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/
31 * lustre/include/lustre/lustre_user.h
33 * Lustre public user-space interface definitions.
36 #ifndef _LUSTRE_USER_H
37 #define _LUSTRE_USER_H
39 /** \defgroup lustreuser lustreuser
44 # define __USE_ISOC99 1
46 # include <stdio.h> /* snprintf() */
47 # include <sys/stat.h>
50 # define __USE_XOPEN2K8 1
51 # define FILEID_LUSTRE 0x97 /* for name_to_handle_at() (and llapi_fd2fid()) */
52 #endif /* !__KERNEL__ */
55 #include <linux/limits.h>
56 #include <linux/kernel.h>
57 #include <linux/string.h>
58 #include <linux/quota.h>
59 #include <linux/types.h>
60 #include <linux/unistd.h>
61 #include <linux/lustre/lustre_fiemap.h>
62 #include <linux/lustre/lustre_ver.h>
64 #if defined(__cplusplus)
68 #ifdef __STRICT_ANSI__
69 #define typeof __typeof__
73 * This is a temporary solution of adding quota type.
74 * Should be removed as soon as system header is updated.
77 #define LL_MAXQUOTAS 3
79 #define INITQFNAMES { \
80 "user", /* USRQUOTA */ \
81 "group", /* GRPQUOTA */ \
82 "project", /* PRJQUOTA */ \
96 * We need to always use 64bit version because the structure
97 * is shared across entire cluster where 32bit and 64bit machines
100 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
101 typedef struct stat64 lstat_t;
102 #define lstat_f lstat64
103 #define fstat_f fstat64
104 #define fstatat_f fstatat64
106 typedef struct stat lstat_t;
107 #define lstat_f lstat
108 #define fstat_f fstat
109 #define fstatat_f fstatat
112 #ifndef STATX_BASIC_STATS
114 * Timestamp structure for the timestamps in struct statx.
116 * tv_sec holds the number of seconds before (negative) or after (positive)
117 * 00:00:00 1st January 1970 UTC.
119 * tv_nsec holds a number of nanoseconds (0..999,999,999) after the tv_sec time.
121 * __reserved is held in case we need a yet finer resolution.
123 struct statx_timestamp {
130 * Structures for the extended file attribute retrieval system call
133 * The caller passes a mask of what they're specifically interested in as a
134 * parameter to statx(). What statx() actually got will be indicated in
135 * st_mask upon return.
137 * For each bit in the mask argument:
139 * - if the datum is not supported:
141 * - the bit will be cleared, and
143 * - the datum will be set to an appropriate fabricated value if one is
144 * available (eg. CIFS can take a default uid and gid), otherwise
146 * - the field will be cleared;
148 * - otherwise, if explicitly requested:
150 * - the datum will be synchronised to the server if AT_STATX_FORCE_SYNC is
151 * set or if the datum is considered out of date, and
153 * - the field will be filled in and the bit will be set;
155 * - otherwise, if not requested, but available in approximate form without any
156 * effort, it will be filled in anyway, and the bit will be set upon return
157 * (it might not be up to date, however, and no attempt will be made to
158 * synchronise the internal state first);
160 * - otherwise the field and the bit will be cleared before returning.
162 * Items in STATX_BASIC_STATS may be marked unavailable on return, but they
163 * will have values installed for compatibility purposes so that stat() and
164 * co. can be emulated in userspace.
168 __u32 stx_mask; /* What results were written [uncond] */
169 __u32 stx_blksize; /* Preferred general I/O size [uncond] */
170 __u64 stx_attributes; /* Flags conveying information about the file [uncond] */
172 __u32 stx_nlink; /* Number of hard links */
173 __u32 stx_uid; /* User ID of owner */
174 __u32 stx_gid; /* Group ID of owner */
175 __u16 stx_mode; /* File mode */
178 __u64 stx_ino; /* Inode number */
179 __u64 stx_size; /* File size */
180 __u64 stx_blocks; /* Number of 512-byte blocks allocated */
181 __u64 stx_attributes_mask; /* Mask to show what's supported in stx_attributes */
183 struct statx_timestamp stx_atime; /* Last access time */
184 struct statx_timestamp stx_btime; /* File creation time */
185 struct statx_timestamp stx_ctime; /* Last attribute change time */
186 struct statx_timestamp stx_mtime; /* Last data modification time */
188 __u32 stx_rdev_major; /* Device ID of special file [if bdev/cdev] */
189 __u32 stx_rdev_minor;
190 __u32 stx_dev_major; /* ID of device containing file [uncond] */
193 __u64 __spare2[14]; /* Spare space for future expansion */
198 * Flags to be stx_mask
200 * Query request/result mask for statx() and struct statx::stx_mask.
202 * These bits should be set in the mask argument of statx() to request
203 * particular items when calling statx().
205 #define STATX_TYPE 0x00000001U /* Want/got stx_mode & S_IFMT */
206 #define STATX_MODE 0x00000002U /* Want/got stx_mode & ~S_IFMT */
207 #define STATX_NLINK 0x00000004U /* Want/got stx_nlink */
208 #define STATX_UID 0x00000008U /* Want/got stx_uid */
209 #define STATX_GID 0x00000010U /* Want/got stx_gid */
210 #define STATX_ATIME 0x00000020U /* Want/got stx_atime */
211 #define STATX_MTIME 0x00000040U /* Want/got stx_mtime */
212 #define STATX_CTIME 0x00000080U /* Want/got stx_ctime */
213 #define STATX_INO 0x00000100U /* Want/got stx_ino */
214 #define STATX_SIZE 0x00000200U /* Want/got stx_size */
215 #define STATX_BLOCKS 0x00000400U /* Want/got stx_blocks */
216 #define STATX_BASIC_STATS 0x000007ffU /* The stuff in the normal stat struct */
217 #define STATX_BTIME 0x00000800U /* Want/got stx_btime */
218 #define STATX_ALL 0x00000fffU /* All currently supported flags */
219 #define STATX__RESERVED 0x80000000U /* Reserved for future struct statx expansion */
222 * Attributes to be found in stx_attributes and masked in stx_attributes_mask.
224 * These give information about the features or the state of a file that might
225 * be of use to ordinary userspace programs such as GUIs or ls rather than
228 * Note that the flags marked [I] correspond to generic FS_IOC_FLAGS
229 * semantically. Where possible, the numerical value is picked to correspond
232 #define STATX_ATTR_COMPRESSED 0x00000004 /* [I] File is compressed by the fs */
233 #define STATX_ATTR_IMMUTABLE 0x00000010 /* [I] File is marked immutable */
234 #define STATX_ATTR_APPEND 0x00000020 /* [I] File is append-only */
235 #define STATX_ATTR_NODUMP 0x00000040 /* [I] File is not to be dumped */
236 #define STATX_ATTR_ENCRYPTED 0x00000800 /* [I] File requires key to decrypt in fs */
238 #define STATX_ATTR_AUTOMOUNT 0x00001000 /* Dir: Automount trigger */
240 #define AT_STATX_SYNC_TYPE 0x6000 /* Type of synchronisation required from statx() */
241 #define AT_STATX_SYNC_AS_STAT 0x0000 /* - Do whatever stat() does */
242 #define AT_STATX_FORCE_SYNC 0x2000 /* - Force the attributes to be sync'd with the server */
243 #define AT_STATX_DONT_SYNC 0x4000 /* - Don't sync attributes with the server */
245 #endif /* STATX_BASIC_STATS */
247 typedef struct statx lstatx_t;
249 #define LUSTRE_EOF 0xffffffffffffffffULL
252 #define LL_SUPER_MAGIC 0x0BD00BD0
254 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
256 /* FIEMAP flags supported by Lustre */
257 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
259 enum obd_statfs_state {
260 OS_STATFS_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
261 OS_STATFS_READONLY = 0x00000002, /**< filesystem is read-only */
262 OS_STATFS_NOPRECREATE = 0x00000004, /**< no object precreation */
263 OS_STATFS_UNUSED1 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
264 OS_STATFS_UNUSED2 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
265 OS_STATFS_ENOSPC = 0x00000020, /**< not enough free space */
266 OS_STATFS_ENOINO = 0x00000040, /**< not enough inodes */
267 OS_STATFS_SUM = 0x00000100, /**< aggregated for all tagrets */
268 OS_STATFS_NONROT = 0x00000200, /**< non-rotational device */
271 /** filesystem statistics/attributes for target device */
273 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
274 __u64 os_blocks; /* total size in #os_bsize blocks */
275 __u64 os_bfree; /* number of unused blocks */
276 __u64 os_bavail; /* blocks available for allocation */
277 __u64 os_files; /* total number of objects */
278 __u64 os_ffree; /* # objects that could be created */
279 __u8 os_fsid[40]; /* identifier for filesystem */
280 __u32 os_bsize; /* block size in bytes for os_blocks */
281 __u32 os_namelen; /* maximum length of filename in bytes*/
282 __u64 os_maxbytes; /* maximum object size in bytes */
283 __u32 os_state; /**< obd_statfs_state OS_STATFS_* */
284 __u32 os_fprecreated; /* objs available now to the caller */
285 /* used in QoS code to find preferred
287 __u32 os_granted; /* space granted for MDS */
288 __u32 os_spare3; /* Unused padding fields. Remember */
289 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
297 /** additional filesystem attributes for target device */
298 struct obd_statfs_info {
299 __u32 os_reserved_mb_low; /* reserved mb low */
300 __u32 os_reserved_mb_high; /* reserved mb high */
301 bool os_enable_pre; /* enable pre create logic */
307 * FID is a cluster-wide unique identifier of a file or an object (stripe).
308 * FIDs are never reused.
312 * FID sequence. Sequence is a unit of migration: all files (objects)
313 * with FIDs from a given sequence are stored on the same server.
314 * Lustre should support 2^64 objects, so even if each sequence
315 * has only a single object we can still enumerate 2^64 objects.
318 /* FID number within sequence. */
321 * FID version, used to distinguish different versions (in the sense
322 * of snapshots, etc.) of the same file system object. Not currently
326 } __attribute__((packed));
328 static inline bool fid_is_zero(const struct lu_fid *fid)
330 return fid->f_seq == 0 && fid->f_oid == 0;
333 /* The data name_to_handle_at() places in a struct file_handle (at f_handle) */
334 struct lustre_file_handle {
335 struct lu_fid lfh_child;
336 struct lu_fid lfh_parent;
339 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
340 * MDT-object's FID::f_ver, instead it is the OST-object index in its
341 * parent MDT-object's layout EA. */
342 #define f_stripe_idx f_ver
345 __u32 ol_stripe_size;
346 __u32 ol_stripe_count;
350 } __attribute__((packed));
352 /* The filter_fid structure has changed several times over its lifetime.
353 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
354 * stripe_index and the "self FID" (objid/seq) to be able to recover the
355 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
356 * the OST, the "trusted.lma" xattr was added to the OST objects to store
357 * the "self FID" to be consistent with the MDT on-disk format, and the
358 * filter_fid only stored the MDT inode parent FID and stripe index.
360 * In 2.10, the addition of PFL composite layouts required more information
361 * to be stored into the filter_fid in order to be able to identify which
362 * component the OST object belonged. As well, the stripe size may vary
363 * between components, so it was no longer safe to assume the stripe size
364 * or stripe_count of a file. This is also more robust for plain layouts.
366 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
367 * enough space to store both the filter_fid and LMA in the inode, so they
368 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
369 * an extra seek for every OST object access.
371 * In 2.11, FLR mirror layouts also need to store the layout version and
372 * range so that writes to old versions of the layout are not allowed.
373 * That ensures that mirrored objects are not modified by evicted clients,
374 * and ensures that the components are correctly marked stale on the MDT.
376 struct filter_fid_18_23 {
377 struct lu_fid ff_parent; /* stripe_idx in f_ver */
382 struct filter_fid_24_29 {
383 struct lu_fid ff_parent; /* stripe_idx in f_ver */
386 struct filter_fid_210 {
387 struct lu_fid ff_parent; /* stripe_idx in f_ver */
388 struct ost_layout ff_layout;
392 struct lu_fid ff_parent; /* stripe_idx in f_ver */
393 struct ost_layout ff_layout;
394 __u32 ff_layout_version;
395 __u32 ff_range; /* range of layout version that
396 * write are allowed */
397 } __attribute__((packed));
399 /* Userspace should treat lu_fid as opaque, and only use the following methods
400 * to print or parse them. Other functions (e.g. compare, swab) could be moved
401 * here from lustre_idl.h if needed. */
405 LMAC_HSM = 0x00000001,
406 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
407 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
408 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
409 * under /O/<seq>/d<x>. */
410 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
411 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
412 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
416 * Masks for all features that should be supported by a Lustre version to
417 * access a specific file.
418 * This information is stored in lustre_mdt_attrs::lma_incompat.
421 LMAI_RELEASED = 0x00000001, /* file is released */
422 LMAI_AGENT = 0x00000002, /* agent inode */
423 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
424 is on the remote MDT */
425 LMAI_STRIPED = 0x00000008, /* striped directory inode */
426 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
427 LMAI_ENCRYPT = 0x00000020, /* inode is encrypted */
428 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
429 LMAI_STRIPED | LMAI_ORPHAN | LMAI_ENCRYPT)
434 * Following struct for object attributes, that will be kept inode's EA.
435 * Introduced in 2.0 release (please see b15993, for details)
436 * Added to all objects since Lustre 2.4 as contains self FID
438 struct lustre_mdt_attrs {
440 * Bitfield for supported data in this structure. From enum lma_compat.
441 * lma_self_fid and lma_flags are always available.
445 * Per-file incompat feature list. Lustre version should support all
446 * flags set in this field. The supported feature mask is available in
450 /** FID of this inode */
451 struct lu_fid lma_self_fid;
454 struct lustre_ost_attrs {
455 /* Use lustre_mdt_attrs directly for now, need a common header
456 * structure if want to change lustre_mdt_attrs in future. */
457 struct lustre_mdt_attrs loa_lma;
459 /* Below five elements are for OST-object's PFID EA, the
460 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
461 * and the stripe_index (low 16 bits), the size should not exceed
462 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
463 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
464 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
465 * loa_comp_* elements are valid. */
466 struct lu_fid loa_parent_fid;
467 __u32 loa_stripe_size;
469 __u64 loa_comp_start;
474 * Prior to 2.4, the LMA structure also included SOM attributes which has since
475 * been moved to a dedicated xattr
476 * lma_flags was also removed because of lma_compat/incompat fields.
478 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
480 enum lustre_som_flags {
481 /* Unknow or no SoM data, must get size from OSTs. */
482 SOM_FL_UNKNOWN = 0x0000,
483 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
484 SOM_FL_STRICT = 0x0001,
485 /* Known stale - was right at some point in the past, but it is
486 * known (or likely) to be incorrect now (e.g. opened for write). */
487 SOM_FL_STALE = 0x0002,
488 /* Approximate, may never have been strictly correct,
489 * need to sync SOM data to achieve eventual consistency. */
490 SOM_FL_LAZY = 0x0004,
493 struct lustre_som_attrs {
495 __u16 lsa_reserved[3];
501 * OST object IDentifier.
509 struct lu_fid oi_fid;
511 } __attribute__((packed));
513 #define DOSTID "%#llx:%llu"
514 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
515 ((unsigned long long)ostid_id(oi))
517 struct ll_futimes_3 {
519 __u64 lfu_atime_nsec;
521 __u64 lfu_mtime_nsec;
523 __u64 lfu_ctime_nsec;
527 * Maximum number of mirrors currently implemented.
529 #define LUSTRE_MIRROR_COUNT_MAX 16
531 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
533 LL_LEASE_RDLCK = 0x01,
534 LL_LEASE_WRLCK = 0x02,
535 LL_LEASE_UNLCK = 0x04,
538 enum ll_lease_flags {
539 LL_LEASE_RESYNC = 0x1,
540 LL_LEASE_RESYNC_DONE = 0x2,
541 LL_LEASE_LAYOUT_MERGE = 0x4,
542 LL_LEASE_LAYOUT_SPLIT = 0x8,
543 LL_LEASE_PCC_ATTACH = 0x10,
546 #define IOC_IDS_MAX 4096
547 struct ll_ioc_lease {
554 struct ll_ioc_lease_id {
565 * The ioctl naming rules:
566 * LL_* - works on the currently opened filehandle instead of parent dir
567 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
568 * *_MDC_* - gets/sets data related to MDC
569 * *_LOV_* - gets/sets data related to OSC/LOV
570 * *FILE* - called on parent dir and passes in a filename
571 * *STRIPE* - set/get lov_user_md
572 * *INFO - set/get lov_user_mds_data
574 /* lustre_ioctl.h 101-150 */
575 /* ioctl codes 128-143 are reserved for fsverity */
576 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
577 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
578 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
579 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
580 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
581 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
582 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
583 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
584 /* LL_IOC_RECREATE_OBJ 157 obsolete */
585 /* LL_IOC_RECREATE_FID 157 obsolete */
586 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
587 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
588 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
589 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
590 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
591 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
592 /* IOC_LOV_GETINFO 165 obsolete */
593 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
594 /* LL_IOC_RMTACL 167 obsolete */
595 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
596 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
597 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
598 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
599 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
600 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
601 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
602 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
603 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
604 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
605 /* lustre_ioctl.h 177-210 */
606 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
607 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
608 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
609 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
610 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
611 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
612 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
613 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
614 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
615 struct lustre_swap_layouts)
616 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
617 struct hsm_current_action)
618 /* lustre_ioctl.h 221-232 */
619 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
620 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
621 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
622 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
623 #define LL_IOC_UNLOCK_FOREIGN _IO('f', 242)
624 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
625 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
626 #define LL_IOC_GET_LEASE _IO('f', 244)
627 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
628 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
629 #define LL_IOC_MIGRATE _IOR('f', 247, int)
630 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
631 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
632 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
633 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
634 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
635 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
636 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
637 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
638 #define LL_IOC_PROJECT _IOW('f', 253, struct lu_project)
640 #ifndef FS_IOC_FSGETXATTR
642 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
645 __u32 fsx_xflags; /* xflags field value (get/set) */
646 __u32 fsx_extsize; /* extsize field value (get/set)*/
647 __u32 fsx_nextents; /* nextents field value (get) */
648 __u32 fsx_projid; /* project identifier (get/set) */
649 unsigned char fsx_pad[12];
651 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
652 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
654 #ifndef FS_XFLAG_PROJINHERIT
655 #define FS_XFLAG_PROJINHERIT 0x00000200
659 #define LL_STATFS_LMV 1
660 #define LL_STATFS_LOV 2
661 #define LL_STATFS_NODELAY 4
663 #define IOC_MDC_TYPE 'i'
664 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
665 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
666 #define IOC_MDC_GETFILEINFO_V1 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
667 #define IOC_MDC_GETFILEINFO_V2 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
668 #define LL_IOC_MDC_GETINFO_V1 _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
669 #define LL_IOC_MDC_GETINFO_V2 _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
670 #define IOC_MDC_GETFILEINFO IOC_MDC_GETFILEINFO_V1
671 #define LL_IOC_MDC_GETINFO LL_IOC_MDC_GETINFO_V1
673 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
675 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
676 * files, but are unlikely to be used in practice and are not harmful if
677 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
678 * devices and are safe for use on new files. See LU-4209. */
679 /* To be compatible with old statically linked binary we keep the check for
680 * the older 0100000000 flag. This is already removed upstream. LU-812. */
681 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
683 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
685 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
686 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
687 O_LOV_DELAY_CREATE_MASK)
688 /* O_FILE_ENC principle is similar to O_LOV_DELAY_CREATE above,
689 * for access to encrypted files without the encryption key.
691 #define O_FILE_ENC (O_NOCTTY | O_NDELAY)
693 #define LL_FILE_IGNORE_LOCK 0x00000001
694 #define LL_FILE_GROUP_LOCKED 0x00000002
695 #define LL_FILE_READAHEA 0x00000004
696 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
697 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
699 #define LOV_USER_MAGIC_V1 0x0BD10BD0
700 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
701 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
702 #define LOV_USER_MAGIC_V3 0x0BD30BD0
703 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
704 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
705 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
706 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
707 #define LOV_USER_MAGIC_SEL 0x0BD80BD0
709 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
710 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
711 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
713 #define LOV_PATTERN_NONE 0x000
714 #define LOV_PATTERN_RAID0 0x001
715 #define LOV_PATTERN_RAID1 0x002
716 #define LOV_PATTERN_MDT 0x100
717 #define LOV_PATTERN_OVERSTRIPING 0x200
718 #define LOV_PATTERN_FOREIGN 0x400
720 #define LOV_PATTERN_F_MASK 0xffff0000
721 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
722 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
723 #define LOV_PATTERN_DEFAULT 0xffffffff
725 #define LOV_OFFSET_DEFAULT ((__u16)-1)
726 #define LMV_OFFSET_DEFAULT ((__u32)-1)
728 static inline bool lov_pattern_supported(__u32 pattern)
730 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
731 (pattern & ~LOV_PATTERN_F_RELEASED) ==
732 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
733 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
736 /* RELEASED and MDT patterns are not valid in many places, so rather than
737 * having many extra checks on lov_pattern_supported, we have this separate
738 * check for non-released, non-DOM components
740 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
742 return pattern == LOV_PATTERN_RAID0 ||
743 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
747 #define LOV_MAXPOOLNAME 15
748 #define LOV_POOLNAMEF "%.15s"
750 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
751 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
752 #define LOV_MAX_STRIPE_COUNT_OLD 160
753 /* This calculation is crafted so that input of 4096 will result in 160
754 * which in turn is equal to old maximal stripe count.
755 * XXX: In fact this is too simpified for now, what it also need is to get
756 * ea_type argument to clearly know how much space each stripe consumes.
758 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
759 * allocation that is sufficient for the current generation of systems.
761 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
762 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
763 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
764 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
766 #define XATTR_LUSTRE_PREFIX "lustre."
767 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
769 /* Please update if XATTR_LUSTRE_LOV".set" groks more flags in the future */
770 #define allowed_lustre_lov(att) (strcmp((att), XATTR_LUSTRE_LOV".add") == 0 || \
771 strcmp((att), XATTR_LUSTRE_LOV".set") == 0 || \
772 strcmp((att), XATTR_LUSTRE_LOV".set.flags") == 0 || \
773 strcmp((att), XATTR_LUSTRE_LOV".del") == 0)
775 #define lov_user_ost_data lov_user_ost_data_v1
776 struct lov_user_ost_data_v1 { /* per-stripe data structure */
777 struct ost_id l_ost_oi; /* OST object ID */
778 __u32 l_ost_gen; /* generation of this OST index */
779 __u32 l_ost_idx; /* OST index in LOV */
780 } __attribute__((packed));
782 #define lov_user_md lov_user_md_v1
783 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
784 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
785 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
786 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
787 __u32 lmm_stripe_size; /* size of stripe in bytes */
788 __u16 lmm_stripe_count; /* num stripes in use for this object */
790 __u16 lmm_stripe_offset; /* starting stripe offset in
791 * lmm_objects, use when writing */
792 __u16 lmm_layout_gen; /* layout generation number
793 * used when reading */
795 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
796 } __attribute__((packed, __may_alias__));
798 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
799 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
800 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
801 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
802 __u32 lmm_stripe_size; /* size of stripe in bytes */
803 __u16 lmm_stripe_count; /* num stripes in use for this object */
805 __u16 lmm_stripe_offset; /* starting stripe offset in
806 * lmm_objects, use when writing */
807 __u16 lmm_layout_gen; /* layout generation number
808 * used when reading */
810 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
811 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
812 } __attribute__((packed, __may_alias__));
814 struct lov_foreign_md {
815 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
816 __u32 lfm_length; /* length of lfm_value */
817 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
818 __u32 lfm_flags; /* flags, type specific */
820 } __attribute__((packed));
822 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
823 offsetof(struct lov_foreign_md, lfm_value))
825 #define foreign_size_le(lfm) \
826 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
827 offsetof(struct lov_foreign_md, lfm_value))
830 * The stripe size fields are shared for the extension size storage, however
831 * the extension size is stored in KB, not bytes.
833 #define SEL_UNIT_SIZE 1024llu
838 } __attribute__((packed));
840 #define DEXT "[%#llx, %#llx)"
841 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
843 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
844 struct lu_extent *e2)
846 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
849 static inline bool lu_extent_is_whole(struct lu_extent *e)
851 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
854 enum lov_comp_md_entry_flags {
855 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
856 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
857 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
858 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
859 LCME_FL_OFFLINE = 0x00000008, /* Not used */
860 LCME_FL_INIT = 0x00000010, /* instantiated */
861 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
862 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
863 LCME_FL_PARITY = 0x00000080, /* EC: a parity code component */
864 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
865 * won't be stored on disk
869 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
870 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
873 /* The component flags can be set by users at creation/modification time. */
874 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
877 /* The mirror flags can be set by users at creation time. */
878 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
880 /* The allowed flags obtained from the client at component creation time. */
881 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
883 /* The mirror flags sent by client */
884 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
886 /* These flags have meaning when set in a default layout and will be inherited
887 * from the default/template layout set on a directory.
889 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
892 /* the highest bit in obdo::o_layout_version is used to mark if the file is
894 #define LU_LAYOUT_RESYNC LCME_FL_NEG
896 /* lcme_id can be specified as certain flags, and the the first
897 * bit of lcme_id is used to indicate that the ID is representing
898 * certain LCME_FL_* but not a real ID. Which implies we can have
899 * at most 31 flags (see LCME_FL_XXX). */
902 LCME_ID_MAX = 0x7FFFFFFF,
903 LCME_ID_ALL = 0xFFFFFFFF,
904 LCME_ID_NOT_ID = LCME_FL_NEG
907 #define LCME_ID_MASK LCME_ID_MAX
909 struct lov_comp_md_entry_v1 {
910 __u32 lcme_id; /* unique id of component */
911 __u32 lcme_flags; /* LCME_FL_XXX */
912 /* file extent for component. If it's an EC code component, its flags
913 * contains LCME_FL_PARITY, and its extent covers the same extent of
914 * its corresponding data component.
916 struct lu_extent lcme_extent;
917 __u32 lcme_offset; /* offset of component blob,
918 start from lov_comp_md_v1 */
919 __u32 lcme_size; /* size of component blob */
920 __u32 lcme_layout_gen;
921 __u64 lcme_timestamp; /* snapshot time if applicable*/
922 __u8 lcme_dstripe_count; /* data stripe count,
925 __u8 lcme_cstripe_count; /* code stripe count,
928 __u16 lcme_padding_1;
929 } __attribute__((packed));
931 #define SEQ_ID_MAX 0x0000FFFF
932 #define SEQ_ID_MASK SEQ_ID_MAX
933 /* bit 30:16 of lcme_id is used to store mirror id */
934 #define MIRROR_ID_MASK 0x7FFF0000
935 #define MIRROR_ID_NEG 0x8000
936 #define MIRROR_ID_SHIFT 16
938 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
940 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
943 static inline __u16 mirror_id_of(__u32 id)
945 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
949 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
951 enum lov_comp_md_flags {
952 /* the least 4 bits are used by FLR to record file state */
955 LCM_FL_WRITE_PENDING = 0x2,
956 LCM_FL_SYNC_PENDING = 0x3,
957 LCM_FL_PCC_RDONLY = 0x8,
958 LCM_FL_FLR_MASK = 0xB,
961 struct lov_comp_md_v1 {
962 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
963 __u32 lcm_size; /* overall size including this struct */
964 __u32 lcm_layout_gen;
966 __u16 lcm_entry_count;
967 /* lcm_mirror_count stores the number of actual mirrors minus 1,
968 * so that non-flr files will have value 0 meaning 1 mirror. */
969 __u16 lcm_mirror_count;
970 /* code components count, non-EC file contains 0 ec_count */
972 __u8 lcm_padding3[1];
973 __u16 lcm_padding1[2];
975 struct lov_comp_md_entry_v1 lcm_entries[0];
976 } __attribute__((packed));
978 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
980 if (stripes == (__u16)-1)
983 if (lmm_magic == LOV_USER_MAGIC_V1)
984 return sizeof(struct lov_user_md_v1) +
985 stripes * sizeof(struct lov_user_ost_data_v1);
986 return sizeof(struct lov_user_md_v3) +
987 stripes * sizeof(struct lov_user_ost_data_v1);
990 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
991 * use this. It is unsafe to #define those values in this header as it
992 * is possible the application has already #included <sys/stat.h>. */
993 #define lov_user_mds_data lov_user_mds_data_v2
994 struct lov_user_mds_data_v1 {
995 lstat_t lmd_st; /* MDS stat struct */
996 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
997 } __attribute__((packed));
999 struct lov_user_mds_data_v2 {
1000 struct lu_fid lmd_fid; /* Lustre FID */
1001 lstatx_t lmd_stx; /* MDS statx struct */
1002 __u64 lmd_flags; /* MDS stat flags */
1003 __u32 lmd_lmmsize; /* LOV EA size */
1004 __u32 lmd_padding; /* unused */
1005 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
1006 } __attribute__((packed));
1008 struct lmv_user_mds_data {
1009 struct lu_fid lum_fid;
1012 } __attribute__((packed, __may_alias__));
1014 enum lmv_hash_type {
1015 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
1016 LMV_HASH_TYPE_ALL_CHARS = 1, /* simple sum of characters */
1017 LMV_HASH_TYPE_FNV_1A_64 = 2, /* reasonable non-cryptographic hash */
1018 LMV_HASH_TYPE_CRUSH = 3, /* double-hash to optimize migration */
1019 LMV_HASH_TYPE_CRUSH2 = 4, /* CRUSH with small fixes, LU-15692 */
1021 LMV_HASH_TYPE_DEFAULT = LMV_HASH_TYPE_FNV_1A_64
1024 static __attribute__((unused)) const char *mdt_hash_name[] = {
1033 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1034 * and the higher part will be the flag to indicate the status of object,
1035 * for example the object is being migrated. And the hash function
1036 * might be interpreted differently with different flags. */
1037 #define LMV_HASH_TYPE_MASK 0x0000ffff
1039 static inline bool lmv_is_known_hash_type(__u32 type)
1041 return (type & LMV_HASH_TYPE_MASK) > LMV_HASH_TYPE_UNKNOWN &&
1042 (type & LMV_HASH_TYPE_MASK) < LMV_HASH_TYPE_MAX;
1045 /* fixed layout, such directories won't split automatically */
1046 /* NB, update LMV_HASH_FLAG_KNOWN when adding new flag */
1047 #define LMV_HASH_FLAG_FIXED 0x02000000
1048 #define LMV_HASH_FLAG_MERGE 0x04000000
1049 #define LMV_HASH_FLAG_SPLIT 0x08000000
1051 /* The striped directory has ever lost its master LMV EA, then LFSCK
1052 * re-generated it. This flag is used to indicate such case. It is an
1054 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
1056 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1057 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1059 #define LMV_HASH_FLAG_LAYOUT_CHANGE \
1060 (LMV_HASH_FLAG_MIGRATION | LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MERGE)
1062 #define LMV_HASH_FLAG_KNOWN 0xbe000000
1064 /* both SPLIT and MIGRATION are set for directory split */
1065 static inline bool lmv_hash_is_splitting(__u32 hash)
1067 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1068 (LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MIGRATION);
1071 /* both MERGE and MIGRATION are set for directory merge */
1072 static inline bool lmv_hash_is_merging(__u32 hash)
1074 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1075 (LMV_HASH_FLAG_MERGE | LMV_HASH_FLAG_MIGRATION);
1078 /* only MIGRATION is set for directory migration */
1079 static inline bool lmv_hash_is_migrating(__u32 hash)
1081 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) == LMV_HASH_FLAG_MIGRATION;
1084 static inline bool lmv_hash_is_restriping(__u32 hash)
1086 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash);
1089 static inline bool lmv_hash_is_layout_changing(__u32 hash)
1091 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash) ||
1092 lmv_hash_is_migrating(hash);
1095 struct lustre_foreign_type {
1097 const char *lft_name;
1101 * LOV/LMV foreign types
1103 enum lustre_foreign_types {
1104 LU_FOREIGN_TYPE_NONE = 0,
1105 LU_FOREIGN_TYPE_SYMLINK = 0xda05,
1106 /* must be the max/last one */
1107 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
1110 extern struct lustre_foreign_type lu_foreign_types[];
1112 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
1113 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
1114 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
1115 #define lmv_user_md lmv_user_md_v1
1116 struct lmv_user_md_v1 {
1117 __u32 lum_magic; /* must be the first field */
1118 __u32 lum_stripe_count; /* dirstripe count */
1119 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
1120 __u32 lum_hash_type; /* Dir stripe policy */
1121 __u32 lum_type; /* LMV type: default */
1122 __u8 lum_max_inherit; /* inherit depth of default LMV */
1123 __u8 lum_max_inherit_rr; /* inherit depth of default LMV to round-robin mkdir */
1127 char lum_pool_name[LOV_MAXPOOLNAME + 1];
1128 struct lmv_user_mds_data lum_objects[0];
1129 } __attribute__((packed));
1131 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
1133 if (size <= sizeof(struct lmv_user_md))
1136 size -= sizeof(struct lmv_user_md);
1137 return (size + sizeof(struct lmv_user_mds_data) - 1) /
1138 sizeof(struct lmv_user_mds_data);
1142 * NB, historically default layout didn't set type, but use XATTR name to differ
1143 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
1144 * and still use the same method.
1147 LMV_TYPE_DEFAULT = 0x0000,
1150 /* lum_max_inherit will be decreased by 1 after each inheritance if it's not
1151 * LMV_INHERIT_UNLIMITED or > LMV_INHERIT_MAX.
1154 /* for historical reason, 0 means unlimited inheritance */
1155 LMV_INHERIT_UNLIMITED = 0,
1156 /* unlimited lum_max_inherit by default for plain stripe (0 or 1) */
1157 LMV_INHERIT_DEFAULT_PLAIN = LMV_INHERIT_UNLIMITED,
1158 /* not inherit any more */
1159 LMV_INHERIT_END = 1,
1160 /* for multiple stripes, the default lum_max_inherit is 3 */
1161 LMV_INHERIT_DEFAULT_STRIPED = 3,
1162 /* max inherit depth */
1163 LMV_INHERIT_MAX = 250,
1164 /* [251, 254] are reserved */
1165 /* not set, or when inherit depth goes beyond end, */
1166 LMV_INHERIT_NONE = 255,
1170 /* not set, or when inherit_rr depth goes beyond end, */
1171 LMV_INHERIT_RR_NONE = 0,
1172 /* disable lum_max_inherit_rr by default */
1173 LMV_INHERIT_RR_DEFAULT = LMV_INHERIT_RR_NONE,
1174 /* not inherit any more */
1175 LMV_INHERIT_RR_END = 1,
1176 /* default inherit_rr of ROOT */
1177 LMV_INHERIT_RR_ROOT = 3,
1178 /* max inherit depth */
1179 LMV_INHERIT_RR_MAX = 250,
1180 /* [251, 254] are reserved */
1181 /* unlimited inheritance */
1182 LMV_INHERIT_RR_UNLIMITED = 255,
1185 static inline int lmv_user_md_size(int stripes, int lmm_magic)
1187 int size = sizeof(struct lmv_user_md);
1189 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
1190 size += stripes * sizeof(struct lmv_user_mds_data);
1195 struct ll_recreate_obj {
1201 __u64 id; /* holds object id */
1202 __u32 generation; /* holds object generation */
1203 __u32 f_type; /* holds object type or stripe idx when passing it to
1204 * OST for saving into EA. */
1209 char uuid[UUID_MAX];
1212 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
1213 const struct obd_uuid *u2)
1215 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
1218 static inline int obd_uuid_empty(struct obd_uuid *uuid)
1220 return uuid->uuid[0] == '\0';
1223 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
1225 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
1226 uuid->uuid[sizeof(*uuid) - 1] = '\0';
1229 /* For printf's only, make sure uuid is terminated */
1230 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
1235 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
1236 /* Obviously not safe, but for printfs, no real harm done...
1237 we're always null-terminated, even in a race. */
1238 static char temp[sizeof(*uuid->uuid)];
1240 memcpy(temp, uuid->uuid, sizeof(*uuid->uuid) - 1);
1241 temp[sizeof(*uuid->uuid) - 1] = '\0';
1245 return (char *)(uuid->uuid);
1248 #define LUSTRE_MAXFSNAME 8
1249 #define LUSTRE_MAXINSTANCE 16
1251 /* Extract fsname from uuid (or target name) of a target
1252 e.g. (myfs-OST0007_UUID -> myfs)
1253 see also deuuidify. */
1254 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1258 strncpy(buf, uuid, buflen - 1);
1259 buf[buflen - 1] = '\0';
1260 p = strrchr(buf, '-');
1265 /* printf display format for Lustre FIDs
1266 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1267 #define FID_NOBRACE_LEN 40
1268 #define FID_LEN (FID_NOBRACE_LEN + 2)
1269 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1270 #define DFID "[" DFID_NOBRACE "]"
1271 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1273 /* scanf input parse format for fids in DFID_NOBRACE format
1274 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1275 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1276 #define SFID "0x%llx:0x%x:0x%x"
1277 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1279 /********* Quotas **********/
1281 /* From linux/fs/quota/quota.c */
1282 static inline __u64 toqb(__kernel_size_t space)
1284 return (space + QIF_DQBLKSIZE - 1) >> QIF_DQBLKSIZE_BITS;
1287 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1288 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1289 #define Q_GETOINFO 0x800102 /* get obd quota info */
1290 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1291 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1293 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1294 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1295 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1296 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1297 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1298 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1299 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1300 /* lustre-specific control commands */
1301 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1302 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1303 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1304 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1305 #define LUSTRE_Q_GETQUOTAPOOL 0x80000f /* get user pool quota */
1306 #define LUSTRE_Q_SETQUOTAPOOL 0x800010 /* set user pool quota */
1307 #define LUSTRE_Q_GETINFOPOOL 0x800011 /* get pool quota info */
1308 #define LUSTRE_Q_SETINFOPOOL 0x800012 /* set pool quota info */
1309 #define LUSTRE_Q_GETDEFAULT_POOL 0x800013 /* get default pool quota*/
1310 #define LUSTRE_Q_SETDEFAULT_POOL 0x800014 /* set default pool quota */
1311 #define LUSTRE_Q_DELETEQID 0x800015 /* delete quota ID */
1312 /* In the current Lustre implementation, the grace time is either the time
1313 * or the timestamp to be used after some quota ID exceeds the soft limt,
1314 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1316 #define LQUOTA_GRACE_BITS 48
1317 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1318 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1319 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1320 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1321 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1323 /* special grace time, only notify the user when its quota is over soft limit
1324 * but doesn't block new writes until the hard limit is reached. */
1325 #define NOTIFY_GRACE "notify"
1326 #define NOTIFY_GRACE_TIME LQUOTA_GRACE_MASK
1328 /* different quota flags */
1330 /* the default quota flag, the corresponding quota ID will use the default
1331 * quota setting, the hardlimit and softlimit of its quota record in the global
1332 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1333 * and high 16 bits will contain this flag (see above comment).
1335 #define LQUOTA_FLAG_DEFAULT 0x0001
1336 #define LQUOTA_FLAG_DELETED 0x0002
1338 #define LUSTRE_Q_CMD_IS_POOL(cmd) \
1339 (cmd == LUSTRE_Q_GETQUOTAPOOL || \
1340 cmd == LUSTRE_Q_SETQUOTAPOOL || \
1341 cmd == LUSTRE_Q_SETINFOPOOL || \
1342 cmd == LUSTRE_Q_GETINFOPOOL || \
1343 cmd == LUSTRE_Q_SETDEFAULT_POOL || \
1344 cmd == LUSTRE_Q_GETDEFAULT_POOL)
1346 #define ALLQUOTA 255 /* set all quota */
1347 static inline const char *qtype_name(int qtype)
1360 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1363 #define N_PERMS_MAX 64
1365 struct perm_downcall_data {
1371 struct identity_downcall_data {
1378 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1379 __u32 idd_groups[0];
1382 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
1383 /* old interface struct is deprecated in 2.14 */
1384 #define SEPOL_DOWNCALL_MAGIC_OLD 0x8b8bb842
1385 struct sepol_downcall_data_old {
1387 __s64 sdd_sepol_mtime;
1388 __u16 sdd_sepol_len;
1393 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb843
1394 struct sepol_downcall_data {
1396 __u16 sdd_sepol_len;
1398 __s64 sdd_sepol_mtime;
1402 #ifdef NEED_QUOTA_DEFS
1404 #define QIF_BLIMITS 1
1406 #define QIF_ILIMITS 4
1407 #define QIF_INODES 8
1408 #define QIF_BTIME 16
1409 #define QIF_ITIME 32
1410 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1411 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1412 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1413 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1416 #endif /* !__KERNEL__ */
1418 /* lustre volatile file support
1419 * file name header: ".^L^S^T^R:volatile"
1421 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1422 #define LUSTRE_VOLATILE_HDR_LEN 14
1424 enum lustre_quota_version {
1428 /* XXX: same as if_dqinfo struct in kernel */
1436 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1438 __u64 dqb_bhardlimit; /* kbytes unit */
1439 __u64 dqb_bsoftlimit; /* kbytes unit */
1440 __u64 dqb_curspace; /* bytes unit */
1441 __u64 dqb_ihardlimit;
1442 __u64 dqb_isoftlimit;
1443 __u64 dqb_curinodes;
1457 struct if_quotactl {
1464 struct obd_dqinfo qc_dqinfo;
1465 struct obd_dqblk qc_dqblk;
1467 struct obd_uuid obd_uuid;
1468 char qc_poolname[0];
1471 /* swap layout flags */
1472 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1473 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1474 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1475 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1476 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1478 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1479 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1480 struct lustre_swap_layouts {
1488 /** Bit-mask of valid attributes */
1489 /* The LA_* flags are written to disk as part of the ChangeLog records
1490 * so they are part of the on-disk and network protocol, and cannot be changed.
1491 * Only the first 12 bits are currently saved.
1494 LA_ATIME = 1 << 0, /* 0x00001 */
1495 LA_MTIME = 1 << 1, /* 0x00002 */
1496 LA_CTIME = 1 << 2, /* 0x00004 */
1497 LA_SIZE = 1 << 3, /* 0x00008 */
1498 LA_MODE = 1 << 4, /* 0x00010 */
1499 LA_UID = 1 << 5, /* 0x00020 */
1500 LA_GID = 1 << 6, /* 0x00040 */
1501 LA_BLOCKS = 1 << 7, /* 0x00080 */
1502 LA_TYPE = 1 << 8, /* 0x00100 */
1503 LA_FLAGS = 1 << 9, /* 0x00200 */
1504 LA_NLINK = 1 << 10, /* 0x00400 */
1505 LA_RDEV = 1 << 11, /* 0x00800 */
1506 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1507 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1508 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1509 LA_PROJID = 1 << 15, /* 0x08000 */
1510 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1511 LA_LSIZE = 1 << 17, /* 0x20000 */
1512 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1513 LA_BTIME = 1 << 19, /* 0x80000 */
1515 * Attributes must be transmitted to OST objects
1517 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1520 #define MDS_FMODE_READ 00000001
1521 #define MDS_FMODE_WRITE 00000002
1523 #define MDS_FMODE_CLOSED 00000000
1524 #define MDS_FMODE_EXEC 00000004
1525 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1526 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1527 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1529 #define MDS_OPEN_CREATED 00000010
1530 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1532 #define MDS_OPEN_CREAT 00000100
1533 #define MDS_OPEN_EXCL 00000200
1534 #define MDS_OPEN_TRUNC 00001000
1535 #define MDS_OPEN_APPEND 00002000
1536 #define MDS_OPEN_SYNC 00010000
1537 #define MDS_OPEN_DIRECTORY 00200000
1539 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1540 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1541 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1542 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1543 * We do not support JOIN FILE
1544 * anymore, reserve this flags
1545 * just for preventing such bit
1548 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1549 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1550 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1551 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1552 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1554 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1556 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1557 * delegation, succeed if it's not
1558 * being opened with conflict mode.
1560 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1562 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1563 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1564 * for newly created file */
1565 #define MDS_OP_WITH_FID 020000000000000ULL /* operation carried out by FID */
1567 /* lustre internal open flags, which should not be set from user space */
1568 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1569 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1570 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1571 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1572 MDS_OPEN_PCC | MDS_OP_WITH_FID)
1575 /********* Changelogs **********/
1576 /** Changelog record types */
1577 enum changelog_rec_type {
1580 CL_CREATE = 1, /* namespace */
1581 CL_MKDIR = 2, /* namespace */
1582 CL_HARDLINK = 3, /* namespace */
1583 CL_SOFTLINK = 4, /* namespace */
1584 CL_MKNOD = 5, /* namespace */
1585 CL_UNLINK = 6, /* namespace */
1586 CL_RMDIR = 7, /* namespace */
1587 CL_RENAME = 8, /* namespace */
1588 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1589 CL_OPEN = 10, /* not currently used */
1590 CL_CLOSE = 11, /* may be written to log only with mtime change */
1591 CL_LAYOUT = 12, /* file layout/striping modified */
1595 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1596 CL_HSM = 16, /* HSM specific events, see flags */
1597 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1601 CL_FLRW = 21, /* FLR: file was firstly written */
1602 CL_RESYNC = 22, /* FLR: file was resync-ed */
1604 CL_DN_OPEN = 24, /* denied open */
1608 static inline const char *changelog_type2str(int type) {
1609 static const char *const changelog_str[] = {
1610 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1611 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1612 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1613 "FLRW", "RESYNC","GXATR", "NOPEN",
1616 if (type >= 0 && type < CL_LAST)
1617 return changelog_str[type];
1621 /* 12 bits of per-record data can be stored in the bottom of the flags */
1622 #define CLF_FLAGSHIFT 12
1623 enum changelog_rec_flags {
1624 CLF_VERSION = 0x1000,
1625 CLF_RENAME = 0x2000,
1627 CLF_EXTRA_FLAGS = 0x8000,
1628 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1630 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1631 CLF_VERMASK = ~CLF_FLAGMASK,
1635 /* Anything under the flagmask may be per-type (if desired) */
1636 /* Flags for unlink */
1637 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1638 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1639 /* HSM cleaning needed */
1640 /* Flags for rename */
1641 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1643 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1644 * has an archive in backend */
1647 /* 12b used (from high weight to low weight):
1652 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1653 #define CLF_HSM_ERR_H 6
1654 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1655 #define CLF_HSM_EVENT_H 9
1656 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1657 #define CLF_HSM_FLAG_H 11
1658 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1659 #define CLF_HSM_SPARE_H 15
1660 #define CLF_HSM_LAST 15
1662 /* Remove bits higher than _h, then extract the value
1663 * between _h and _l by shifting lower weigth to bit 0. */
1664 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1665 >> (CLF_HSM_LAST - _h + _l))
1667 #define CLF_HSM_SUCCESS 0x00
1668 #define CLF_HSM_MAXERROR 0x7E
1669 #define CLF_HSM_ERROVERFLOW 0x7F
1671 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1673 /* 3 bits field => 8 values allowed */
1685 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1687 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1691 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1694 *clf_flags = (enum changelog_rec_flags)
1695 (*clf_flags | (he << CLF_HSM_EVENT_L));
1698 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1700 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1703 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1706 *clf_flags = (enum changelog_rec_flags)
1707 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1710 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1712 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1715 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1718 *clf_flags = (enum changelog_rec_flags)
1719 (*clf_flags | (error << CLF_HSM_ERR_L));
1722 enum changelog_rec_extra_flags {
1724 CLFE_UIDGID = 0x0001,
1727 CLFE_XATTR = 0x0008,
1728 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1731 enum changelog_send_flag {
1732 /* Not yet implemented */
1733 CHANGELOG_FLAG_FOLLOW = 0x01,
1734 /* Blocking IO makes sense in case of slow user parsing of the records,
1735 * but it also prevents us from cleaning up if the records are not
1737 CHANGELOG_FLAG_BLOCK = 0x02,
1738 /* Pack jobid into the changelog records if available. */
1739 CHANGELOG_FLAG_JOBID = 0x04,
1740 /* Pack additional flag bits into the changelog record */
1741 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1744 enum changelog_send_extra_flag {
1745 /* Pack uid/gid into the changelog record */
1746 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1747 /* Pack nid into the changelog record */
1748 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1749 /* Pack open mode into the changelog record */
1750 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1751 /* Pack xattr name into the changelog record */
1752 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1755 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1756 changelog_rec_offset(CLF_SUPPORTED, \
1759 /* 31 usable bytes string + null terminator. */
1760 #define LUSTRE_JOBID_SIZE 32
1762 /* This is the minimal changelog record. It can contain extensions
1763 * such as rename fields or process jobid. Its exact content is described
1764 * by the cr_flags and cr_extra_flags.
1766 * Extensions are packed in the same order as their corresponding flags,
1767 * then in the same order as their corresponding extra flags.
1769 struct changelog_rec {
1771 __u16 cr_flags; /**< \a changelog_rec_flags */
1772 __u32 cr_type; /**< \a changelog_rec_type */
1773 __u64 cr_index; /**< changelog record number */
1774 __u64 cr_prev; /**< last index for this target fid */
1777 struct lu_fid cr_tfid; /**< target fid */
1778 __u32 cr_markerflags; /**< CL_MARK flags */
1780 struct lu_fid cr_pfid; /**< parent fid */
1781 } __attribute__ ((packed));
1783 /* Changelog extension for RENAME. */
1784 struct changelog_ext_rename {
1785 struct lu_fid cr_sfid; /**< source fid, or zero */
1786 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1789 /* Changelog extension to include JOBID. */
1790 struct changelog_ext_jobid {
1791 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1794 /* Changelog extension to include additional flags. */
1795 struct changelog_ext_extra_flags {
1796 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1799 /* Changelog extra extension to include UID/GID. */
1800 struct changelog_ext_uidgid {
1805 /* Changelog extra extension to include NID. */
1806 struct changelog_ext_nid {
1807 /* have __u64 instead of lnet_nid_t type for use by client api */
1809 /* for use when IPv6 support is added */
1814 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1815 struct changelog_ext_openmode {
1819 /* Changelog extra extension to include xattr */
1820 struct changelog_ext_xattr {
1821 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1824 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1825 const struct changelog_rec *rec);
1827 static inline __kernel_size_t changelog_rec_offset(enum changelog_rec_flags crf,
1828 enum changelog_rec_extra_flags cref)
1830 __kernel_size_t size = sizeof(struct changelog_rec);
1832 if (crf & CLF_RENAME)
1833 size += sizeof(struct changelog_ext_rename);
1835 if (crf & CLF_JOBID)
1836 size += sizeof(struct changelog_ext_jobid);
1838 if (crf & CLF_EXTRA_FLAGS) {
1839 size += sizeof(struct changelog_ext_extra_flags);
1840 if (cref & CLFE_UIDGID)
1841 size += sizeof(struct changelog_ext_uidgid);
1842 if (cref & CLFE_NID)
1843 size += sizeof(struct changelog_ext_nid);
1844 if (cref & CLFE_OPEN)
1845 size += sizeof(struct changelog_ext_openmode);
1846 if (cref & CLFE_XATTR)
1847 size += sizeof(struct changelog_ext_xattr);
1853 static inline __kernel_size_t changelog_rec_size(const struct changelog_rec *rec)
1855 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1857 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1858 cref = (enum changelog_rec_extra_flags)
1859 changelog_rec_extra_flags(rec)->cr_extra_flags;
1861 return changelog_rec_offset(
1862 (enum changelog_rec_flags)rec->cr_flags, cref);
1865 static inline __kernel_size_t changelog_rec_varsize(const struct changelog_rec *rec)
1867 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1871 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1873 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1874 (rec->cr_flags & CLF_VERSION);
1876 return (struct changelog_ext_rename *)((char *)rec +
1877 changelog_rec_offset(crf,
1881 /* The jobid follows the rename extension, if present */
1883 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1885 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1886 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1888 return (struct changelog_ext_jobid *)((char *)rec +
1889 changelog_rec_offset(crf,
1893 /* The additional flags follow the rename and jobid extensions, if present */
1895 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1896 const struct changelog_rec *rec)
1898 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1899 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1901 return (struct changelog_ext_extra_flags *)((char *)rec +
1902 changelog_rec_offset(crf,
1906 /* The uid/gid is the first extra extension */
1908 struct changelog_ext_uidgid *changelog_rec_uidgid(
1909 const struct changelog_rec *rec)
1911 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1913 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1915 return (struct changelog_ext_uidgid *)((char *)rec +
1916 changelog_rec_offset(crf,
1920 /* The nid is the second extra extension */
1922 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1924 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1926 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1927 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1929 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1930 cref = (enum changelog_rec_extra_flags)
1931 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1934 return (struct changelog_ext_nid *)((char *)rec +
1935 changelog_rec_offset(crf, cref));
1938 /* The OPEN mode is the third extra extension */
1940 struct changelog_ext_openmode *changelog_rec_openmode(
1941 const struct changelog_rec *rec)
1943 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1945 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1946 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1948 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1949 cref = (enum changelog_rec_extra_flags)
1950 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1951 (CLFE_UIDGID | CLFE_NID));
1954 return (struct changelog_ext_openmode *)((char *)rec +
1955 changelog_rec_offset(crf, cref));
1958 /* The xattr name is the fourth extra extension */
1960 struct changelog_ext_xattr *changelog_rec_xattr(
1961 const struct changelog_rec *rec)
1963 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1965 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1966 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1968 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1969 cref = (enum changelog_rec_extra_flags)
1970 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1971 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
1973 return (struct changelog_ext_xattr *)((char *)rec +
1974 changelog_rec_offset(crf, cref));
1977 /* The name follows the rename, jobid and extra flags extns, if present */
1978 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1980 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1982 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1983 cref = (enum changelog_rec_extra_flags)
1984 changelog_rec_extra_flags(rec)->cr_extra_flags;
1986 return (char *)rec + changelog_rec_offset(
1987 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
1988 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
1991 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1993 return strchrnul(changelog_rec_name(rec), '\0') + 1;
1996 static inline __kernel_size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1998 return strlen(changelog_rec_sname(rec));
2002 * Remap a record to the desired format as specified by the crf flags.
2003 * The record must be big enough to contain the final remapped version.
2004 * Superfluous extension fields are removed and missing ones are added
2005 * and zeroed. The flags of the record are updated accordingly.
2007 * The jobid and rename extensions can be added to a record, to match the
2008 * format an application expects, typically. In this case, the newly added
2009 * fields will be zeroed.
2010 * The Jobid field can be removed, to guarantee compatibility with older
2011 * clients that don't expect this field in the records they process.
2013 * The following assumptions are being made:
2014 * - CLF_RENAME will not be removed
2015 * - CLF_JOBID will not be added without CLF_RENAME being added too
2016 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
2018 * @param[in,out] rec The record to remap.
2019 * @param[in] crf_wanted Flags describing the desired extensions.
2020 * @param[in] cref_want Flags describing the desired extra extensions.
2022 static inline void changelog_remap_rec(struct changelog_rec *rec,
2023 enum changelog_rec_flags crf_wanted,
2024 enum changelog_rec_extra_flags cref_want)
2026 char *xattr_mov = NULL;
2027 char *omd_mov = NULL;
2028 char *nid_mov = NULL;
2029 char *uidgid_mov = NULL;
2033 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2035 crf_wanted = (enum changelog_rec_flags)
2036 (crf_wanted & CLF_SUPPORTED);
2037 cref_want = (enum changelog_rec_extra_flags)
2038 (cref_want & CLFE_SUPPORTED);
2040 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
2041 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
2042 (rec->cr_flags & CLF_EXTRA_FLAGS &&
2043 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2049 /* First move the variable-length name field */
2050 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
2051 changelog_rec_name(rec), rec->cr_namelen);
2053 /* Locations of extensions in the remapped record */
2054 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
2055 xattr_mov = (char *)rec +
2056 changelog_rec_offset(
2057 (enum changelog_rec_flags)
2058 (crf_wanted & CLF_SUPPORTED),
2059 (enum changelog_rec_extra_flags)
2060 (cref_want & ~CLFE_XATTR));
2061 omd_mov = (char *)rec +
2062 changelog_rec_offset(
2063 (enum changelog_rec_flags)
2064 (crf_wanted & CLF_SUPPORTED),
2065 (enum changelog_rec_extra_flags)
2066 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
2067 nid_mov = (char *)rec +
2068 changelog_rec_offset(
2069 (enum changelog_rec_flags)
2070 (crf_wanted & CLF_SUPPORTED),
2071 (enum changelog_rec_extra_flags)
2073 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
2074 uidgid_mov = (char *)rec +
2075 changelog_rec_offset(
2076 (enum changelog_rec_flags)
2077 (crf_wanted & CLF_SUPPORTED),
2078 (enum changelog_rec_extra_flags)
2079 (cref_want & ~(CLFE_UIDGID |
2083 cref = (enum changelog_rec_extra_flags)
2084 changelog_rec_extra_flags(rec)->cr_extra_flags;
2087 ef_mov = (char *)rec +
2088 changelog_rec_offset(
2089 (enum changelog_rec_flags)
2090 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
2091 jid_mov = (char *)rec +
2092 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2093 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
2095 rnm_mov = (char *)rec +
2096 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2102 /* Move the extension fields to the desired positions */
2103 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2104 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
2105 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
2106 memmove(xattr_mov, changelog_rec_xattr(rec),
2107 sizeof(struct changelog_ext_xattr));
2109 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
2110 memmove(omd_mov, changelog_rec_openmode(rec),
2111 sizeof(struct changelog_ext_openmode));
2113 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
2114 memmove(nid_mov, changelog_rec_nid(rec),
2115 sizeof(struct changelog_ext_nid));
2117 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
2118 memmove(uidgid_mov, changelog_rec_uidgid(rec),
2119 sizeof(struct changelog_ext_uidgid));
2121 memmove(ef_mov, changelog_rec_extra_flags(rec),
2122 sizeof(struct changelog_ext_extra_flags));
2125 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
2126 memmove(jid_mov, changelog_rec_jobid(rec),
2127 sizeof(struct changelog_ext_jobid));
2129 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
2130 memmove(rnm_mov, changelog_rec_rename(rec),
2131 sizeof(struct changelog_ext_rename));
2133 /* Clear newly added fields */
2134 if (xattr_mov && (cref_want & CLFE_XATTR) &&
2135 !(cref & CLFE_XATTR))
2136 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
2138 if (omd_mov && (cref_want & CLFE_OPEN) &&
2139 !(cref & CLFE_OPEN))
2140 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
2142 if (nid_mov && (cref_want & CLFE_NID) &&
2144 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
2146 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
2147 !(cref & CLFE_UIDGID))
2148 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
2150 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2151 !(rec->cr_flags & CLF_EXTRA_FLAGS))
2152 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
2154 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
2155 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
2157 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
2158 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
2160 /* Update the record's flags accordingly */
2161 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
2162 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2163 changelog_rec_extra_flags(rec)->cr_extra_flags =
2164 changelog_rec_extra_flags(rec)->cr_extra_flags |
2168 enum changelog_message_type {
2169 CL_RECORD = 10, /* message is a changelog_rec */
2170 CL_EOF = 11, /* at end of current changelog */
2173 /********* Misc **********/
2175 struct ioc_data_version {
2177 __u32 idv_layout_version; /* FLR: layout version for OST objects */
2178 __u32 idv_flags; /* enum ioc_data_version_flags */
2181 enum ioc_data_version_flags {
2182 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
2183 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
2184 LL_DV_SZ_UPDATE = (1 << 2), /* Update the file size on the client */
2188 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
2191 #define dot_lustre_name ".lustre"
2192 #define dot_fscrypt_name ".fscrypt"
2195 /********* HSM **********/
2197 /** HSM per-file state
2198 * See HSM_FLAGS below.
2201 HS_NONE = 0x00000000,
2202 HS_EXISTS = 0x00000001,
2203 HS_DIRTY = 0x00000002,
2204 HS_RELEASED = 0x00000004,
2205 HS_ARCHIVED = 0x00000008,
2206 HS_NORELEASE = 0x00000010,
2207 HS_NOARCHIVE = 0x00000020,
2208 HS_LOST = 0x00000040,
2211 /* HSM user-setable flags. */
2212 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
2214 /* Other HSM flags. */
2215 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
2218 * All HSM-related possible flags that could be applied to a file.
2219 * This should be kept in sync with hsm_states.
2221 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
2224 * HSM request progress state
2226 enum hsm_progress_states {
2233 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
2236 case HPS_WAITING: return "waiting";
2237 case HPS_RUNNING: return "running";
2238 case HPS_DONE: return "done";
2239 default: return "unknown";
2246 } __attribute__((packed));
2249 * Current HSM states of a Lustre file.
2251 * This structure purpose is to be sent to user-space mainly. It describes the
2252 * current HSM flags and in-progress action.
2254 struct hsm_user_state {
2255 /** Current HSM states, from enum hsm_states. */
2257 __u32 hus_archive_id;
2258 /** The current undergoing action, if there is one */
2259 __u32 hus_in_progress_state;
2260 __u32 hus_in_progress_action;
2261 struct hsm_extent hus_in_progress_location;
2262 char hus_extended_info[];
2265 struct hsm_state_set_ioc {
2266 struct lu_fid hssi_fid;
2268 __u64 hssi_clearmask;
2272 * This structure describes the current in-progress action for a file.
2273 * it is retuned to user space and send over the wire
2275 struct hsm_current_action {
2276 /** The current undergoing action, if there is one */
2277 /* state is one of hsm_progress_states */
2279 /* action is one of hsm_user_action */
2281 struct hsm_extent hca_location;
2284 /***** HSM user requests ******/
2285 /* User-generated (lfs/ioctl) request types */
2286 enum hsm_user_action {
2287 HUA_NONE = 1, /* no action (noop) */
2288 HUA_ARCHIVE = 10, /* copy to hsm */
2289 HUA_RESTORE = 11, /* prestage */
2290 HUA_RELEASE = 12, /* drop ost objects */
2291 HUA_REMOVE = 13, /* remove from archive */
2292 HUA_CANCEL = 14 /* cancel a request */
2295 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2298 case HUA_NONE: return "NOOP";
2299 case HUA_ARCHIVE: return "ARCHIVE";
2300 case HUA_RESTORE: return "RESTORE";
2301 case HUA_RELEASE: return "RELEASE";
2302 case HUA_REMOVE: return "REMOVE";
2303 case HUA_CANCEL: return "CANCEL";
2304 default: return "UNKNOWN";
2309 * List of hr_flags (bit field)
2311 #define HSM_FORCE_ACTION 0x0001
2312 /* used by CT, cannot be set by user */
2313 #define HSM_GHOST_COPY 0x0002
2316 * Contains all the fixed part of struct hsm_user_request.
2319 struct hsm_request {
2320 __u32 hr_action; /* enum hsm_user_action */
2321 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2322 __u64 hr_flags; /* request flags */
2323 __u32 hr_itemcount; /* item count in hur_user_item vector */
2327 struct hsm_user_item {
2328 struct lu_fid hui_fid;
2329 struct hsm_extent hui_extent;
2330 } __attribute__((packed));
2332 struct hsm_user_request {
2333 struct hsm_request hur_request;
2334 struct hsm_user_item hur_user_item[0];
2335 /* extra data blob at end of struct (after all
2336 * hur_user_items), only use helpers to access it
2338 } __attribute__((packed));
2340 /** Return pointer to data field in a hsm user request */
2341 static inline void *hur_data(struct hsm_user_request *hur)
2343 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2347 * Compute the current length of the provided hsm_user_request. This returns -1
2348 * instead of an errno because __kernel_ssize_t is defined to be only
2351 * return -1 on bounds check error.
2353 static inline __kernel_size_t hur_len(struct hsm_user_request *hur)
2357 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2358 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2359 (__u64)hur->hur_request.hr_itemcount *
2360 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2362 if ((__kernel_ssize_t)size < 0)
2368 /****** HSM RPCs to copytool *****/
2369 /* Message types the copytool may receive */
2370 enum hsm_message_type {
2371 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2374 /* Actions the copytool may be instructed to take for a given action_item */
2375 enum hsm_copytool_action {
2376 HSMA_NONE = 10, /* no action */
2377 HSMA_ARCHIVE = 20, /* arbitrary offset */
2383 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2386 case HSMA_NONE: return "NOOP";
2387 case HSMA_ARCHIVE: return "ARCHIVE";
2388 case HSMA_RESTORE: return "RESTORE";
2389 case HSMA_REMOVE: return "REMOVE";
2390 case HSMA_CANCEL: return "CANCEL";
2391 default: return "UNKNOWN";
2395 /* Copytool item action description */
2396 struct hsm_action_item {
2397 __u32 hai_len; /* valid size of this struct */
2398 __u32 hai_action; /* hsm_copytool_action, but use known size */
2399 struct lu_fid hai_fid; /* Lustre FID to operate on */
2400 struct lu_fid hai_dfid; /* fid used for data access */
2401 struct hsm_extent hai_extent; /* byte range to operate on */
2402 __u64 hai_cookie; /* action cookie from coordinator */
2403 __u64 hai_gid; /* grouplock id */
2404 char hai_data[0]; /* variable length */
2405 } __attribute__((packed));
2408 * helper function which print in hexa the first bytes of
2411 * \param hai [IN] record to print
2412 * \param buffer [IN,OUT] buffer to write the hex string to
2413 * \param len [IN] max buffer length
2417 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2418 char *buffer, __kernel_size_t len)
2425 data_len = hai->hai_len - sizeof(*hai);
2426 for (i = 0; (i < data_len) && (len > 2); i++) {
2427 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2437 /* Copytool action list */
2438 #define HAL_VERSION 1
2439 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2440 struct hsm_action_list {
2442 __u32 hal_count; /* number of hai's to follow */
2443 __u64 hal_compound_id; /* returned by coordinator, ignored */
2445 __u32 hal_archive_id; /* which archive backend */
2447 char hal_fsname[0]; /* null-terminated */
2448 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2449 boundaries. See hai_zero */
2450 } __attribute__((packed));
2452 /* Return pointer to first hai in action list */
2453 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2455 __kernel_size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2457 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2460 /* Return pointer to next hai */
2461 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2463 __kernel_size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2465 return (struct hsm_action_item *)((char *)hai + offset);
2468 /* Return size of an hsm_action_list */
2469 static inline __kernel_size_t hal_size(struct hsm_action_list *hal)
2473 struct hsm_action_item *hai;
2475 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2476 hai = hai_first(hal);
2477 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2478 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2484 * describe the attributes to be set on imported file
2486 struct hsm_user_import {
2495 __u32 hui_archive_id;
2498 /* Copytool progress reporting */
2499 #define HP_FLAG_COMPLETED 0x01
2500 #define HP_FLAG_RETRY 0x02
2502 struct hsm_progress {
2503 struct lu_fid hp_fid;
2505 struct hsm_extent hp_extent;
2507 __u16 hp_errval; /* positive val */
2512 __u64 hc_data_version;
2514 __u16 hc_errval; /* positive val */
2516 struct hsm_action_item hc_hai;
2519 enum lu_ladvise_type {
2520 LU_LADVISE_INVALID = 0,
2521 LU_LADVISE_WILLREAD = 1,
2522 LU_LADVISE_DONTNEED = 2,
2523 LU_LADVISE_LOCKNOEXPAND = 3,
2524 LU_LADVISE_LOCKAHEAD = 4,
2528 #define LU_LADVISE_NAMES { \
2529 [LU_LADVISE_WILLREAD] = "willread", \
2530 [LU_LADVISE_DONTNEED] = "dontneed", \
2531 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2532 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2535 /* This is the userspace argument for ladvise. It is currently the same as
2536 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2537 * may need info which is only used locally. */
2538 struct llapi_lu_ladvise {
2539 __u16 lla_advice; /* advice type */
2540 __u16 lla_value1; /* values for different advice types */
2542 __u64 lla_start; /* first byte of extent for advice */
2543 __u64 lla_end; /* last byte of extent for advice */
2549 LF_ASYNC = 0x00000001,
2550 LF_UNSET = 0x00000002,
2553 #define LADVISE_MAGIC 0x1ADF1CE0
2554 /* Masks of valid flags for each advice */
2555 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2556 /* Flags valid for all advices not explicitly specified */
2557 #define LF_DEFAULT_MASK LF_ASYNC
2559 #define LF_MASK (LF_ASYNC | LF_UNSET)
2561 #define lla_lockahead_mode lla_value1
2562 #define lla_peradvice_flags lla_value2
2563 #define lla_lockahead_result lla_value3
2565 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2566 * is used on the wire. It is defined separately as we may need info which is
2567 * only used locally. */
2568 struct llapi_ladvise_hdr {
2569 __u32 lah_magic; /* LADVISE_MAGIC */
2570 __u32 lah_count; /* number of advices */
2571 __u64 lah_flags; /* from enum ladvise_flag */
2572 __u32 lah_value1; /* unused */
2573 __u32 lah_value2; /* unused */
2574 __u64 lah_value3; /* unused */
2575 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2578 #define LAH_COUNT_MAX (1024)
2582 SK_CRYPT_INVALID = -1,
2584 SK_CRYPT_AES256_CTR = 1,
2588 SK_HMAC_INVALID = -1,
2594 struct sk_crypt_type {
2595 const char *sct_name;
2599 struct sk_hmac_type {
2600 const char *sht_name;
2604 enum lock_mode_user {
2610 #define LOCK_MODE_NAMES { \
2611 [MODE_READ_USER] = "READ",\
2612 [MODE_WRITE_USER] = "WRITE"\
2615 enum lockahead_results {
2616 LLA_RESULT_SENT = 0,
2617 LLA_RESULT_DIFFERENT,
2621 enum lu_heat_flag_bit {
2622 LU_HEAT_FLAG_BIT_INVALID = 0,
2623 LU_HEAT_FLAG_BIT_OFF,
2624 LU_HEAT_FLAG_BIT_CLEAR,
2628 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2629 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2632 enum obd_heat_type {
2633 OBD_HEAT_READSAMPLE = 0,
2634 OBD_HEAT_WRITESAMPLE = 1,
2635 OBD_HEAT_READBYTE = 2,
2636 OBD_HEAT_WRITEBYTE = 3,
2640 #define LU_HEAT_NAMES { \
2641 [OBD_HEAT_READSAMPLE] = "readsample", \
2642 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2643 [OBD_HEAT_READBYTE] = "readbyte", \
2644 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2659 static inline const char *pcc_type2string(enum lu_pcc_type type)
2664 case LU_PCC_READWRITE:
2671 struct lu_pcc_attach {
2672 __u32 pcca_type; /* PCC type */
2673 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2676 enum lu_pcc_detach_opts {
2677 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2678 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2681 struct lu_pcc_detach_fid {
2682 /* fid of the file to detach */
2683 struct lu_fid pccd_fid;
2687 struct lu_pcc_detach {
2691 enum lu_pcc_state_flags {
2692 PCC_STATE_FL_NONE = 0x0,
2693 /* The inode attr is cached locally */
2694 PCC_STATE_FL_ATTR_VALID = 0x01,
2695 /* The file is being attached into PCC */
2696 PCC_STATE_FL_ATTACHING = 0x02,
2699 struct lu_pcc_state {
2700 __u32 pccs_type; /* enum lu_pcc_type */
2701 __u32 pccs_open_count;
2702 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2704 char pccs_path[PATH_MAX];
2707 enum lu_project_type {
2708 LU_PROJECT_NONE = 0,
2715 __u32 project_type; /* enum lu_project_type */
2717 __u32 project_xflags;
2718 __u32 project_reserved;
2719 char project_name[NAME_MAX + 1];
2724 /* make header's size equal lu_fid */
2727 struct lu_fid fa_fids[0];
2729 #define OBD_MAX_FIDS_IN_ARRAY 4096
2731 /* more types could be defined upon need for more complex
2732 * format to be used in foreign symlink LOV/LMV EAs, like
2733 * one to describe a delimiter string and occurence number
2734 * of delimited sub-string, ...
2736 enum ll_foreign_symlink_upcall_item_type {
2742 /* may need to be modified to allow for more format items to be defined, and
2743 * like for ll_foreign_symlink_upcall_item_type enum
2745 struct ll_foreign_symlink_upcall_item {
2755 /* internal storage of constant string */
2757 /* upcall stores constant string in a raw */
2764 #define POSLEN_ITEM_SZ (offsetof(struct ll_foreign_symlink_upcall_item, len) + \
2765 sizeof(((struct ll_foreign_symlink_upcall_item *)0)->len))
2766 #define STRING_ITEM_SZ(sz) ( \
2767 offsetof(struct ll_foreign_symlink_upcall_item, bytestring) + \
2768 (sz + sizeof(__u32) - 1) / sizeof(__u32) * sizeof(__u32))
2770 /* presently limited to not cause max stack frame size to be reached
2771 * because of temporary automatic array of
2772 * "struct ll_foreign_symlink_upcall_item" presently used in
2773 * foreign_symlink_upcall_info_store()
2775 #define MAX_NB_UPCALL_ITEMS 32
2777 #if defined(__cplusplus)
2781 /** @} lustreuser */
2783 #endif /* _LUSTRE_USER_H */