */
912 #define MIRROR_ID_MASK 0x7FFF0000
913 #define MIRROR_ID_NEG 0x8000
914 #define MIRROR_ID_SHIFT 16
916 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
918 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
921 static inline __u16 mirror_id_of(__u32 id)
923 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
927 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
929 enum lov_comp_md_flags {
930 /* the least 2 bits are used by FLR to record file state */
933 LCM_FL_WRITE_PENDING = 2,
934 LCM_FL_SYNC_PENDING = 3,
935 LCM_FL_FLR_MASK = 0x3,
938 struct lov_comp_md_v1 {
939 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
940 __u32 lcm_size; /* overall size including this struct */
941 __u32 lcm_layout_gen;
943 __u16 lcm_entry_count;
944 /* lcm_mirror_count stores the number of actual mirrors minus 1,
945 * so that non-flr files will have value 0 meaning 1 mirror. */
946 __u16 lcm_mirror_count;
947 __u16 lcm_padding1[3];
949 struct lov_comp_md_entry_v1 lcm_entries[0];
950 } __attribute__((packed));
952 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
954 if (stripes == (__u16)-1)
957 if (lmm_magic == LOV_USER_MAGIC_V1)
958 return sizeof(struct lov_user_md_v1) +
959 stripes * sizeof(struct lov_user_ost_data_v1);
960 return sizeof(struct lov_user_md_v3) +
961 stripes * sizeof(struct lov_user_ost_data_v1);
964 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
965 * use this. It is unsafe to #define those values in this header as it
966 * is possible the application has already #included <sys/stat.h>. */
967 #ifdef HAVE_LOV_USER_MDS_DATA
968 #define lov_user_mds_data lov_user_mds_data_v2
969 struct lov_user_mds_data_v1 {
970 lstat_t lmd_st; /* MDS stat struct */
971 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
972 } __attribute__((packed));
974 struct lov_user_mds_data_v2 {
975 struct lu_fid lmd_fid; /* Lustre FID */
976 lstatx_t lmd_stx; /* MDS statx struct */
977 __u64 lmd_flags; /* MDS stat flags */
978 __u32 lmd_lmmsize; /* LOV EA size */
979 __u32 lmd_padding; /* unused */
980 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
981 } __attribute__((packed));
984 struct lmv_user_mds_data {
985 struct lu_fid lum_fid;
988 } __attribute__((packed, __may_alias__));
991 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
992 LMV_HASH_TYPE_ALL_CHARS = 1,
993 LMV_HASH_TYPE_FNV_1A_64 = 2,
994 LMV_HASH_TYPE_CRUSH = 3,
998 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
1000 #define LMV_HASH_NAME_ALL_CHARS "all_char"
1001 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
1002 #define LMV_HASH_NAME_CRUSH "crush"
1004 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1005 * and the higher part will be the flag to indicate the status of object,
1006 * for example the object is being migrated. And the hash function
1007 * might be interpreted differently with different flags. */
1008 #define LMV_HASH_TYPE_MASK 0x0000ffff
1010 static inline bool lmv_is_known_hash_type(__u32 type)
1012 return (type & LMV_HASH_TYPE_MASK) == LMV_HASH_TYPE_FNV_1A_64 ||
1013 (type & LMV_HASH_TYPE_MASK) == LMV_HASH_TYPE_ALL_CHARS ||
1014 (type & LMV_HASH_TYPE_MASK) == LMV_HASH_TYPE_CRUSH;
1017 /* The striped directory has ever lost its master LMV EA, then LFSCK
1018 * re-generated it. This flag is used to indicate such case. It is an
1020 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
1022 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1023 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1025 #define LMV_HASH_FLAG_LAYOUT_CHANGE LMV_HASH_FLAG_MIGRATION
1027 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
1029 struct lustre_foreign_type {
1031 const char *lft_name;
1035 * LOV/LMV foreign types
1037 enum lustre_foreign_types {
1038 LU_FOREIGN_TYPE_NONE = 0,
1039 LU_FOREIGN_TYPE_DAOS = 0xda05,
1040 /* must be the max/last one */
1041 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
1044 extern struct lustre_foreign_type lu_foreign_types[];
1046 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
1047 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
1048 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
1049 #define lmv_user_md lmv_user_md_v1
1050 struct lmv_user_md_v1 {
1051 __u32 lum_magic; /* must be the first field */
1052 __u32 lum_stripe_count; /* dirstripe count */
1053 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
1054 __u32 lum_hash_type; /* Dir stripe policy */
1055 __u32 lum_type; /* LMV type: default */
1059 char lum_pool_name[LOV_MAXPOOLNAME + 1];
1060 struct lmv_user_mds_data lum_objects[0];
1061 } __attribute__((packed));
1063 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
1065 if (size <= sizeof(struct lmv_user_md))
1068 size -= sizeof(struct lmv_user_md);
1069 return (size + sizeof(struct lmv_user_mds_data) - 1) /
1070 sizeof(struct lmv_user_mds_data);
1074 * NB, historically default layout didn't set type, but use XATTR name to differ
1075 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
1076 * and still use the same method.
1079 LMV_TYPE_DEFAULT = 0x0000,
1082 static inline int lmv_user_md_size(int stripes, int lmm_magic)
1084 int size = sizeof(struct lmv_user_md);
1086 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
1087 size += stripes * sizeof(struct lmv_user_mds_data);
1092 struct ll_recreate_obj {
1098 __u64 id; /* holds object id */
1099 __u32 generation; /* holds object generation */
1100 __u32 f_type; /* holds object type or stripe idx when passing it to
1101 * OST for saving into EA. */
1106 char uuid[UUID_MAX];
1109 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
1110 const struct obd_uuid *u2)
1112 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
1115 static inline int obd_uuid_empty(struct obd_uuid *uuid)
1117 return uuid->uuid[0] == '\0';
1120 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
1122 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
1123 uuid->uuid[sizeof(*uuid) - 1] = '\0';
1126 /* For printf's only, make sure uuid is terminated */
1127 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
1132 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
1133 /* Obviously not safe, but for printfs, no real harm done...
1134 we're always null-terminated, even in a race. */
1135 static char temp[sizeof(*uuid)];
1136 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
1137 temp[sizeof(*uuid) - 1] = '\0';
1140 return (char *)(uuid->uuid);
1143 #define LUSTRE_MAXFSNAME 8
1144 #define LUSTRE_MAXINSTANCE 16
1146 /* Extract fsname from uuid (or target name) of a target
1147 e.g. (myfs-OST0007_UUID -> myfs)
1148 see also deuuidify. */
1149 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1153 strncpy(buf, uuid, buflen - 1);
1154 buf[buflen - 1] = '\0';
1155 p = strrchr(buf, '-');
1160 /* printf display format for Lustre FIDs
1161 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1162 #define FID_NOBRACE_LEN 40
1163 #define FID_LEN (FID_NOBRACE_LEN + 2)
1164 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1165 #define DFID "[" DFID_NOBRACE "]"
1166 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1168 /* scanf input parse format for fids in DFID_NOBRACE format
1169 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1170 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1171 #define SFID "0x%llx:0x%x:0x%x"
1172 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1174 /********* Quotas **********/
1176 #define LUSTRE_QUOTABLOCK_BITS 10
1177 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1179 static inline __u64 lustre_stoqb(__kernel_size_t space)
1181 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1184 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1185 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1186 #define Q_GETOINFO 0x800102 /* get obd quota info */
1187 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1188 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1190 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1191 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1192 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1193 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1194 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1195 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1196 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1197 /* lustre-specific control commands */
1198 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1199 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1200 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1201 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1203 /* In the current Lustre implementation, the grace time is either the time
1204 * or the timestamp to be used after some quota ID exceeds the soft limt,
1205 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1207 #define LQUOTA_GRACE_BITS 48
1208 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1209 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1210 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1211 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1212 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1214 /* special grace time, only notify the user when its quota is over soft limit
1215 * but doesn't block new writes until the hard limit is reached. */
1216 #define NOTIFY_GRACE "notify"
1217 #define NOTIFY_GRACE_TIME LQUOTA_GRACE_MASK
1219 /* different quota flags */
1221 /* the default quota flag, the corresponding quota ID will use the default
1222 * quota setting, the hardlimit and softlimit of its quota record in the global
1223 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1224 * and high 16 bits will contain this flag (see above comment).
1226 #define LQUOTA_FLAG_DEFAULT 0x0001
1228 #define ALLQUOTA 255 /* set all quota */
1229 static inline const char *qtype_name(int qtype)
1242 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1243 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1246 #define N_PERMS_MAX 64
1248 struct perm_downcall_data {
1254 struct identity_downcall_data {
1261 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1262 __u32 idd_groups[0];
1265 struct sepol_downcall_data {
1267 __kernel_time_t sdd_sepol_mtime;
1268 __u16 sdd_sepol_len;
1272 #ifdef NEED_QUOTA_DEFS
1274 #define QIF_BLIMITS 1
1276 #define QIF_ILIMITS 4
1277 #define QIF_INODES 8
1278 #define QIF_BTIME 16
1279 #define QIF_ITIME 32
1280 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1281 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1282 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1283 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1286 #endif /* !__KERNEL__ */
1288 /* lustre volatile file support
1289 * file name header: ".^L^S^T^R:volatile"
1291 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1292 #define LUSTRE_VOLATILE_HDR_LEN 14
1294 enum lustre_quota_version {
1298 /* XXX: same as if_dqinfo struct in kernel */
1306 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1308 __u64 dqb_bhardlimit;
1309 __u64 dqb_bsoftlimit;
1311 __u64 dqb_ihardlimit;
1312 __u64 dqb_isoftlimit;
1313 __u64 dqb_curinodes;
1327 struct if_quotactl {
1334 struct obd_dqinfo qc_dqinfo;
1335 struct obd_dqblk qc_dqblk;
1337 struct obd_uuid obd_uuid;
1340 /* swap layout flags */
1341 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1342 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1343 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1344 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1345 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1347 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1348 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1349 struct lustre_swap_layouts {
1357 /** Bit-mask of valid attributes */
1358 /* The LA_* flags are written to disk as part of the ChangeLog records
1359 * so they are part of the on-disk and network protocol, and cannot be changed.
1360 * Only the first 12 bits are currently saved.
1363 LA_ATIME = 1 << 0, /* 0x00001 */
1364 LA_MTIME = 1 << 1, /* 0x00002 */
1365 LA_CTIME = 1 << 2, /* 0x00004 */
1366 LA_SIZE = 1 << 3, /* 0x00008 */
1367 LA_MODE = 1 << 4, /* 0x00010 */
1368 LA_UID = 1 << 5, /* 0x00020 */
1369 LA_GID = 1 << 6, /* 0x00040 */
1370 LA_BLOCKS = 1 << 7, /* 0x00080 */
1371 LA_TYPE = 1 << 8, /* 0x00100 */
1372 LA_FLAGS = 1 << 9, /* 0x00200 */
1373 LA_NLINK = 1 << 10, /* 0x00400 */
1374 LA_RDEV = 1 << 11, /* 0x00800 */
1375 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1376 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1377 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1378 LA_PROJID = 1 << 15, /* 0x08000 */
1379 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1380 LA_LSIZE = 1 << 17, /* 0x20000 */
1381 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1383 * Attributes must be transmitted to OST objects
1385 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1388 #define MDS_FMODE_READ 00000001
1389 #define MDS_FMODE_WRITE 00000002
1391 #define MDS_FMODE_CLOSED 00000000
1392 #define MDS_FMODE_EXEC 00000004
1393 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1394 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1395 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1397 #define MDS_OPEN_CREATED 00000010
1398 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1400 #define MDS_OPEN_CREAT 00000100
1401 #define MDS_OPEN_EXCL 00000200
1402 #define MDS_OPEN_TRUNC 00001000
1403 #define MDS_OPEN_APPEND 00002000
1404 #define MDS_OPEN_SYNC 00010000
1405 #define MDS_OPEN_DIRECTORY 00200000
1407 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1408 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1409 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1410 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1411 * We do not support JOIN FILE
1412 * anymore, reserve this flags
1413 * just for preventing such bit
1416 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1417 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1418 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1419 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1420 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1422 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1424 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1425 * delegation, succeed if it's not
1426 * being opened with conflict mode.
1428 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1430 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1431 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1432 * for newly created file */
1434 /* lustre internal open flags, which should not be set from user space */
1435 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1436 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1437 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1438 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1442 /********* Changelogs **********/
1443 /** Changelog record types */
1444 enum changelog_rec_type {
1447 CL_CREATE = 1, /* namespace */
1448 CL_MKDIR = 2, /* namespace */
1449 CL_HARDLINK = 3, /* namespace */
1450 CL_SOFTLINK = 4, /* namespace */
1451 CL_MKNOD = 5, /* namespace */
1452 CL_UNLINK = 6, /* namespace */
1453 CL_RMDIR = 7, /* namespace */
1454 CL_RENAME = 8, /* namespace */
1455 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1456 CL_OPEN = 10, /* not currently used */
1457 CL_CLOSE = 11, /* may be written to log only with mtime change */
1458 CL_LAYOUT = 12, /* file layout/striping modified */
1462 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1463 CL_HSM = 16, /* HSM specific events, see flags */
1464 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1468 CL_FLRW = 21, /* FLR: file was firstly written */
1469 CL_RESYNC = 22, /* FLR: file was resync-ed */
1471 CL_DN_OPEN = 24, /* denied open */
1475 static inline const char *changelog_type2str(int type) {
1476 static const char *changelog_str[] = {
1477 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1478 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1479 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1480 "FLRW", "RESYNC","GXATR", "NOPEN",
1483 if (type >= 0 && type < CL_LAST)
1484 return changelog_str[type];
1488 /* 12 bits of per-record data can be stored in the bottom of the flags */
1489 #define CLF_FLAGSHIFT 12
1490 enum changelog_rec_flags {
1491 CLF_VERSION = 0x1000,
1492 CLF_RENAME = 0x2000,
1494 CLF_EXTRA_FLAGS = 0x8000,
1495 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1497 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1498 CLF_VERMASK = ~CLF_FLAGMASK,
1502 /* Anything under the flagmask may be per-type (if desired) */
1503 /* Flags for unlink */
1504 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1505 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1506 /* HSM cleaning needed */
1507 /* Flags for rename */
1508 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1510 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1511 * has an archive in backend */
1514 /* 12b used (from high weight to low weight):
1519 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1520 #define CLF_HSM_ERR_H 6
1521 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1522 #define CLF_HSM_EVENT_H 9
1523 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1524 #define CLF_HSM_FLAG_H 11
1525 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1526 #define CLF_HSM_SPARE_H 15
1527 #define CLF_HSM_LAST 15
1529 /* Remove bits higher than _h, then extract the value
1530 * between _h and _l by shifting lower weigth to bit 0. */
1531 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1532 >> (CLF_HSM_LAST - _h + _l))
1534 #define CLF_HSM_SUCCESS 0x00
1535 #define CLF_HSM_MAXERROR 0x7E
1536 #define CLF_HSM_ERROVERFLOW 0x7F
1538 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1540 /* 3 bits field => 8 values allowed */
1552 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1554 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1558 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1561 *clf_flags = (enum changelog_rec_flags)
1562 (*clf_flags | (he << CLF_HSM_EVENT_L));
1565 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1567 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1570 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1573 *clf_flags = (enum changelog_rec_flags)
1574 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1577 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1579 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1582 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1585 *clf_flags = (enum changelog_rec_flags)
1586 (*clf_flags | (error << CLF_HSM_ERR_L));
1589 enum changelog_rec_extra_flags {
1591 CLFE_UIDGID = 0x0001,
1594 CLFE_XATTR = 0x0008,
1595 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1598 enum changelog_send_flag {
1599 /* Not yet implemented */
1600 CHANGELOG_FLAG_FOLLOW = 0x01,
1601 /* Blocking IO makes sense in case of slow user parsing of the records,
1602 * but it also prevents us from cleaning up if the records are not
1604 CHANGELOG_FLAG_BLOCK = 0x02,
1605 /* Pack jobid into the changelog records if available. */
1606 CHANGELOG_FLAG_JOBID = 0x04,
1607 /* Pack additional flag bits into the changelog record */
1608 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1611 enum changelog_send_extra_flag {
1612 /* Pack uid/gid into the changelog record */
1613 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1614 /* Pack nid into the changelog record */
1615 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1616 /* Pack open mode into the changelog record */
1617 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1618 /* Pack xattr name into the changelog record */
1619 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1622 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1623 changelog_rec_offset(CLF_SUPPORTED, \
1626 /* 31 usable bytes string + null terminator. */
1627 #define LUSTRE_JOBID_SIZE 32
1629 /* This is the minimal changelog record. It can contain extensions
1630 * such as rename fields or process jobid. Its exact content is described
1631 * by the cr_flags and cr_extra_flags.
1633 * Extensions are packed in the same order as their corresponding flags,
1634 * then in the same order as their corresponding extra flags.
1636 struct changelog_rec {
1638 __u16 cr_flags; /**< \a changelog_rec_flags */
1639 __u32 cr_type; /**< \a changelog_rec_type */
1640 __u64 cr_index; /**< changelog record number */
1641 __u64 cr_prev; /**< last index for this target fid */
1644 struct lu_fid cr_tfid; /**< target fid */
1645 __u32 cr_markerflags; /**< CL_MARK flags */
1647 struct lu_fid cr_pfid; /**< parent fid */
1648 } __attribute__ ((packed));
1650 /* Changelog extension for RENAME. */
1651 struct changelog_ext_rename {
1652 struct lu_fid cr_sfid; /**< source fid, or zero */
1653 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1656 /* Changelog extension to include JOBID. */
1657 struct changelog_ext_jobid {
1658 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1661 /* Changelog extension to include additional flags. */
1662 struct changelog_ext_extra_flags {
1663 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1666 /* Changelog extra extension to include UID/GID. */
1667 struct changelog_ext_uidgid {
1672 /* Changelog extra extension to include NID. */
1673 struct changelog_ext_nid {
1674 /* have __u64 instead of lnet_nid_t type for use by client api */
1676 /* for use when IPv6 support is added */
1681 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1682 struct changelog_ext_openmode {
1686 /* Changelog extra extension to include xattr */
1687 struct changelog_ext_xattr {
1688 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1691 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1692 const struct changelog_rec *rec);
1694 static inline __kernel_size_t changelog_rec_offset(enum changelog_rec_flags crf,
1695 enum changelog_rec_extra_flags cref)
1697 __kernel_size_t size = sizeof(struct changelog_rec);
1699 if (crf & CLF_RENAME)
1700 size += sizeof(struct changelog_ext_rename);
1702 if (crf & CLF_JOBID)
1703 size += sizeof(struct changelog_ext_jobid);
1705 if (crf & CLF_EXTRA_FLAGS) {
1706 size += sizeof(struct changelog_ext_extra_flags);
1707 if (cref & CLFE_UIDGID)
1708 size += sizeof(struct changelog_ext_uidgid);
1709 if (cref & CLFE_NID)
1710 size += sizeof(struct changelog_ext_nid);
1711 if (cref & CLFE_OPEN)
1712 size += sizeof(struct changelog_ext_openmode);
1713 if (cref & CLFE_XATTR)
1714 size += sizeof(struct changelog_ext_xattr);
1720 static inline __kernel_size_t changelog_rec_size(const struct changelog_rec *rec)
1722 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1724 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1725 cref = (enum changelog_rec_extra_flags)
1726 changelog_rec_extra_flags(rec)->cr_extra_flags;
1728 return changelog_rec_offset(
1729 (enum changelog_rec_flags)rec->cr_flags, cref);
1732 static inline __kernel_size_t changelog_rec_varsize(const struct changelog_rec *rec)
1734 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1738 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1740 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1741 (rec->cr_flags & CLF_VERSION);
1743 return (struct changelog_ext_rename *)((char *)rec +
1744 changelog_rec_offset(crf,
1748 /* The jobid follows the rename extension, if present */
1750 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1752 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1753 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1755 return (struct changelog_ext_jobid *)((char *)rec +
1756 changelog_rec_offset(crf,
1760 /* The additional flags follow the rename and jobid extensions, if present */
1762 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1763 const struct changelog_rec *rec)
1765 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1766 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1768 return (struct changelog_ext_extra_flags *)((char *)rec +
1769 changelog_rec_offset(crf,
1773 /* The uid/gid is the first extra extension */
1775 struct changelog_ext_uidgid *changelog_rec_uidgid(
1776 const struct changelog_rec *rec)
1778 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1780 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1782 return (struct changelog_ext_uidgid *)((char *)rec +
1783 changelog_rec_offset(crf,
1787 /* The nid is the second extra extension */
1789 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1791 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1793 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1794 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1796 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1797 cref = (enum changelog_rec_extra_flags)
1798 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1801 return (struct changelog_ext_nid *)((char *)rec +
1802 changelog_rec_offset(crf, cref));
1805 /* The OPEN mode is the third extra extension */
1807 struct changelog_ext_openmode *changelog_rec_openmode(
1808 const struct changelog_rec *rec)
1810 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1812 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1813 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1815 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1816 cref = (enum changelog_rec_extra_flags)
1817 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1818 (CLFE_UIDGID | CLFE_NID));
1821 return (struct changelog_ext_openmode *)((char *)rec +
1822 changelog_rec_offset(crf, cref));
1825 /* The xattr name is the fourth extra extension */
1827 struct changelog_ext_xattr *changelog_rec_xattr(
1828 const struct changelog_rec *rec)
1830 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1832 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1833 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1835 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1836 cref = (enum changelog_rec_extra_flags)
1837 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1838 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
1840 return (struct changelog_ext_xattr *)((char *)rec +
1841 changelog_rec_offset(crf, cref));
1844 /* The name follows the rename, jobid and extra flags extns, if present */
1845 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1847 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1849 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1850 cref = (enum changelog_rec_extra_flags)
1851 changelog_rec_extra_flags(rec)->cr_extra_flags;
1853 return (char *)rec + changelog_rec_offset(
1854 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
1855 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
1858 static inline __kernel_size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1860 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1863 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1865 char *cr_name = changelog_rec_name(rec);
1867 return cr_name + strlen(cr_name) + 1;
1871 * Remap a record to the desired format as specified by the crf flags.
1872 * The record must be big enough to contain the final remapped version.
1873 * Superfluous extension fields are removed and missing ones are added
1874 * and zeroed. The flags of the record are updated accordingly.
1876 * The jobid and rename extensions can be added to a record, to match the
1877 * format an application expects, typically. In this case, the newly added
1878 * fields will be zeroed.
1879 * The Jobid field can be removed, to guarantee compatibility with older
1880 * clients that don't expect this field in the records they process.
1882 * The following assumptions are being made:
1883 * - CLF_RENAME will not be removed
1884 * - CLF_JOBID will not be added without CLF_RENAME being added too
1885 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1887 * @param[in,out] rec The record to remap.
1888 * @param[in] crf_wanted Flags describing the desired extensions.
1889 * @param[in] cref_want Flags describing the desired extra extensions.
1891 static inline void changelog_remap_rec(struct changelog_rec *rec,
1892 enum changelog_rec_flags crf_wanted,
1893 enum changelog_rec_extra_flags cref_want)
1895 char *xattr_mov = NULL;
1896 char *omd_mov = NULL;
1897 char *nid_mov = NULL;
1898 char *uidgid_mov = NULL;
1902 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1904 crf_wanted = (enum changelog_rec_flags)
1905 (crf_wanted & CLF_SUPPORTED);
1906 cref_want = (enum changelog_rec_extra_flags)
1907 (cref_want & CLFE_SUPPORTED);
1909 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1910 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1911 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1912 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1918 /* First move the variable-length name field */
1919 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1920 changelog_rec_name(rec), rec->cr_namelen);
1922 /* Locations of extensions in the remapped record */
1923 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1924 xattr_mov = (char *)rec +
1925 changelog_rec_offset(
1926 (enum changelog_rec_flags)
1927 (crf_wanted & CLF_SUPPORTED),
1928 (enum changelog_rec_extra_flags)
1929 (cref_want & ~CLFE_XATTR));
1930 omd_mov = (char *)rec +
1931 changelog_rec_offset(
1932 (enum changelog_rec_flags)
1933 (crf_wanted & CLF_SUPPORTED),
1934 (enum changelog_rec_extra_flags)
1935 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
1936 nid_mov = (char *)rec +
1937 changelog_rec_offset(
1938 (enum changelog_rec_flags)
1939 (crf_wanted & CLF_SUPPORTED),
1940 (enum changelog_rec_extra_flags)
1942 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
1943 uidgid_mov = (char *)rec +
1944 changelog_rec_offset(
1945 (enum changelog_rec_flags)
1946 (crf_wanted & CLF_SUPPORTED),
1947 (enum changelog_rec_extra_flags)
1948 (cref_want & ~(CLFE_UIDGID |
1952 cref = (enum changelog_rec_extra_flags)
1953 changelog_rec_extra_flags(rec)->cr_extra_flags;
1956 ef_mov = (char *)rec +
1957 changelog_rec_offset(
1958 (enum changelog_rec_flags)
1959 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
1960 jid_mov = (char *)rec +
1961 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
1962 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
1964 rnm_mov = (char *)rec +
1965 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
1971 /* Move the extension fields to the desired positions */
1972 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1973 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1974 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1975 memmove(xattr_mov, changelog_rec_xattr(rec),
1976 sizeof(struct changelog_ext_xattr));
1978 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1979 memmove(omd_mov, changelog_rec_openmode(rec),
1980 sizeof(struct changelog_ext_openmode));
1982 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1983 memmove(nid_mov, changelog_rec_nid(rec),
1984 sizeof(struct changelog_ext_nid));
1986 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1987 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1988 sizeof(struct changelog_ext_uidgid));
1990 memmove(ef_mov, changelog_rec_extra_flags(rec),
1991 sizeof(struct changelog_ext_extra_flags));
1994 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1995 memmove(jid_mov, changelog_rec_jobid(rec),
1996 sizeof(struct changelog_ext_jobid));
1998 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1999 memmove(rnm_mov, changelog_rec_rename(rec),
2000 sizeof(struct changelog_ext_rename));
2002 /* Clear newly added fields */
2003 if (xattr_mov && (cref_want & CLFE_XATTR) &&
2004 !(cref & CLFE_XATTR))
2005 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
2007 if (omd_mov && (cref_want & CLFE_OPEN) &&
2008 !(cref & CLFE_OPEN))
2009 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
2011 if (nid_mov && (cref_want & CLFE_NID) &&
2013 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
2015 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
2016 !(cref & CLFE_UIDGID))
2017 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
2019 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2020 !(rec->cr_flags & CLF_EXTRA_FLAGS))
2021 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
2023 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
2024 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
2026 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
2027 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
2029 /* Update the record's flags accordingly */
2030 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
2031 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2032 changelog_rec_extra_flags(rec)->cr_extra_flags =
2033 changelog_rec_extra_flags(rec)->cr_extra_flags |
2037 enum changelog_message_type {
2038 CL_RECORD = 10, /* message is a changelog_rec */
2039 CL_EOF = 11, /* at end of current changelog */
2042 /********* Misc **********/
2044 struct ioc_data_version {
2046 __u32 idv_layout_version; /* FLR: layout version for OST objects */
2047 __u32 idv_flags; /* enum ioc_data_version_flags */
2050 enum ioc_data_version_flags {
2051 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
2052 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
2056 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
2059 #define dot_lustre_name ".lustre"
2062 /********* HSM **********/
2064 /** HSM per-file state
2065 * See HSM_FLAGS below.
2068 HS_NONE = 0x00000000,
2069 HS_EXISTS = 0x00000001,
2070 HS_DIRTY = 0x00000002,
2071 HS_RELEASED = 0x00000004,
2072 HS_ARCHIVED = 0x00000008,
2073 HS_NORELEASE = 0x00000010,
2074 HS_NOARCHIVE = 0x00000020,
2075 HS_LOST = 0x00000040,
2078 /* HSM user-setable flags. */
2079 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
2081 /* Other HSM flags. */
2082 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
2085 * All HSM-related possible flags that could be applied to a file.
2086 * This should be kept in sync with hsm_states.
2088 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
2091 * HSM request progress state
2093 enum hsm_progress_states {
2100 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
2103 case HPS_WAITING: return "waiting";
2104 case HPS_RUNNING: return "running";
2105 case HPS_DONE: return "done";
2106 default: return "unknown";
2113 } __attribute__((packed));
2116 * Current HSM states of a Lustre file.
2118 * This structure purpose is to be sent to user-space mainly. It describes the
2119 * current HSM flags and in-progress action.
2121 struct hsm_user_state {
2122 /** Current HSM states, from enum hsm_states. */
2124 __u32 hus_archive_id;
2125 /** The current undergoing action, if there is one */
2126 __u32 hus_in_progress_state;
2127 __u32 hus_in_progress_action;
2128 struct hsm_extent hus_in_progress_location;
2129 char hus_extended_info[];
2132 struct hsm_state_set_ioc {
2133 struct lu_fid hssi_fid;
2135 __u64 hssi_clearmask;
2139 * This structure describes the current in-progress action for a file.
2140 * it is retuned to user space and send over the wire
2142 struct hsm_current_action {
2143 /** The current undergoing action, if there is one */
2144 /* state is one of hsm_progress_states */
2146 /* action is one of hsm_user_action */
2148 struct hsm_extent hca_location;
2151 /***** HSM user requests ******/
2152 /* User-generated (lfs/ioctl) request types */
2153 enum hsm_user_action {
2154 HUA_NONE = 1, /* no action (noop) */
2155 HUA_ARCHIVE = 10, /* copy to hsm */
2156 HUA_RESTORE = 11, /* prestage */
2157 HUA_RELEASE = 12, /* drop ost objects */
2158 HUA_REMOVE = 13, /* remove from archive */
2159 HUA_CANCEL = 14 /* cancel a request */
2162 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2165 case HUA_NONE: return "NOOP";
2166 case HUA_ARCHIVE: return "ARCHIVE";
2167 case HUA_RESTORE: return "RESTORE";
2168 case HUA_RELEASE: return "RELEASE";
2169 case HUA_REMOVE: return "REMOVE";
2170 case HUA_CANCEL: return "CANCEL";
2171 default: return "UNKNOWN";
2176 * List of hr_flags (bit field)
2178 #define HSM_FORCE_ACTION 0x0001
2179 /* used by CT, cannot be set by user */
2180 #define HSM_GHOST_COPY 0x0002
2183 * Contains all the fixed part of struct hsm_user_request.
2186 struct hsm_request {
2187 __u32 hr_action; /* enum hsm_user_action */
2188 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2189 __u64 hr_flags; /* request flags */
2190 __u32 hr_itemcount; /* item count in hur_user_item vector */
2194 struct hsm_user_item {
2195 struct lu_fid hui_fid;
2196 struct hsm_extent hui_extent;
2197 } __attribute__((packed));
2199 struct hsm_user_request {
2200 struct hsm_request hur_request;
2201 struct hsm_user_item hur_user_item[0];
2202 /* extra data blob at end of struct (after all
2203 * hur_user_items), only use helpers to access it
2205 } __attribute__((packed));
2207 /** Return pointer to data field in a hsm user request */
2208 static inline void *hur_data(struct hsm_user_request *hur)
2210 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2214 * Compute the current length of the provided hsm_user_request. This returns -1
2215 * instead of an errno because __kernel_ssize_t is defined to be only
2218 * return -1 on bounds check error.
2220 static inline __kernel_size_t hur_len(struct hsm_user_request *hur)
2224 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2225 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2226 (__u64)hur->hur_request.hr_itemcount *
2227 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2229 if ((__kernel_ssize_t)size < 0)
2235 /****** HSM RPCs to copytool *****/
2236 /* Message types the copytool may receive */
2237 enum hsm_message_type {
2238 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2241 /* Actions the copytool may be instructed to take for a given action_item */
2242 enum hsm_copytool_action {
2243 HSMA_NONE = 10, /* no action */
2244 HSMA_ARCHIVE = 20, /* arbitrary offset */
2250 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2253 case HSMA_NONE: return "NOOP";
2254 case HSMA_ARCHIVE: return "ARCHIVE";
2255 case HSMA_RESTORE: return "RESTORE";
2256 case HSMA_REMOVE: return "REMOVE";
2257 case HSMA_CANCEL: return "CANCEL";
2258 default: return "UNKNOWN";
2262 /* Copytool item action description */
2263 struct hsm_action_item {
2264 __u32 hai_len; /* valid size of this struct */
2265 __u32 hai_action; /* hsm_copytool_action, but use known size */
2266 struct lu_fid hai_fid; /* Lustre FID to operate on */
2267 struct lu_fid hai_dfid; /* fid used for data access */
2268 struct hsm_extent hai_extent; /* byte range to operate on */
2269 __u64 hai_cookie; /* action cookie from coordinator */
2270 __u64 hai_gid; /* grouplock id */
2271 char hai_data[0]; /* variable length */
2272 } __attribute__((packed));
2275 * helper function which print in hexa the first bytes of
2278 * \param hai [IN] record to print
2279 * \param buffer [IN,OUT] buffer to write the hex string to
2280 * \param len [IN] max buffer length
2284 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2285 char *buffer, __kernel_size_t len)
2292 data_len = hai->hai_len - sizeof(*hai);
2293 for (i = 0; (i < data_len) && (len > 2); i++) {
2294 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2304 /* Copytool action list */
2305 #define HAL_VERSION 1
2306 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2307 struct hsm_action_list {
2309 __u32 hal_count; /* number of hai's to follow */
2310 __u64 hal_compound_id; /* returned by coordinator, ignored */
2312 __u32 hal_archive_id; /* which archive backend */
2314 char hal_fsname[0]; /* null-terminated */
2315 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2316 boundaries. See hai_zero */
2317 } __attribute__((packed));
2319 /* Return pointer to first hai in action list */
2320 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2322 __kernel_size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2324 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2327 /* Return pointer to next hai */
2328 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2330 __kernel_size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2332 return (struct hsm_action_item *)((char *)hai + offset);
2335 /* Return size of an hsm_action_list */
2336 static inline __kernel_size_t hal_size(struct hsm_action_list *hal)
2340 struct hsm_action_item *hai;
2342 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2343 hai = hai_first(hal);
2344 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2345 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2351 * describe the attributes to be set on imported file
2353 struct hsm_user_import {
2362 __u32 hui_archive_id;
2365 /* Copytool progress reporting */
2366 #define HP_FLAG_COMPLETED 0x01
2367 #define HP_FLAG_RETRY 0x02
2369 struct hsm_progress {
2370 struct lu_fid hp_fid;
2372 struct hsm_extent hp_extent;
2374 __u16 hp_errval; /* positive val */
2379 __u64 hc_data_version;
2381 __u16 hc_errval; /* positive val */
2383 struct hsm_action_item hc_hai;
2387 enum llapi_json_types {
2388 LLAPI_JSON_INTEGER = 1,
2394 struct llapi_json_item {
2403 struct llapi_json_item *lji_next;
2406 struct llapi_json_item_list {
2407 int ljil_item_count;
2408 struct llapi_json_item *ljil_items;
2411 enum lu_ladvise_type {
2412 LU_LADVISE_INVALID = 0,
2413 LU_LADVISE_WILLREAD = 1,
2414 LU_LADVISE_DONTNEED = 2,
2415 LU_LADVISE_LOCKNOEXPAND = 3,
2416 LU_LADVISE_LOCKAHEAD = 4,
2420 #define LU_LADVISE_NAMES { \
2421 [LU_LADVISE_WILLREAD] = "willread", \
2422 [LU_LADVISE_DONTNEED] = "dontneed", \
2423 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2424 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2427 /* This is the userspace argument for ladvise. It is currently the same as
2428 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2429 * may need info which is only used locally. */
2430 struct llapi_lu_ladvise {
2431 __u16 lla_advice; /* advice type */
2432 __u16 lla_value1; /* values for different advice types */
2434 __u64 lla_start; /* first byte of extent for advice */
2435 __u64 lla_end; /* last byte of extent for advice */
2441 LF_ASYNC = 0x00000001,
2442 LF_UNSET = 0x00000002,
2445 #define LADVISE_MAGIC 0x1ADF1CE0
2446 /* Masks of valid flags for each advice */
2447 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2448 /* Flags valid for all advices not explicitly specified */
2449 #define LF_DEFAULT_MASK LF_ASYNC
2451 #define LF_MASK (LF_ASYNC | LF_UNSET)
2453 #define lla_lockahead_mode lla_value1
2454 #define lla_peradvice_flags lla_value2
2455 #define lla_lockahead_result lla_value3
2457 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2458 * is used on the wire. It is defined separately as we may need info which is
2459 * only used locally. */
2460 struct llapi_ladvise_hdr {
2461 __u32 lah_magic; /* LADVISE_MAGIC */
2462 __u32 lah_count; /* number of advices */
2463 __u64 lah_flags; /* from enum ladvise_flag */
2464 __u32 lah_value1; /* unused */
2465 __u32 lah_value2; /* unused */
2466 __u64 lah_value3; /* unused */
2467 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2470 #define LAH_COUNT_MAX (1024)
2474 SK_CRYPT_INVALID = -1,
2476 SK_CRYPT_AES256_CTR = 1,
2480 SK_HMAC_INVALID = -1,
2486 struct sk_crypt_type {
2487 const char *sct_name;
2491 struct sk_hmac_type {
2492 const char *sht_name;
2496 enum lock_mode_user {
2502 #define LOCK_MODE_NAMES { \
2503 [MODE_READ_USER] = "READ",\
2504 [MODE_WRITE_USER] = "WRITE"\
2507 enum lockahead_results {
2508 LLA_RESULT_SENT = 0,
2509 LLA_RESULT_DIFFERENT,
2513 enum lu_heat_flag_bit {
2514 LU_HEAT_FLAG_BIT_INVALID = 0,
2515 LU_HEAT_FLAG_BIT_OFF,
2516 LU_HEAT_FLAG_BIT_CLEAR,
2520 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2521 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2524 enum obd_heat_type {
2525 OBD_HEAT_READSAMPLE = 0,
2526 OBD_HEAT_WRITESAMPLE = 1,
2527 OBD_HEAT_READBYTE = 2,
2528 OBD_HEAT_WRITEBYTE = 3,
2532 #define LU_HEAT_NAMES { \
2533 [OBD_HEAT_READSAMPLE] = "readsample", \
2534 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2535 [OBD_HEAT_READBYTE] = "readbyte", \
2536 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2551 static inline const char *pcc_type2string(enum lu_pcc_type type)
2556 case LU_PCC_READWRITE:
2563 struct lu_pcc_attach {
2564 __u32 pcca_type; /* PCC type */
2565 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2568 enum lu_pcc_detach_opts {
2569 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2570 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2573 struct lu_pcc_detach_fid {
2574 /* fid of the file to detach */
2575 struct lu_fid pccd_fid;
2579 struct lu_pcc_detach {
2583 enum lu_pcc_state_flags {
2584 PCC_STATE_FL_NONE = 0x0,
2585 /* The inode attr is cached locally */
2586 PCC_STATE_FL_ATTR_VALID = 0x01,
2587 /* The file is being attached into PCC */
2588 PCC_STATE_FL_ATTACHING = 0x02,
2591 struct lu_pcc_state {
2592 __u32 pccs_type; /* enum lu_pcc_type */
2593 __u32 pccs_open_count;
2594 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2596 char pccs_path[PATH_MAX];
2601 /* make header's size equal lu_fid */
2604 struct lu_fid fa_fids[0];
2606 #define OBD_MAX_FIDS_IN_ARRAY 4096
2608 #if defined(__cplusplus)
2612 /** @} lustreuser */
2614 #endif /* _LUSTRE_USER_H */