4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_user.h
34 * Lustre public user-space interface definitions.
37 #ifndef _LUSTRE_USER_H
38 #define _LUSTRE_USER_H
40 /** \defgroup lustreuser lustreuser
45 #include <linux/kernel.h>
46 #include <linux/types.h>
49 # include <linux/fs.h>
50 # include <linux/quota.h>
51 # include <linux/string.h> /* snprintf() */
52 # include <linux/version.h>
53 # include <uapi/linux/lustre/lustre_fiemap.h>
54 #else /* !__KERNEL__ */
57 # include <stdio.h> /* snprintf() */
59 # define NEED_QUOTA_DEFS
60 /* # include <sys/quota.h> - this causes complaints about caddr_t */
61 # include <sys/stat.h>
62 # include <linux/lustre/lustre_fiemap.h>
63 #endif /* __KERNEL__ */
65 /* Handle older distros */
66 #ifndef __ALIGN_KERNEL
67 # define __ALIGN_KERNEL(x, a) __ALIGN_KERNEL_MASK(x, (typeof(x))(a) - 1)
68 # define __ALIGN_KERNEL_MASK(x, mask) (((x) + (mask)) & ~(mask))
71 #if defined(__cplusplus)
76 * This is a temporary solution of adding quota type.
77 * Should be removed as soon as system header is updated.
80 #define LL_MAXQUOTAS 3
82 #define INITQFNAMES { \
83 "user", /* USRQUOTA */ \
84 "group", /* GRPQUOTA */ \
85 "project", /* PRJQUOTA */ \
99 * We need to always use 64bit version because the structure
100 * is shared across entire cluster where 32bit and 64bit machines
103 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
104 typedef struct stat64 lstat_t;
105 #define lstat_f lstat64
106 #define fstat_f fstat64
107 #define fstatat_f fstatat64
109 typedef struct stat lstat_t;
110 #define lstat_f lstat
111 #define fstat_f fstat
112 #define fstatat_f fstatat
115 #ifndef STATX_BASIC_STATS
117 * Timestamp structure for the timestamps in struct statx.
119 * tv_sec holds the number of seconds before (negative) or after (positive)
120 * 00:00:00 1st January 1970 UTC.
122 * tv_nsec holds a number of nanoseconds (0..999,999,999) after the tv_sec time.
124 * __reserved is held in case we need a yet finer resolution.
126 struct statx_timestamp {
133 * Structures for the extended file attribute retrieval system call
136 * The caller passes a mask of what they're specifically interested in as a
137 * parameter to statx(). What statx() actually got will be indicated in
138 * st_mask upon return.
140 * For each bit in the mask argument:
142 * - if the datum is not supported:
144 * - the bit will be cleared, and
146 * - the datum will be set to an appropriate fabricated value if one is
147 * available (eg. CIFS can take a default uid and gid), otherwise
149 * - the field will be cleared;
151 * - otherwise, if explicitly requested:
153 * - the datum will be synchronised to the server if AT_STATX_FORCE_SYNC is
154 * set or if the datum is considered out of date, and
156 * - the field will be filled in and the bit will be set;
158 * - otherwise, if not requested, but available in approximate form without any
159 * effort, it will be filled in anyway, and the bit will be set upon return
160 * (it might not be up to date, however, and no attempt will be made to
161 * synchronise the internal state first);
163 * - otherwise the field and the bit will be cleared before returning.
165 * Items in STATX_BASIC_STATS may be marked unavailable on return, but they
166 * will have values installed for compatibility purposes so that stat() and
167 * co. can be emulated in userspace.
171 __u32 stx_mask; /* What results were written [uncond] */
172 __u32 stx_blksize; /* Preferred general I/O size [uncond] */
173 __u64 stx_attributes; /* Flags conveying information about the file [uncond] */
175 __u32 stx_nlink; /* Number of hard links */
176 __u32 stx_uid; /* User ID of owner */
177 __u32 stx_gid; /* Group ID of owner */
178 __u16 stx_mode; /* File mode */
181 __u64 stx_ino; /* Inode number */
182 __u64 stx_size; /* File size */
183 __u64 stx_blocks; /* Number of 512-byte blocks allocated */
184 __u64 stx_attributes_mask; /* Mask to show what's supported in stx_attributes */
186 struct statx_timestamp stx_atime; /* Last access time */
187 struct statx_timestamp stx_btime; /* File creation time */
188 struct statx_timestamp stx_ctime; /* Last attribute change time */
189 struct statx_timestamp stx_mtime; /* Last data modification time */
191 __u32 stx_rdev_major; /* Device ID of special file [if bdev/cdev] */
192 __u32 stx_rdev_minor;
193 __u32 stx_dev_major; /* ID of device containing file [uncond] */
196 __u64 __spare2[14]; /* Spare space for future expansion */
201 * Flags to be stx_mask
203 * Query request/result mask for statx() and struct statx::stx_mask.
205 * These bits should be set in the mask argument of statx() to request
206 * particular items when calling statx().
208 #define STATX_TYPE 0x00000001U /* Want/got stx_mode & S_IFMT */
209 #define STATX_MODE 0x00000002U /* Want/got stx_mode & ~S_IFMT */
210 #define STATX_NLINK 0x00000004U /* Want/got stx_nlink */
211 #define STATX_UID 0x00000008U /* Want/got stx_uid */
212 #define STATX_GID 0x00000010U /* Want/got stx_gid */
213 #define STATX_ATIME 0x00000020U /* Want/got stx_atime */
214 #define STATX_MTIME 0x00000040U /* Want/got stx_mtime */
215 #define STATX_CTIME 0x00000080U /* Want/got stx_ctime */
216 #define STATX_INO 0x00000100U /* Want/got stx_ino */
217 #define STATX_SIZE 0x00000200U /* Want/got stx_size */
218 #define STATX_BLOCKS 0x00000400U /* Want/got stx_blocks */
219 #define STATX_BASIC_STATS 0x000007ffU /* The stuff in the normal stat struct */
220 #define STATX_BTIME 0x00000800U /* Want/got stx_btime */
221 #define STATX_ALL 0x00000fffU /* All currently supported flags */
222 #define STATX__RESERVED 0x80000000U /* Reserved for future struct statx expansion */
225 * Attributes to be found in stx_attributes and masked in stx_attributes_mask.
227 * These give information about the features or the state of a file that might
228 * be of use to ordinary userspace programs such as GUIs or ls rather than
231 * Note that the flags marked [I] correspond to generic FS_IOC_FLAGS
232 * semantically. Where possible, the numerical value is picked to correspond
235 #define STATX_ATTR_COMPRESSED 0x00000004 /* [I] File is compressed by the fs */
236 #define STATX_ATTR_IMMUTABLE 0x00000010 /* [I] File is marked immutable */
237 #define STATX_ATTR_APPEND 0x00000020 /* [I] File is append-only */
238 #define STATX_ATTR_NODUMP 0x00000040 /* [I] File is not to be dumped */
239 #define STATX_ATTR_ENCRYPTED 0x00000800 /* [I] File requires key to decrypt in fs */
241 #define STATX_ATTR_AUTOMOUNT 0x00001000 /* Dir: Automount trigger */
245 typedef struct statx lstatx_t;
247 #define HAVE_LOV_USER_MDS_DATA
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_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
261 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
262 OS_STATE_NOPRECREATE = 0x00000004, /**< no object precreation */
263 OS_STATE_UNUSED1 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
264 OS_STATE_UNUSED2 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
265 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
266 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
267 OS_STATE_SUM = 0x00000100, /**< aggregated for all tagrets */
268 OS_STATE_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_STATE_* flag */
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() */
300 * FID is a cluster-wide unique identifier of a file or an object (stripe).
301 * FIDs are never reused.
305 * FID sequence. Sequence is a unit of migration: all files (objects)
306 * with FIDs from a given sequence are stored on the same server.
307 * Lustre should support 2^64 objects, so even if each sequence
308 * has only a single object we can still enumerate 2^64 objects.
311 /* FID number within sequence. */
314 * FID version, used to distinguish different versions (in the sense
315 * of snapshots, etc.) of the same file system object. Not currently
321 static inline bool fid_is_zero(const struct lu_fid *fid)
323 return fid->f_seq == 0 && fid->f_oid == 0;
326 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
327 * MDT-object's FID::f_ver, instead it is the OST-object index in its
328 * parent MDT-object's layout EA. */
329 #define f_stripe_idx f_ver
332 __u32 ol_stripe_size;
333 __u32 ol_stripe_count;
337 } __attribute__((packed));
339 /* The filter_fid structure has changed several times over its lifetime.
340 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
341 * stripe_index and the "self FID" (objid/seq) to be able to recover the
342 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
343 * the OST, the "trusted.lma" xattr was added to the OST objects to store
344 * the "self FID" to be consistent with the MDT on-disk format, and the
345 * filter_fid only stored the MDT inode parent FID and stripe index.
347 * In 2.10, the addition of PFL composite layouts required more information
348 * to be stored into the filter_fid in order to be able to identify which
349 * component the OST object belonged. As well, the stripe size may vary
350 * between components, so it was no longer safe to assume the stripe size
351 * or stripe_count of a file. This is also more robust for plain layouts.
353 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
354 * enough space to store both the filter_fid and LMA in the inode, so they
355 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
356 * an extra seek for every OST object access.
358 * In 2.11, FLR mirror layouts also need to store the layout version and
359 * range so that writes to old versions of the layout are not allowed.
360 * That ensures that mirrored objects are not modified by evicted clients,
361 * and ensures that the components are correctly marked stale on the MDT.
363 struct filter_fid_18_23 {
364 struct lu_fid ff_parent; /* stripe_idx in f_ver */
369 struct filter_fid_24_29 {
370 struct lu_fid ff_parent; /* stripe_idx in f_ver */
373 struct filter_fid_210 {
374 struct lu_fid ff_parent; /* stripe_idx in f_ver */
375 struct ost_layout ff_layout;
379 struct lu_fid ff_parent; /* stripe_idx in f_ver */
380 struct ost_layout ff_layout;
381 __u32 ff_layout_version;
382 __u32 ff_range; /* range of layout version that
383 * write are allowed */
384 } __attribute__((packed));
386 /* Userspace should treat lu_fid as opaque, and only use the following methods
387 * to print or parse them. Other functions (e.g. compare, swab) could be moved
388 * here from lustre_idl.h if needed. */
392 LMAC_HSM = 0x00000001,
393 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
394 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
395 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
396 * under /O/<seq>/d<x>. */
397 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
398 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
399 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
403 * Masks for all features that should be supported by a Lustre version to
404 * access a specific file.
405 * This information is stored in lustre_mdt_attrs::lma_incompat.
408 LMAI_RELEASED = 0x00000001, /* file is released */
409 LMAI_AGENT = 0x00000002, /* agent inode */
410 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
411 is on the remote MDT */
412 LMAI_STRIPED = 0x00000008, /* striped directory inode */
413 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
414 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
415 LMAI_STRIPED | LMAI_ORPHAN)
420 * Following struct for object attributes, that will be kept inode's EA.
421 * Introduced in 2.0 release (please see b15993, for details)
422 * Added to all objects since Lustre 2.4 as contains self FID
424 struct lustre_mdt_attrs {
426 * Bitfield for supported data in this structure. From enum lma_compat.
427 * lma_self_fid and lma_flags are always available.
431 * Per-file incompat feature list. Lustre version should support all
432 * flags set in this field. The supported feature mask is available in
436 /** FID of this inode */
437 struct lu_fid lma_self_fid;
440 struct lustre_ost_attrs {
441 /* Use lustre_mdt_attrs directly for now, need a common header
442 * structure if want to change lustre_mdt_attrs in future. */
443 struct lustre_mdt_attrs loa_lma;
445 /* Below five elements are for OST-object's PFID EA, the
446 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
447 * and the stripe_index (low 16 bits), the size should not exceed
448 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
449 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
450 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
451 * loa_comp_* elements are valid. */
452 struct lu_fid loa_parent_fid;
453 __u32 loa_stripe_size;
455 __u64 loa_comp_start;
460 * Prior to 2.4, the LMA structure also included SOM attributes which has since
461 * been moved to a dedicated xattr
462 * lma_flags was also removed because of lma_compat/incompat fields.
464 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
466 enum lustre_som_flags {
467 /* Unknow or no SoM data, must get size from OSTs. */
468 SOM_FL_UNKNOWN = 0x0000,
469 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
470 SOM_FL_STRICT = 0x0001,
471 /* Known stale - was right at some point in the past, but it is
472 * known (or likely) to be incorrect now (e.g. opened for write). */
473 SOM_FL_STALE = 0x0002,
474 /* Approximate, may never have been strictly correct,
475 * need to sync SOM data to achieve eventual consistency. */
476 SOM_FL_LAZY = 0x0004,
479 struct lustre_som_attrs {
481 __u16 lsa_reserved[3];
487 * OST object IDentifier.
495 struct lu_fid oi_fid;
499 #define DOSTID "%#llx:%llu"
500 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
501 ((unsigned long long)ostid_id(oi))
503 struct ll_futimes_3 {
505 __u64 lfu_atime_nsec;
507 __u64 lfu_mtime_nsec;
509 __u64 lfu_ctime_nsec;
513 * Maximum number of mirrors currently implemented.
515 #define LUSTRE_MIRROR_COUNT_MAX 16
517 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
519 LL_LEASE_RDLCK = 0x01,
520 LL_LEASE_WRLCK = 0x02,
521 LL_LEASE_UNLCK = 0x04,
524 enum ll_lease_flags {
525 LL_LEASE_RESYNC = 0x1,
526 LL_LEASE_RESYNC_DONE = 0x2,
527 LL_LEASE_LAYOUT_MERGE = 0x4,
528 LL_LEASE_LAYOUT_SPLIT = 0x8,
531 #define IOC_IDS_MAX 4096
532 struct ll_ioc_lease {
539 struct ll_ioc_lease_id {
550 * The ioctl naming rules:
551 * LL_* - works on the currently opened filehandle instead of parent dir
552 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
553 * *_MDC_* - gets/sets data related to MDC
554 * *_LOV_* - gets/sets data related to OSC/LOV
555 * *FILE* - called on parent dir and passes in a filename
556 * *STRIPE* - set/get lov_user_md
557 * *INFO - set/get lov_user_mds_data
559 /* lustre_ioctl.h 101-150 */
560 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
561 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
562 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
563 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
564 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
565 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
566 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
567 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
568 /* LL_IOC_RECREATE_OBJ 157 obsolete */
569 /* LL_IOC_RECREATE_FID 157 obsolete */
570 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
571 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
572 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
573 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
574 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
575 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
576 /* IOC_LOV_GETINFO 165 obsolete */
577 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
578 /* LL_IOC_RMTACL 167 obsolete */
579 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
580 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
581 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
582 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
583 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
584 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
585 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
586 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
587 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
588 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
589 /* lustre_ioctl.h 177-210 */
590 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
591 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
592 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
593 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
594 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
595 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
596 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
597 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
598 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
599 struct lustre_swap_layouts)
600 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
601 struct hsm_current_action)
602 /* lustre_ioctl.h 221-232 */
603 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
604 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
605 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
606 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
607 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
608 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
609 #define LL_IOC_GET_LEASE _IO('f', 244)
610 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
611 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
612 #define LL_IOC_MIGRATE _IOR('f', 247, int)
613 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
614 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
615 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
617 #ifndef FS_IOC_FSGETXATTR
619 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
622 __u32 fsx_xflags; /* xflags field value (get/set) */
623 __u32 fsx_extsize; /* extsize field value (get/set)*/
624 __u32 fsx_nextents; /* nextents field value (get) */
625 __u32 fsx_projid; /* project identifier (get/set) */
626 unsigned char fsx_pad[12];
628 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
629 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
631 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
632 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
633 #ifndef FS_XFLAG_PROJINHERIT
634 #define FS_XFLAG_PROJINHERIT 0x00000200
638 #define LL_STATFS_LMV 1
639 #define LL_STATFS_LOV 2
640 #define LL_STATFS_NODELAY 4
642 #define IOC_MDC_TYPE 'i'
643 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
644 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
645 #ifdef HAVE_LOV_USER_MDS_DATA
646 #define IOC_MDC_GETFILEINFO_OLD _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
647 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
648 #define LL_IOC_MDC_GETINFO_OLD _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
649 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
652 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
654 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
655 * files, but are unlikely to be used in practice and are not harmful if
656 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
657 * devices and are safe for use on new files. See LU-4209. */
658 /* To be compatible with old statically linked binary we keep the check for
659 * the older 0100000000 flag. This is already removed upstream. LU-812. */
660 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
662 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
664 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
665 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
666 O_LOV_DELAY_CREATE_MASK)
668 #define LL_FILE_IGNORE_LOCK 0x00000001
669 #define LL_FILE_GROUP_LOCKED 0x00000002
670 #define LL_FILE_READAHEA 0x00000004
671 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
672 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
674 #define LOV_USER_MAGIC_V1 0x0BD10BD0
675 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
676 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
677 #define LOV_USER_MAGIC_V3 0x0BD30BD0
678 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
679 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
680 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
682 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
683 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
684 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
686 #define LOV_PATTERN_NONE 0x000
687 #define LOV_PATTERN_RAID0 0x001
688 #define LOV_PATTERN_RAID1 0x002
689 #define LOV_PATTERN_MDT 0x100
690 #define LOV_PATTERN_CMOBD 0x200
692 #define LOV_PATTERN_F_MASK 0xffff0000
693 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
694 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
695 #define LOV_PATTERN_DEFAULT 0xffffffff
697 static inline bool lov_pattern_supported(__u32 pattern)
699 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
700 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
703 #define LOV_MAXPOOLNAME 15
704 #define LOV_POOLNAMEF "%.15s"
706 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
707 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
708 #define LOV_MAX_STRIPE_COUNT_OLD 160
709 /* This calculation is crafted so that input of 4096 will result in 160
710 * which in turn is equal to old maximal stripe count.
711 * XXX: In fact this is too simpified for now, what it also need is to get
712 * ea_type argument to clearly know how much space each stripe consumes.
714 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
715 * allocation that is sufficient for the current generation of systems.
717 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
718 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
719 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
720 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
722 #define XATTR_LUSTRE_PREFIX "lustre."
723 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
725 #define lov_user_ost_data lov_user_ost_data_v1
726 struct lov_user_ost_data_v1 { /* per-stripe data structure */
727 struct ost_id l_ost_oi; /* OST object ID */
728 __u32 l_ost_gen; /* generation of this OST index */
729 __u32 l_ost_idx; /* OST index in LOV */
730 } __attribute__((packed));
732 #define lov_user_md lov_user_md_v1
733 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
734 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
735 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
736 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
737 __u32 lmm_stripe_size; /* size of stripe in bytes */
738 __u16 lmm_stripe_count; /* num stripes in use for this object */
740 __u16 lmm_stripe_offset; /* starting stripe offset in
741 * lmm_objects, use when writing */
742 __u16 lmm_layout_gen; /* layout generation number
743 * used when reading */
745 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
746 } __attribute__((packed, __may_alias__));
748 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
749 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
750 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
751 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
752 __u32 lmm_stripe_size; /* size of stripe in bytes */
753 __u16 lmm_stripe_count; /* num stripes in use for this object */
755 __u16 lmm_stripe_offset; /* starting stripe offset in
756 * lmm_objects, use when writing */
757 __u16 lmm_layout_gen; /* layout generation number
758 * used when reading */
760 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
761 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
762 } __attribute__((packed));
769 #define DEXT "[%#llx, %#llx)"
770 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
772 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
773 struct lu_extent *e2)
775 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
778 static inline bool lu_extent_is_whole(struct lu_extent *e)
780 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
783 enum lov_comp_md_entry_flags {
784 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
785 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
786 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
787 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
788 LCME_FL_OFFLINE = 0x00000008, /* Not used */
789 LCME_FL_INIT = 0x00000010, /* instantiated */
790 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
791 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
792 won't be stored on disk */
795 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
796 LCME_FL_PREF_RW | LCME_FL_NOSYNC)
797 /* The flags can be set by users at mirror creation time. */
798 #define LCME_USER_FLAGS (LCME_FL_PREF_RW)
800 /* The flags are for mirrors */
801 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
803 /* These flags have meaning when set in a default layout and will be inherited
804 * from the default/template layout set on a directory.
806 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC)
808 /* the highest bit in obdo::o_layout_version is used to mark if the file is
810 #define LU_LAYOUT_RESYNC LCME_FL_NEG
812 /* lcme_id can be specified as certain flags, and the the first
813 * bit of lcme_id is used to indicate that the ID is representing
814 * certain LCME_FL_* but not a real ID. Which implies we can have
815 * at most 31 flags (see LCME_FL_XXX). */
818 LCME_ID_MAX = 0x7FFFFFFF,
819 LCME_ID_ALL = 0xFFFFFFFF,
820 LCME_ID_NOT_ID = LCME_FL_NEG
823 #define LCME_ID_MASK LCME_ID_MAX
825 struct lov_comp_md_entry_v1 {
826 __u32 lcme_id; /* unique id of component */
827 __u32 lcme_flags; /* LCME_FL_XXX */
828 struct lu_extent lcme_extent; /* file extent for component */
829 __u32 lcme_offset; /* offset of component blob,
830 start from lov_comp_md_v1 */
831 __u32 lcme_size; /* size of component blob */
832 __u32 lcme_layout_gen;
833 __u64 lcme_timestamp; /* snapshot time if applicable*/
834 __u32 lcme_padding_1;
835 } __attribute__((packed));
837 #define SEQ_ID_MAX 0x0000FFFF
838 #define SEQ_ID_MASK SEQ_ID_MAX
839 /* bit 30:16 of lcme_id is used to store mirror id */
840 #define MIRROR_ID_MASK 0x7FFF0000
841 #define MIRROR_ID_NEG 0x8000
842 #define MIRROR_ID_SHIFT 16
844 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
846 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
849 static inline __u16 mirror_id_of(__u32 id)
851 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
855 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
857 enum lov_comp_md_flags {
858 /* the least 2 bits are used by FLR to record file state */
861 LCM_FL_WRITE_PENDING = 2,
862 LCM_FL_SYNC_PENDING = 3,
863 LCM_FL_FLR_MASK = 0x3,
866 struct lov_comp_md_v1 {
867 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
868 __u32 lcm_size; /* overall size including this struct */
869 __u32 lcm_layout_gen;
871 __u16 lcm_entry_count;
872 /* lcm_mirror_count stores the number of actual mirrors minus 1,
873 * so that non-flr files will have value 0 meaning 1 mirror. */
874 __u16 lcm_mirror_count;
875 __u16 lcm_padding1[3];
877 struct lov_comp_md_entry_v1 lcm_entries[0];
878 } __attribute__((packed));
880 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
882 if (stripes == (__u16)-1)
885 if (lmm_magic == LOV_USER_MAGIC_V1)
886 return sizeof(struct lov_user_md_v1) +
887 stripes * sizeof(struct lov_user_ost_data_v1);
888 return sizeof(struct lov_user_md_v3) +
889 stripes * sizeof(struct lov_user_ost_data_v1);
892 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
893 * use this. It is unsafe to #define those values in this header as it
894 * is possible the application has already #included <sys/stat.h>. */
895 #ifdef HAVE_LOV_USER_MDS_DATA
896 #define lov_user_mds_data lov_user_mds_data_v2
897 struct lov_user_mds_data_v1 {
898 lstat_t lmd_st; /* MDS stat struct */
899 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
900 } __attribute__((packed));
902 struct lov_user_mds_data_v2 {
903 struct lu_fid lmd_fid; /* Lustre FID */
904 lstatx_t lmd_stx; /* MDS statx struct */
905 __u64 lmd_flags; /* MDS stat flags */
906 __u32 lmd_lmmsize; /* LOV EA size */
907 __u32 lmd_padding; /* unused */
908 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
909 } __attribute__((packed));
912 struct lmv_user_mds_data {
913 struct lu_fid lum_fid;
919 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
920 LMV_HASH_TYPE_ALL_CHARS = 1,
921 LMV_HASH_TYPE_FNV_1A_64 = 2,
925 #define LMV_HASH_NAME_ALL_CHARS "all_char"
926 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
928 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
930 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
931 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
932 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
933 #define lmv_user_md lmv_user_md_v1
934 struct lmv_user_md_v1 {
935 __u32 lum_magic; /* must be the first field */
936 __u32 lum_stripe_count; /* dirstripe count */
937 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
938 __u32 lum_hash_type; /* Dir stripe policy */
939 __u32 lum_type; /* LMV type: default or normal */
943 char lum_pool_name[LOV_MAXPOOLNAME + 1];
944 struct lmv_user_mds_data lum_objects[0];
945 } __attribute__((packed));
947 static inline int lmv_user_md_size(int stripes, int lmm_magic)
949 int size = sizeof(struct lmv_user_md);
951 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
952 size += stripes * sizeof(struct lmv_user_mds_data);
957 struct ll_recreate_obj {
963 __u64 id; /* holds object id */
964 __u32 generation; /* holds object generation */
965 __u32 f_type; /* holds object type or stripe idx when passing it to
966 * OST for saving into EA. */
974 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
975 const struct obd_uuid *u2)
977 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
980 static inline int obd_uuid_empty(struct obd_uuid *uuid)
982 return uuid->uuid[0] == '\0';
985 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
987 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
988 uuid->uuid[sizeof(*uuid) - 1] = '\0';
991 /* For printf's only, make sure uuid is terminated */
992 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
997 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
998 /* Obviously not safe, but for printfs, no real harm done...
999 we're always null-terminated, even in a race. */
1000 static char temp[sizeof(*uuid)];
1001 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
1002 temp[sizeof(*uuid) - 1] = '\0';
1005 return (char *)(uuid->uuid);
1008 #define LUSTRE_MAXFSNAME 8
1010 /* Extract fsname from uuid (or target name) of a target
1011 e.g. (myfs-OST0007_UUID -> myfs)
1012 see also deuuidify. */
1013 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1017 strncpy(buf, uuid, buflen - 1);
1018 buf[buflen - 1] = '\0';
1019 p = strrchr(buf, '-');
1024 /* printf display format for Lustre FIDs
1025 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1026 #define FID_NOBRACE_LEN 40
1027 #define FID_LEN (FID_NOBRACE_LEN + 2)
1028 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1029 #define DFID "["DFID_NOBRACE"]"
1030 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1032 /* scanf input parse format for fids in DFID_NOBRACE format
1033 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1034 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1035 #define SFID "0x%llx:0x%x:0x%x"
1036 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1038 /********* Quotas **********/
1040 #define LUSTRE_QUOTABLOCK_BITS 10
1041 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1043 static inline __u64 lustre_stoqb(size_t space)
1045 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1048 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1049 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1050 #define Q_GETOINFO 0x800102 /* get obd quota info */
1051 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1052 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1054 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1055 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1056 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1057 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1058 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1059 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1060 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1061 /* lustre-specific control commands */
1062 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1063 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1064 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1065 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1067 /* In the current Lustre implementation, the grace time is either the time
1068 * or the timestamp to be used after some quota ID exceeds the soft limt,
1069 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1071 #define LQUOTA_GRACE_BITS 48
1072 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1073 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1074 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1075 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1076 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1078 /* different quota flags */
1080 /* the default quota flag, the corresponding quota ID will use the default
1081 * quota setting, the hardlimit and softlimit of its quota record in the global
1082 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1083 * and high 16 bits will contain this flag (see above comment).
1085 #define LQUOTA_FLAG_DEFAULT 0x0001
1087 #define ALLQUOTA 255 /* set all quota */
1088 static inline char *qtype_name(int qtype)
1101 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1102 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1105 #define N_PERMS_MAX 64
1107 struct perm_downcall_data {
1113 struct identity_downcall_data {
1120 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1121 __u32 idd_groups[0];
1124 struct sepol_downcall_data {
1126 time_t sdd_sepol_mtime;
1127 __u16 sdd_sepol_len;
1131 #ifdef NEED_QUOTA_DEFS
1133 #define QIF_BLIMITS 1
1135 #define QIF_ILIMITS 4
1136 #define QIF_INODES 8
1137 #define QIF_BTIME 16
1138 #define QIF_ITIME 32
1139 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1140 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1141 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1142 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1145 #endif /* !__KERNEL__ */
1147 /* lustre volatile file support
1148 * file name header: ".^L^S^T^R:volatile"
1150 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1151 #define LUSTRE_VOLATILE_HDR_LEN 14
1153 enum lustre_quota_version {
1157 /* XXX: same as if_dqinfo struct in kernel */
1165 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1167 __u64 dqb_bhardlimit;
1168 __u64 dqb_bsoftlimit;
1170 __u64 dqb_ihardlimit;
1171 __u64 dqb_isoftlimit;
1172 __u64 dqb_curinodes;
1186 struct if_quotactl {
1193 struct obd_dqinfo qc_dqinfo;
1194 struct obd_dqblk qc_dqblk;
1196 struct obd_uuid obd_uuid;
1199 /* swap layout flags */
1200 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1201 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1202 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1203 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1204 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1206 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1207 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1208 struct lustre_swap_layouts {
1216 /** Bit-mask of valid attributes */
1217 /* The LA_* flags are written to disk as part of the ChangeLog records
1218 * so they are part of the on-disk and network protocol, and cannot be changed.
1219 * Only the first 12 bits are currently saved.
1222 LA_ATIME = 1 << 0, /* 0x00001 */
1223 LA_MTIME = 1 << 1, /* 0x00002 */
1224 LA_CTIME = 1 << 2, /* 0x00004 */
1225 LA_SIZE = 1 << 3, /* 0x00008 */
1226 LA_MODE = 1 << 4, /* 0x00010 */
1227 LA_UID = 1 << 5, /* 0x00020 */
1228 LA_GID = 1 << 6, /* 0x00040 */
1229 LA_BLOCKS = 1 << 7, /* 0x00080 */
1230 LA_TYPE = 1 << 8, /* 0x00100 */
1231 LA_FLAGS = 1 << 9, /* 0x00200 */
1232 LA_NLINK = 1 << 10, /* 0x00400 */
1233 LA_RDEV = 1 << 11, /* 0x00800 */
1234 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1235 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1236 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1237 LA_PROJID = 1 << 15, /* 0x08000 */
1238 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1239 LA_LSIZE = 1 << 17, /* 0x20000 */
1240 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1242 * Attributes must be transmitted to OST objects
1244 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1247 #define MDS_FMODE_READ 00000001
1248 #define MDS_FMODE_WRITE 00000002
1250 #define MDS_FMODE_CLOSED 00000000
1251 #define MDS_FMODE_EXEC 00000004
1252 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1253 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1254 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1256 #define MDS_OPEN_CREATED 00000010
1257 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1259 #define MDS_OPEN_CREAT 00000100
1260 #define MDS_OPEN_EXCL 00000200
1261 #define MDS_OPEN_TRUNC 00001000
1262 #define MDS_OPEN_APPEND 00002000
1263 #define MDS_OPEN_SYNC 00010000
1264 #define MDS_OPEN_DIRECTORY 00200000
1266 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1267 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1268 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1269 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1270 * We do not support JOIN FILE
1271 * anymore, reserve this flags
1272 * just for preventing such bit
1275 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1276 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1277 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1278 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1279 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1281 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1283 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1284 * delegation, succeed if it's not
1285 * being opened with conflict mode.
1287 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1289 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1291 /* lustre internal open flags, which should not be set from user space */
1292 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1293 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1294 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1295 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC)
1298 /********* Changelogs **********/
1299 /** Changelog record types */
1300 enum changelog_rec_type {
1303 CL_CREATE = 1, /* namespace */
1304 CL_MKDIR = 2, /* namespace */
1305 CL_HARDLINK = 3, /* namespace */
1306 CL_SOFTLINK = 4, /* namespace */
1307 CL_MKNOD = 5, /* namespace */
1308 CL_UNLINK = 6, /* namespace */
1309 CL_RMDIR = 7, /* namespace */
1310 CL_RENAME = 8, /* namespace */
1311 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1312 CL_OPEN = 10, /* not currently used */
1313 CL_CLOSE = 11, /* may be written to log only with mtime change */
1314 CL_LAYOUT = 12, /* file layout/striping modified */
1318 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1319 CL_HSM = 16, /* HSM specific events, see flags */
1320 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1324 CL_FLRW = 21, /* FLR: file was firstly written */
1325 CL_RESYNC = 22, /* FLR: file was resync-ed */
1327 CL_DN_OPEN = 24, /* denied open */
1331 static inline const char *changelog_type2str(int type) {
1332 static const char *changelog_str[] = {
1333 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1334 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1335 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1336 "FLRW", "RESYNC","GXATR", "NOPEN",
1339 if (type >= 0 && type < CL_LAST)
1340 return changelog_str[type];
1344 /* 12 bits of per-record data can be stored in the bottom of the flags */
1345 #define CLF_FLAGSHIFT 12
1346 enum changelog_rec_flags {
1347 CLF_VERSION = 0x1000,
1348 CLF_RENAME = 0x2000,
1350 CLF_EXTRA_FLAGS = 0x8000,
1351 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1353 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1354 CLF_VERMASK = ~CLF_FLAGMASK,
1358 /* Anything under the flagmask may be per-type (if desired) */
1359 /* Flags for unlink */
1360 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1361 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1362 /* HSM cleaning needed */
1363 /* Flags for rename */
1364 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1366 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1367 * has an archive in backend */
1370 /* 12b used (from high weight to low weight):
1375 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1376 #define CLF_HSM_ERR_H 6
1377 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1378 #define CLF_HSM_EVENT_H 9
1379 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1380 #define CLF_HSM_FLAG_H 11
1381 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1382 #define CLF_HSM_SPARE_H 15
1383 #define CLF_HSM_LAST 15
1385 /* Remove bits higher than _h, then extract the value
1386 * between _h and _l by shifting lower weigth to bit 0. */
1387 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1388 >> (CLF_HSM_LAST - _h + _l))
1390 #define CLF_HSM_SUCCESS 0x00
1391 #define CLF_HSM_MAXERROR 0x7E
1392 #define CLF_HSM_ERROVERFLOW 0x7F
1394 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1396 /* 3 bits field => 8 values allowed */
1408 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1410 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1414 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1417 *clf_flags |= (he << CLF_HSM_EVENT_L);
1420 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1422 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1425 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1428 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1431 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1433 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1436 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1439 *clf_flags |= (error << CLF_HSM_ERR_L);
1442 enum changelog_rec_extra_flags {
1444 CLFE_UIDGID = 0x0001,
1447 CLFE_XATTR = 0x0008,
1448 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1451 enum changelog_send_flag {
1452 /* Not yet implemented */
1453 CHANGELOG_FLAG_FOLLOW = 0x01,
1454 /* Blocking IO makes sense in case of slow user parsing of the records,
1455 * but it also prevents us from cleaning up if the records are not
1457 CHANGELOG_FLAG_BLOCK = 0x02,
1458 /* Pack jobid into the changelog records if available. */
1459 CHANGELOG_FLAG_JOBID = 0x04,
1460 /* Pack additional flag bits into the changelog record */
1461 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1464 enum changelog_send_extra_flag {
1465 /* Pack uid/gid into the changelog record */
1466 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1467 /* Pack nid into the changelog record */
1468 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1469 /* Pack open mode into the changelog record */
1470 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1471 /* Pack xattr name into the changelog record */
1472 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1475 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1476 changelog_rec_offset(CLF_SUPPORTED, \
1479 /* 31 usable bytes string + null terminator. */
1480 #define LUSTRE_JOBID_SIZE 32
1482 /* This is the minimal changelog record. It can contain extensions
1483 * such as rename fields or process jobid. Its exact content is described
1484 * by the cr_flags and cr_extra_flags.
1486 * Extensions are packed in the same order as their corresponding flags,
1487 * then in the same order as their corresponding extra flags.
1489 struct changelog_rec {
1491 __u16 cr_flags; /**< \a changelog_rec_flags */
1492 __u32 cr_type; /**< \a changelog_rec_type */
1493 __u64 cr_index; /**< changelog record number */
1494 __u64 cr_prev; /**< last index for this target fid */
1497 struct lu_fid cr_tfid; /**< target fid */
1498 __u32 cr_markerflags; /**< CL_MARK flags */
1500 struct lu_fid cr_pfid; /**< parent fid */
1503 /* Changelog extension for RENAME. */
1504 struct changelog_ext_rename {
1505 struct lu_fid cr_sfid; /**< source fid, or zero */
1506 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1509 /* Changelog extension to include JOBID. */
1510 struct changelog_ext_jobid {
1511 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1514 /* Changelog extension to include additional flags. */
1515 struct changelog_ext_extra_flags {
1516 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1519 /* Changelog extra extension to include UID/GID. */
1520 struct changelog_ext_uidgid {
1525 /* Changelog extra extension to include NID. */
1526 struct changelog_ext_nid {
1527 /* have __u64 instead of lnet_nid_t type for use by client api */
1529 /* for use when IPv6 support is added */
1534 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1535 struct changelog_ext_openmode {
1539 /* Changelog extra extension to include xattr */
1540 struct changelog_ext_xattr {
1541 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1544 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1545 const struct changelog_rec *rec);
1547 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1548 enum changelog_rec_extra_flags cref)
1550 size_t size = sizeof(struct changelog_rec);
1552 if (crf & CLF_RENAME)
1553 size += sizeof(struct changelog_ext_rename);
1555 if (crf & CLF_JOBID)
1556 size += sizeof(struct changelog_ext_jobid);
1558 if (crf & CLF_EXTRA_FLAGS) {
1559 size += sizeof(struct changelog_ext_extra_flags);
1560 if (cref & CLFE_UIDGID)
1561 size += sizeof(struct changelog_ext_uidgid);
1562 if (cref & CLFE_NID)
1563 size += sizeof(struct changelog_ext_nid);
1564 if (cref & CLFE_OPEN)
1565 size += sizeof(struct changelog_ext_openmode);
1566 if (cref & CLFE_XATTR)
1567 size += sizeof(struct changelog_ext_xattr);
1573 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1575 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1577 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1578 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1580 return changelog_rec_offset(rec->cr_flags, cref);
1583 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1585 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1589 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1591 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1593 return (struct changelog_ext_rename *)((char *)rec +
1594 changelog_rec_offset(crf,
1598 /* The jobid follows the rename extension, if present */
1600 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1602 enum changelog_rec_flags crf = rec->cr_flags &
1603 (CLF_VERSION | CLF_RENAME);
1605 return (struct changelog_ext_jobid *)((char *)rec +
1606 changelog_rec_offset(crf,
1610 /* The additional flags follow the rename and jobid extensions, if present */
1612 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1613 const struct changelog_rec *rec)
1615 enum changelog_rec_flags crf = rec->cr_flags &
1616 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1618 return (struct changelog_ext_extra_flags *)((char *)rec +
1619 changelog_rec_offset(crf,
1623 /* The uid/gid is the first extra extension */
1625 struct changelog_ext_uidgid *changelog_rec_uidgid(
1626 const struct changelog_rec *rec)
1628 enum changelog_rec_flags crf = rec->cr_flags &
1629 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1631 return (struct changelog_ext_uidgid *)((char *)rec +
1632 changelog_rec_offset(crf,
1636 /* The nid is the second extra extension */
1638 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1640 enum changelog_rec_flags crf = rec->cr_flags &
1641 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1642 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1644 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1645 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1648 return (struct changelog_ext_nid *)((char *)rec +
1649 changelog_rec_offset(crf, cref));
1652 /* The OPEN mode is the third extra extension */
1654 struct changelog_ext_openmode *changelog_rec_openmode(
1655 const struct changelog_rec *rec)
1657 enum changelog_rec_flags crf = rec->cr_flags &
1658 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1659 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1661 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1662 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1663 (CLFE_UIDGID | CLFE_NID);
1665 return (struct changelog_ext_openmode *)((char *)rec +
1666 changelog_rec_offset(crf, cref));
1669 /* The xattr name is the fourth extra extension */
1671 struct changelog_ext_xattr *changelog_rec_xattr(
1672 const struct changelog_rec *rec)
1674 enum changelog_rec_flags crf = rec->cr_flags &
1675 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1676 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1678 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1679 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1680 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1682 return (struct changelog_ext_xattr *)((char *)rec +
1683 changelog_rec_offset(crf, cref));
1686 /* The name follows the rename, jobid and extra flags extns, if present */
1687 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1689 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1691 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1692 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1694 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1695 cref & CLFE_SUPPORTED);
1698 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1700 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1703 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1705 char *cr_name = changelog_rec_name(rec);
1707 return cr_name + strlen(cr_name) + 1;
1711 * Remap a record to the desired format as specified by the crf flags.
1712 * The record must be big enough to contain the final remapped version.
1713 * Superfluous extension fields are removed and missing ones are added
1714 * and zeroed. The flags of the record are updated accordingly.
1716 * The jobid and rename extensions can be added to a record, to match the
1717 * format an application expects, typically. In this case, the newly added
1718 * fields will be zeroed.
1719 * The Jobid field can be removed, to guarantee compatibility with older
1720 * clients that don't expect this field in the records they process.
1722 * The following assumptions are being made:
1723 * - CLF_RENAME will not be removed
1724 * - CLF_JOBID will not be added without CLF_RENAME being added too
1725 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1727 * @param[in,out] rec The record to remap.
1728 * @param[in] crf_wanted Flags describing the desired extensions.
1729 * @param[in] cref_want Flags describing the desired extra extensions.
1731 static inline void changelog_remap_rec(struct changelog_rec *rec,
1732 enum changelog_rec_flags crf_wanted,
1733 enum changelog_rec_extra_flags cref_want)
1735 char *xattr_mov = NULL;
1736 char *omd_mov = NULL;
1737 char *nid_mov = NULL;
1738 char *uidgid_mov = NULL;
1742 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1744 crf_wanted &= CLF_SUPPORTED;
1745 cref_want &= CLFE_SUPPORTED;
1747 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1748 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1749 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1750 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1756 /* First move the variable-length name field */
1757 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1758 changelog_rec_name(rec), rec->cr_namelen);
1760 /* Locations of extensions in the remapped record */
1761 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1762 xattr_mov = (char *)rec +
1763 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1764 cref_want & ~CLFE_XATTR);
1765 omd_mov = (char *)rec +
1766 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1767 cref_want & ~(CLFE_OPEN |
1769 nid_mov = (char *)rec +
1770 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1771 cref_want & ~(CLFE_NID |
1774 uidgid_mov = (char *)rec +
1775 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1776 cref_want & ~(CLFE_UIDGID |
1780 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1783 ef_mov = (char *)rec +
1784 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1786 jid_mov = (char *)rec +
1787 changelog_rec_offset(crf_wanted &
1788 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1790 rnm_mov = (char *)rec +
1791 changelog_rec_offset(crf_wanted &
1797 /* Move the extension fields to the desired positions */
1798 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1799 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1800 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1801 memmove(xattr_mov, changelog_rec_xattr(rec),
1802 sizeof(struct changelog_ext_xattr));
1804 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1805 memmove(omd_mov, changelog_rec_openmode(rec),
1806 sizeof(struct changelog_ext_openmode));
1808 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1809 memmove(nid_mov, changelog_rec_nid(rec),
1810 sizeof(struct changelog_ext_nid));
1812 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1813 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1814 sizeof(struct changelog_ext_uidgid));
1816 memmove(ef_mov, changelog_rec_extra_flags(rec),
1817 sizeof(struct changelog_ext_extra_flags));
1820 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1821 memmove(jid_mov, changelog_rec_jobid(rec),
1822 sizeof(struct changelog_ext_jobid));
1824 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1825 memmove(rnm_mov, changelog_rec_rename(rec),
1826 sizeof(struct changelog_ext_rename));
1828 /* Clear newly added fields */
1829 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1830 !(cref & CLFE_XATTR))
1831 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1833 if (omd_mov && (cref_want & CLFE_OPEN) &&
1834 !(cref & CLFE_OPEN))
1835 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1837 if (nid_mov && (cref_want & CLFE_NID) &&
1839 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1841 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1842 !(cref & CLFE_UIDGID))
1843 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1845 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1846 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1847 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1849 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1850 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1852 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1853 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1855 /* Update the record's flags accordingly */
1856 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1857 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1858 changelog_rec_extra_flags(rec)->cr_extra_flags =
1859 changelog_rec_extra_flags(rec)->cr_extra_flags |
1863 enum changelog_message_type {
1864 CL_RECORD = 10, /* message is a changelog_rec */
1865 CL_EOF = 11, /* at end of current changelog */
1868 /********* Misc **********/
1870 struct ioc_data_version {
1872 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1873 __u32 idv_flags; /* enum ioc_data_version_flags */
1876 enum ioc_data_version_flags {
1877 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1878 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1882 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1885 #define dot_lustre_name ".lustre"
1888 /********* HSM **********/
1890 /** HSM per-file state
1891 * See HSM_FLAGS below.
1894 HS_NONE = 0x00000000,
1895 HS_EXISTS = 0x00000001,
1896 HS_DIRTY = 0x00000002,
1897 HS_RELEASED = 0x00000004,
1898 HS_ARCHIVED = 0x00000008,
1899 HS_NORELEASE = 0x00000010,
1900 HS_NOARCHIVE = 0x00000020,
1901 HS_LOST = 0x00000040,
1904 /* HSM user-setable flags. */
1905 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1907 /* Other HSM flags. */
1908 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1911 * All HSM-related possible flags that could be applied to a file.
1912 * This should be kept in sync with hsm_states.
1914 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1917 * HSM request progress state
1919 enum hsm_progress_states {
1926 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1929 case HPS_WAITING: return "waiting";
1930 case HPS_RUNNING: return "running";
1931 case HPS_DONE: return "done";
1932 default: return "unknown";
1939 } __attribute__((packed));
1942 * Current HSM states of a Lustre file.
1944 * This structure purpose is to be sent to user-space mainly. It describes the
1945 * current HSM flags and in-progress action.
1947 struct hsm_user_state {
1948 /** Current HSM states, from enum hsm_states. */
1950 __u32 hus_archive_id;
1951 /** The current undergoing action, if there is one */
1952 __u32 hus_in_progress_state;
1953 __u32 hus_in_progress_action;
1954 struct hsm_extent hus_in_progress_location;
1955 char hus_extended_info[];
1958 struct hsm_state_set_ioc {
1959 struct lu_fid hssi_fid;
1961 __u64 hssi_clearmask;
1965 * This structure describes the current in-progress action for a file.
1966 * it is retuned to user space and send over the wire
1968 struct hsm_current_action {
1969 /** The current undergoing action, if there is one */
1970 /* state is one of hsm_progress_states */
1972 /* action is one of hsm_user_action */
1974 struct hsm_extent hca_location;
1977 /***** HSM user requests ******/
1978 /* User-generated (lfs/ioctl) request types */
1979 enum hsm_user_action {
1980 HUA_NONE = 1, /* no action (noop) */
1981 HUA_ARCHIVE = 10, /* copy to hsm */
1982 HUA_RESTORE = 11, /* prestage */
1983 HUA_RELEASE = 12, /* drop ost objects */
1984 HUA_REMOVE = 13, /* remove from archive */
1985 HUA_CANCEL = 14 /* cancel a request */
1988 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1991 case HUA_NONE: return "NOOP";
1992 case HUA_ARCHIVE: return "ARCHIVE";
1993 case HUA_RESTORE: return "RESTORE";
1994 case HUA_RELEASE: return "RELEASE";
1995 case HUA_REMOVE: return "REMOVE";
1996 case HUA_CANCEL: return "CANCEL";
1997 default: return "UNKNOWN";
2002 * List of hr_flags (bit field)
2004 #define HSM_FORCE_ACTION 0x0001
2005 /* used by CT, cannot be set by user */
2006 #define HSM_GHOST_COPY 0x0002
2009 * Contains all the fixed part of struct hsm_user_request.
2012 struct hsm_request {
2013 __u32 hr_action; /* enum hsm_user_action */
2014 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2015 __u64 hr_flags; /* request flags */
2016 __u32 hr_itemcount; /* item count in hur_user_item vector */
2020 struct hsm_user_item {
2021 struct lu_fid hui_fid;
2022 struct hsm_extent hui_extent;
2023 } __attribute__((packed));
2025 struct hsm_user_request {
2026 struct hsm_request hur_request;
2027 struct hsm_user_item hur_user_item[0];
2028 /* extra data blob at end of struct (after all
2029 * hur_user_items), only use helpers to access it
2031 } __attribute__((packed));
2033 /** Return pointer to data field in a hsm user request */
2034 static inline void *hur_data(struct hsm_user_request *hur)
2036 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2040 * Compute the current length of the provided hsm_user_request. This returns -1
2041 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
2043 * return -1 on bounds check error.
2045 static inline ssize_t hur_len(struct hsm_user_request *hur)
2049 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2050 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2051 (__u64)hur->hur_request.hr_itemcount *
2052 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2054 if (size != (ssize_t)size)
2060 /****** HSM RPCs to copytool *****/
2061 /* Message types the copytool may receive */
2062 enum hsm_message_type {
2063 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2066 /* Actions the copytool may be instructed to take for a given action_item */
2067 enum hsm_copytool_action {
2068 HSMA_NONE = 10, /* no action */
2069 HSMA_ARCHIVE = 20, /* arbitrary offset */
2075 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2078 case HSMA_NONE: return "NOOP";
2079 case HSMA_ARCHIVE: return "ARCHIVE";
2080 case HSMA_RESTORE: return "RESTORE";
2081 case HSMA_REMOVE: return "REMOVE";
2082 case HSMA_CANCEL: return "CANCEL";
2083 default: return "UNKNOWN";
2087 /* Copytool item action description */
2088 struct hsm_action_item {
2089 __u32 hai_len; /* valid size of this struct */
2090 __u32 hai_action; /* hsm_copytool_action, but use known size */
2091 struct lu_fid hai_fid; /* Lustre FID to operate on */
2092 struct lu_fid hai_dfid; /* fid used for data access */
2093 struct hsm_extent hai_extent; /* byte range to operate on */
2094 __u64 hai_cookie; /* action cookie from coordinator */
2095 __u64 hai_gid; /* grouplock id */
2096 char hai_data[0]; /* variable length */
2097 } __attribute__((packed));
2100 * helper function which print in hexa the first bytes of
2103 * \param hai [IN] record to print
2104 * \param buffer [IN,OUT] buffer to write the hex string to
2105 * \param len [IN] max buffer length
2109 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2110 char *buffer, size_t len)
2117 data_len = hai->hai_len - sizeof(*hai);
2118 for (i = 0; (i < data_len) && (len > 2); i++) {
2119 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2129 /* Copytool action list */
2130 #define HAL_VERSION 1
2131 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2132 struct hsm_action_list {
2134 __u32 hal_count; /* number of hai's to follow */
2135 __u64 hal_compound_id; /* returned by coordinator, ignored */
2137 __u32 hal_archive_id; /* which archive backend */
2139 char hal_fsname[0]; /* null-terminated */
2140 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2141 boundaries. See hai_zero */
2142 } __attribute__((packed));
2144 /* Return pointer to first hai in action list */
2145 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2147 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2149 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2152 /* Return pointer to next hai */
2153 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2155 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2157 return (struct hsm_action_item *)((char *)hai + offset);
2160 /* Return size of an hsm_action_list */
2161 static inline size_t hal_size(struct hsm_action_list *hal)
2165 struct hsm_action_item *hai;
2167 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2168 hai = hai_first(hal);
2169 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2170 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2176 * describe the attributes to be set on imported file
2178 struct hsm_user_import {
2187 __u32 hui_archive_id;
2190 /* Copytool progress reporting */
2191 #define HP_FLAG_COMPLETED 0x01
2192 #define HP_FLAG_RETRY 0x02
2194 struct hsm_progress {
2195 struct lu_fid hp_fid;
2197 struct hsm_extent hp_extent;
2199 __u16 hp_errval; /* positive val */
2204 __u64 hc_data_version;
2206 __u16 hc_errval; /* positive val */
2208 struct hsm_action_item hc_hai;
2212 enum llapi_json_types {
2213 LLAPI_JSON_INTEGER = 1,
2219 struct llapi_json_item {
2228 struct llapi_json_item *lji_next;
2231 struct llapi_json_item_list {
2232 int ljil_item_count;
2233 struct llapi_json_item *ljil_items;
2236 enum lu_ladvise_type {
2237 LU_LADVISE_INVALID = 0,
2238 LU_LADVISE_WILLREAD = 1,
2239 LU_LADVISE_DONTNEED = 2,
2240 LU_LADVISE_LOCKNOEXPAND = 3,
2241 LU_LADVISE_LOCKAHEAD = 4,
2245 #define LU_LADVISE_NAMES { \
2246 [LU_LADVISE_WILLREAD] = "willread", \
2247 [LU_LADVISE_DONTNEED] = "dontneed", \
2248 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2249 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2252 /* This is the userspace argument for ladvise. It is currently the same as
2253 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2254 * may need info which is only used locally. */
2255 struct llapi_lu_ladvise {
2256 __u16 lla_advice; /* advice type */
2257 __u16 lla_value1; /* values for different advice types */
2259 __u64 lla_start; /* first byte of extent for advice */
2260 __u64 lla_end; /* last byte of extent for advice */
2266 LF_ASYNC = 0x00000001,
2267 LF_UNSET = 0x00000002,
2270 #define LADVISE_MAGIC 0x1ADF1CE0
2271 /* Masks of valid flags for each advice */
2272 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2273 /* Flags valid for all advices not explicitly specified */
2274 #define LF_DEFAULT_MASK LF_ASYNC
2276 #define LF_MASK (LF_ASYNC | LF_UNSET)
2278 #define lla_lockahead_mode lla_value1
2279 #define lla_peradvice_flags lla_value2
2280 #define lla_lockahead_result lla_value3
2282 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2283 * is used on the wire. It is defined separately as we may need info which is
2284 * only used locally. */
2285 struct llapi_ladvise_hdr {
2286 __u32 lah_magic; /* LADVISE_MAGIC */
2287 __u32 lah_count; /* number of advices */
2288 __u64 lah_flags; /* from enum ladvise_flag */
2289 __u32 lah_value1; /* unused */
2290 __u32 lah_value2; /* unused */
2291 __u64 lah_value3; /* unused */
2292 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2295 #define LAH_COUNT_MAX (1024)
2299 SK_CRYPT_INVALID = -1,
2301 SK_CRYPT_AES256_CTR = 1,
2305 SK_HMAC_INVALID = -1,
2311 struct sk_crypt_type {
2312 const char *sct_name;
2316 struct sk_hmac_type {
2317 const char *sht_name;
2321 enum lock_mode_user {
2327 #define LOCK_MODE_NAMES { \
2328 [MODE_READ_USER] = "READ",\
2329 [MODE_WRITE_USER] = "WRITE"\
2332 enum lockahead_results {
2333 LLA_RESULT_SENT = 0,
2334 LLA_RESULT_DIFFERENT,
2340 /* make header's size equal lu_fid */
2343 struct lu_fid fa_fids[0];
2345 #define OBD_MAX_FIDS_IN_ARRAY 4096
2347 #if defined(__cplusplus)
2351 /** @} lustreuser */
2353 #endif /* _LUSTRE_USER_H */