4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_user.h
34 * Lustre public user-space interface definitions.
37 #ifndef _LUSTRE_USER_H
38 #define _LUSTRE_USER_H
40 /** \defgroup lustreuser lustreuser
46 #include <linux/limits.h>
47 #include <linux/kernel.h>
48 #include <linux/string.h>
49 #include <linux/quota.h>
50 #include <linux/types.h>
51 #include <linux/unistd.h>
52 #include <linux/lustre/lustre_fiemap.h>
55 # define __USE_ISOC99 1
57 # include <stdio.h> /* snprintf() */
58 # include <sys/stat.h>
59 # define FILEID_LUSTRE 0x97 /* for name_to_handle_at() (and llapi_fd2fid()) */
60 #endif /* !__KERNEL__ */
62 #if defined(__cplusplus)
66 #ifdef __STRICT_ANSI__
67 #define typeof __typeof__
71 * This is a temporary solution of adding quota type.
72 * Should be removed as soon as system header is updated.
75 #define LL_MAXQUOTAS 3
77 #define INITQFNAMES { \
78 "user", /* USRQUOTA */ \
79 "group", /* GRPQUOTA */ \
80 "project", /* PRJQUOTA */ \
94 * We need to always use 64bit version because the structure
95 * is shared across entire cluster where 32bit and 64bit machines
98 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
99 typedef struct stat64 lstat_t;
100 #define lstat_f lstat64
101 #define fstat_f fstat64
102 #define fstatat_f fstatat64
104 typedef struct stat lstat_t;
105 #define lstat_f lstat
106 #define fstat_f fstat
107 #define fstatat_f fstatat
110 #ifndef STATX_BASIC_STATS
112 * Timestamp structure for the timestamps in struct statx.
114 * tv_sec holds the number of seconds before (negative) or after (positive)
115 * 00:00:00 1st January 1970 UTC.
117 * tv_nsec holds a number of nanoseconds (0..999,999,999) after the tv_sec time.
119 * __reserved is held in case we need a yet finer resolution.
121 struct statx_timestamp {
128 * Structures for the extended file attribute retrieval system call
131 * The caller passes a mask of what they're specifically interested in as a
132 * parameter to statx(). What statx() actually got will be indicated in
133 * st_mask upon return.
135 * For each bit in the mask argument:
137 * - if the datum is not supported:
139 * - the bit will be cleared, and
141 * - the datum will be set to an appropriate fabricated value if one is
142 * available (eg. CIFS can take a default uid and gid), otherwise
144 * - the field will be cleared;
146 * - otherwise, if explicitly requested:
148 * - the datum will be synchronised to the server if AT_STATX_FORCE_SYNC is
149 * set or if the datum is considered out of date, and
151 * - the field will be filled in and the bit will be set;
153 * - otherwise, if not requested, but available in approximate form without any
154 * effort, it will be filled in anyway, and the bit will be set upon return
155 * (it might not be up to date, however, and no attempt will be made to
156 * synchronise the internal state first);
158 * - otherwise the field and the bit will be cleared before returning.
160 * Items in STATX_BASIC_STATS may be marked unavailable on return, but they
161 * will have values installed for compatibility purposes so that stat() and
162 * co. can be emulated in userspace.
166 __u32 stx_mask; /* What results were written [uncond] */
167 __u32 stx_blksize; /* Preferred general I/O size [uncond] */
168 __u64 stx_attributes; /* Flags conveying information about the file [uncond] */
170 __u32 stx_nlink; /* Number of hard links */
171 __u32 stx_uid; /* User ID of owner */
172 __u32 stx_gid; /* Group ID of owner */
173 __u16 stx_mode; /* File mode */
176 __u64 stx_ino; /* Inode number */
177 __u64 stx_size; /* File size */
178 __u64 stx_blocks; /* Number of 512-byte blocks allocated */
179 __u64 stx_attributes_mask; /* Mask to show what's supported in stx_attributes */
181 struct statx_timestamp stx_atime; /* Last access time */
182 struct statx_timestamp stx_btime; /* File creation time */
183 struct statx_timestamp stx_ctime; /* Last attribute change time */
184 struct statx_timestamp stx_mtime; /* Last data modification time */
186 __u32 stx_rdev_major; /* Device ID of special file [if bdev/cdev] */
187 __u32 stx_rdev_minor;
188 __u32 stx_dev_major; /* ID of device containing file [uncond] */
191 __u64 __spare2[14]; /* Spare space for future expansion */
196 * Flags to be stx_mask
198 * Query request/result mask for statx() and struct statx::stx_mask.
200 * These bits should be set in the mask argument of statx() to request
201 * particular items when calling statx().
203 #define STATX_TYPE 0x00000001U /* Want/got stx_mode & S_IFMT */
204 #define STATX_MODE 0x00000002U /* Want/got stx_mode & ~S_IFMT */
205 #define STATX_NLINK 0x00000004U /* Want/got stx_nlink */
206 #define STATX_UID 0x00000008U /* Want/got stx_uid */
207 #define STATX_GID 0x00000010U /* Want/got stx_gid */
208 #define STATX_ATIME 0x00000020U /* Want/got stx_atime */
209 #define STATX_MTIME 0x00000040U /* Want/got stx_mtime */
210 #define STATX_CTIME 0x00000080U /* Want/got stx_ctime */
211 #define STATX_INO 0x00000100U /* Want/got stx_ino */
212 #define STATX_SIZE 0x00000200U /* Want/got stx_size */
213 #define STATX_BLOCKS 0x00000400U /* Want/got stx_blocks */
214 #define STATX_BASIC_STATS 0x000007ffU /* The stuff in the normal stat struct */
215 #define STATX_BTIME 0x00000800U /* Want/got stx_btime */
216 #define STATX_ALL 0x00000fffU /* All currently supported flags */
217 #define STATX__RESERVED 0x80000000U /* Reserved for future struct statx expansion */
220 * Attributes to be found in stx_attributes and masked in stx_attributes_mask.
222 * These give information about the features or the state of a file that might
223 * be of use to ordinary userspace programs such as GUIs or ls rather than
226 * Note that the flags marked [I] correspond to generic FS_IOC_FLAGS
227 * semantically. Where possible, the numerical value is picked to correspond
230 #define STATX_ATTR_COMPRESSED 0x00000004 /* [I] File is compressed by the fs */
231 #define STATX_ATTR_IMMUTABLE 0x00000010 /* [I] File is marked immutable */
232 #define STATX_ATTR_APPEND 0x00000020 /* [I] File is append-only */
233 #define STATX_ATTR_NODUMP 0x00000040 /* [I] File is not to be dumped */
234 #define STATX_ATTR_ENCRYPTED 0x00000800 /* [I] File requires key to decrypt in fs */
236 #define STATX_ATTR_AUTOMOUNT 0x00001000 /* Dir: Automount trigger */
238 #define AT_STATX_SYNC_TYPE 0x6000 /* Type of synchronisation required from statx() */
239 #define AT_STATX_SYNC_AS_STAT 0x0000 /* - Do whatever stat() does */
240 #define AT_STATX_FORCE_SYNC 0x2000 /* - Force the attributes to be sync'd with the server */
241 #define AT_STATX_DONT_SYNC 0x4000 /* - Don't sync attributes with the server */
243 #endif /* STATX_BASIC_STATS */
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() */
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 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
576 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
577 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
578 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
579 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
580 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
581 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
582 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
583 /* LL_IOC_RECREATE_OBJ 157 obsolete */
584 /* LL_IOC_RECREATE_FID 157 obsolete */
585 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
586 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
587 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
588 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
589 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
590 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
591 /* IOC_LOV_GETINFO 165 obsolete */
592 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
593 /* LL_IOC_RMTACL 167 obsolete */
594 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
595 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
596 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
597 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
598 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
599 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
600 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
601 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
602 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
603 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
604 /* lustre_ioctl.h 177-210 */
605 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
606 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
607 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
608 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
609 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
610 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
611 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
612 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
613 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
614 struct lustre_swap_layouts)
615 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
616 struct hsm_current_action)
617 /* lustre_ioctl.h 221-232 */
618 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
619 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
620 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
621 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
622 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
623 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
624 #define LL_IOC_GET_LEASE _IO('f', 244)
625 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
626 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
627 #define LL_IOC_MIGRATE _IOR('f', 247, int)
628 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
629 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
630 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
631 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
632 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
633 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
634 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
635 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
637 #ifndef FS_IOC_FSGETXATTR
639 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
642 __u32 fsx_xflags; /* xflags field value (get/set) */
643 __u32 fsx_extsize; /* extsize field value (get/set)*/
644 __u32 fsx_nextents; /* nextents field value (get) */
645 __u32 fsx_projid; /* project identifier (get/set) */
646 unsigned char fsx_pad[12];
648 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
649 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
651 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
652 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
653 #ifndef FS_XFLAG_PROJINHERIT
654 #define FS_XFLAG_PROJINHERIT 0x00000200
658 #define LL_STATFS_LMV 1
659 #define LL_STATFS_LOV 2
660 #define LL_STATFS_NODELAY 4
662 #define IOC_MDC_TYPE 'i'
663 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
664 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
665 #ifdef HAVE_LOV_USER_MDS_DATA
666 #define IOC_MDC_GETFILEINFO_OLD _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
667 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
668 #define LL_IOC_MDC_GETINFO_OLD _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
669 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
672 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
674 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
675 * files, but are unlikely to be used in practice and are not harmful if
676 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
677 * devices and are safe for use on new files. See LU-4209. */
678 /* To be compatible with old statically linked binary we keep the check for
679 * the older 0100000000 flag. This is already removed upstream. LU-812. */
680 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
682 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
684 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
685 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
686 O_LOV_DELAY_CREATE_MASK)
688 #define LL_FILE_IGNORE_LOCK 0x00000001
689 #define LL_FILE_GROUP_LOCKED 0x00000002
690 #define LL_FILE_READAHEA 0x00000004
691 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
692 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
693 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
695 #define LOV_USER_MAGIC_V1 0x0BD10BD0
696 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
697 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
698 #define LOV_USER_MAGIC_V3 0x0BD30BD0
699 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
700 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
701 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
702 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
703 #define LOV_USER_MAGIC_SEL 0x0BD80BD0
705 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
706 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
707 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
709 #define LOV_PATTERN_NONE 0x000
710 #define LOV_PATTERN_RAID0 0x001
711 #define LOV_PATTERN_RAID1 0x002
712 #define LOV_PATTERN_MDT 0x100
713 #define LOV_PATTERN_OVERSTRIPING 0x200
715 #define LOV_PATTERN_F_MASK 0xffff0000
716 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
717 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
718 #define LOV_PATTERN_DEFAULT 0xffffffff
720 #define LOV_OFFSET_DEFAULT ((__u16)-1)
721 #define LMV_OFFSET_DEFAULT ((__u32)-1)
723 static inline bool lov_pattern_supported(__u32 pattern)
725 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
726 (pattern & ~LOV_PATTERN_F_RELEASED) ==
727 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
728 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
731 /* RELEASED and MDT patterns are not valid in many places, so rather than
732 * having many extra checks on lov_pattern_supported, we have this separate
733 * check for non-released, non-DOM components
735 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
737 return pattern == LOV_PATTERN_RAID0 ||
738 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
742 #define LOV_MAXPOOLNAME 15
743 #define LOV_POOLNAMEF "%.15s"
745 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
746 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
747 #define LOV_MAX_STRIPE_COUNT_OLD 160
748 /* This calculation is crafted so that input of 4096 will result in 160
749 * which in turn is equal to old maximal stripe count.
750 * XXX: In fact this is too simpified for now, what it also need is to get
751 * ea_type argument to clearly know how much space each stripe consumes.
753 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
754 * allocation that is sufficient for the current generation of systems.
756 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
757 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
758 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
759 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
761 #define XATTR_LUSTRE_PREFIX "lustre."
762 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
764 /* Please update if XATTR_LUSTRE_LOV".set" groks more flags in the future */
765 #define allowed_lustre_lov(att) (strcmp((att), XATTR_LUSTRE_LOV".add") == 0 || \
766 strcmp((att), XATTR_LUSTRE_LOV".set") == 0 || \
767 strcmp((att), XATTR_LUSTRE_LOV".set.flags") == 0 || \
768 strcmp((att), XATTR_LUSTRE_LOV".del") == 0)
770 #define lov_user_ost_data lov_user_ost_data_v1
771 struct lov_user_ost_data_v1 { /* per-stripe data structure */
772 struct ost_id l_ost_oi; /* OST object ID */
773 __u32 l_ost_gen; /* generation of this OST index */
774 __u32 l_ost_idx; /* OST index in LOV */
775 } __attribute__((packed));
777 #define lov_user_md lov_user_md_v1
778 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
779 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
780 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
781 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
782 __u32 lmm_stripe_size; /* size of stripe in bytes */
783 __u16 lmm_stripe_count; /* num stripes in use for this object */
785 __u16 lmm_stripe_offset; /* starting stripe offset in
786 * lmm_objects, use when writing */
787 __u16 lmm_layout_gen; /* layout generation number
788 * used when reading */
790 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
791 } __attribute__((packed, __may_alias__));
793 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
794 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
795 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
796 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
797 __u32 lmm_stripe_size; /* size of stripe in bytes */
798 __u16 lmm_stripe_count; /* num stripes in use for this object */
800 __u16 lmm_stripe_offset; /* starting stripe offset in
801 * lmm_objects, use when writing */
802 __u16 lmm_layout_gen; /* layout generation number
803 * used when reading */
805 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
806 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
807 } __attribute__((packed, __may_alias__));
809 struct lov_foreign_md {
810 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
811 __u32 lfm_length; /* length of lfm_value */
812 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
813 __u32 lfm_flags; /* flags, type specific */
815 } __attribute__((packed));
817 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
818 offsetof(struct lov_foreign_md, lfm_value))
820 #define foreign_size_le(lfm) \
821 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
822 offsetof(struct lov_foreign_md, lfm_value))
825 * The stripe size fields are shared for the extension size storage, however
826 * the extension size is stored in KB, not bytes.
828 #define SEL_UNIT_SIZE 1024llu
833 } __attribute__((packed));
835 #define DEXT "[%#llx, %#llx)"
836 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
838 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
839 struct lu_extent *e2)
841 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
844 static inline bool lu_extent_is_whole(struct lu_extent *e)
846 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
849 enum lov_comp_md_entry_flags {
850 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
851 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
852 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
853 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
854 LCME_FL_OFFLINE = 0x00000008, /* Not used */
855 LCME_FL_INIT = 0x00000010, /* instantiated */
856 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
857 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
858 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
859 * won't be stored on disk
863 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
864 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
867 /* The component flags can be set by users at creation/modification time. */
868 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
871 /* The mirror flags can be set by users at creation time. */
872 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
874 /* The allowed flags obtained from the client at component creation time. */
875 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
877 /* The mirror flags sent by client */
878 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
880 /* These flags have meaning when set in a default layout and will be inherited
881 * from the default/template layout set on a directory.
883 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
886 /* the highest bit in obdo::o_layout_version is used to mark if the file is
888 #define LU_LAYOUT_RESYNC LCME_FL_NEG
890 /* lcme_id can be specified as certain flags, and the the first
891 * bit of lcme_id is used to indicate that the ID is representing
892 * certain LCME_FL_* but not a real ID. Which implies we can have
893 * at most 31 flags (see LCME_FL_XXX). */
896 LCME_ID_MAX = 0x7FFFFFFF,
897 LCME_ID_ALL = 0xFFFFFFFF,
898 LCME_ID_NOT_ID = LCME_FL_NEG
901 #define LCME_ID_MASK LCME_ID_MAX
903 struct lov_comp_md_entry_v1 {
904 __u32 lcme_id; /* unique id of component */
905 __u32 lcme_flags; /* LCME_FL_XXX */
906 struct lu_extent lcme_extent; /* file extent for component */
907 __u32 lcme_offset; /* offset of component blob,
908 start from lov_comp_md_v1 */
909 __u32 lcme_size; /* size of component blob */
910 __u32 lcme_layout_gen;
911 __u64 lcme_timestamp; /* snapshot time if applicable*/
912 __u32 lcme_padding_1;
913 } __attribute__((packed));
915 #define SEQ_ID_MAX 0x0000FFFF
916 #define SEQ_ID_MASK SEQ_ID_MAX
917 /* bit 30:16 of lcme_id is used to store mirror id */
918 #define MIRROR_ID_MASK 0x7FFF0000
919 #define MIRROR_ID_NEG 0x8000
920 #define MIRROR_ID_SHIFT 16
922 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
924 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
927 static inline __u16 mirror_id_of(__u32 id)
929 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
933 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
935 enum lov_comp_md_flags {
936 /* the least 2 bits are used by FLR to record file state */
939 LCM_FL_WRITE_PENDING = 2,
940 LCM_FL_SYNC_PENDING = 3,
941 LCM_FL_FLR_MASK = 0x3,
944 struct lov_comp_md_v1 {
945 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
946 __u32 lcm_size; /* overall size including this struct */
947 __u32 lcm_layout_gen;
949 __u16 lcm_entry_count;
950 /* lcm_mirror_count stores the number of actual mirrors minus 1,
951 * so that non-flr files will have value 0 meaning 1 mirror. */
952 __u16 lcm_mirror_count;
953 __u16 lcm_padding1[3];
955 struct lov_comp_md_entry_v1 lcm_entries[0];
956 } __attribute__((packed));
958 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
960 if (stripes == (__u16)-1)
963 if (lmm_magic == LOV_USER_MAGIC_V1)
964 return sizeof(struct lov_user_md_v1) +
965 stripes * sizeof(struct lov_user_ost_data_v1);
966 return sizeof(struct lov_user_md_v3) +
967 stripes * sizeof(struct lov_user_ost_data_v1);
970 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
971 * use this. It is unsafe to #define those values in this header as it
972 * is possible the application has already #included <sys/stat.h>. */
973 #ifdef HAVE_LOV_USER_MDS_DATA
974 #define lov_user_mds_data lov_user_mds_data_v2
975 struct lov_user_mds_data_v1 {
976 lstat_t lmd_st; /* MDS stat struct */
977 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
978 } __attribute__((packed));
980 struct lov_user_mds_data_v2 {
981 struct lu_fid lmd_fid; /* Lustre FID */
982 lstatx_t lmd_stx; /* MDS statx struct */
983 __u64 lmd_flags; /* MDS stat flags */
984 __u32 lmd_lmmsize; /* LOV EA size */
985 __u32 lmd_padding; /* unused */
986 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
987 } __attribute__((packed));
990 struct lmv_user_mds_data {
991 struct lu_fid lum_fid;
994 } __attribute__((packed, __may_alias__));
997 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
998 LMV_HASH_TYPE_ALL_CHARS = 1,
999 LMV_HASH_TYPE_FNV_1A_64 = 2,
1000 LMV_HASH_TYPE_CRUSH = 3,
1004 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
1006 #define LMV_HASH_NAME_ALL_CHARS "all_char"
1007 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
1008 #define LMV_HASH_NAME_CRUSH "crush"
1010 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1011 * and the higher part will be the flag to indicate the status of object,
1012 * for example the object is being migrated. And the hash function
1013 * might be interpreted differently with different flags. */
1014 #define LMV_HASH_TYPE_MASK 0x0000ffff
1016 static inline bool lmv_is_known_hash_type(__u32 type)
1018 return (type & LMV_HASH_TYPE_MASK) == LMV_HASH_TYPE_FNV_1A_64 ||
1019 (type & LMV_HASH_TYPE_MASK) == LMV_HASH_TYPE_ALL_CHARS ||
1020 (type & LMV_HASH_TYPE_MASK) == LMV_HASH_TYPE_CRUSH;
1023 #define LMV_HASH_FLAG_MERGE 0x04000000
1024 #define LMV_HASH_FLAG_SPLIT 0x08000000
1026 /* The striped directory has ever lost its master LMV EA, then LFSCK
1027 * re-generated it. This flag is used to indicate such case. It is an
1029 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
1031 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1032 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1034 #define LMV_HASH_FLAG_LAYOUT_CHANGE \
1035 (LMV_HASH_FLAG_MIGRATION | LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MERGE)
1037 /* both SPLIT and MIGRATION are set for directory split */
1038 static inline bool lmv_hash_is_splitting(__u32 hash)
1040 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1041 (LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MIGRATION);
1044 /* both MERGE and MIGRATION are set for directory merge */
1045 static inline bool lmv_hash_is_merging(__u32 hash)
1047 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1048 (LMV_HASH_FLAG_MERGE | LMV_HASH_FLAG_MIGRATION);
1051 /* only MIGRATION is set for directory migration */
1052 static inline bool lmv_hash_is_migrating(__u32 hash)
1054 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) == LMV_HASH_FLAG_MIGRATION;
1057 static inline bool lmv_hash_is_restriping(__u32 hash)
1059 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash);
1062 static inline bool lmv_hash_is_layout_changing(__u32 hash)
1064 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash) ||
1065 lmv_hash_is_migrating(hash);
1068 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
1070 struct lustre_foreign_type {
1072 const char *lft_name;
1076 * LOV/LMV foreign types
1078 enum lustre_foreign_types {
1079 LU_FOREIGN_TYPE_NONE = 0,
1080 LU_FOREIGN_TYPE_DAOS = 0xda05,
1081 /* must be the max/last one */
1082 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
1085 extern struct lustre_foreign_type lu_foreign_types[];
1087 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
1088 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
1089 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
1090 #define lmv_user_md lmv_user_md_v1
1091 struct lmv_user_md_v1 {
1092 __u32 lum_magic; /* must be the first field */
1093 __u32 lum_stripe_count; /* dirstripe count */
1094 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
1095 __u32 lum_hash_type; /* Dir stripe policy */
1096 __u32 lum_type; /* LMV type: default */
1100 char lum_pool_name[LOV_MAXPOOLNAME + 1];
1101 struct lmv_user_mds_data lum_objects[0];
1102 } __attribute__((packed));
1104 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
1106 if (size <= sizeof(struct lmv_user_md))
1109 size -= sizeof(struct lmv_user_md);
1110 return (size + sizeof(struct lmv_user_mds_data) - 1) /
1111 sizeof(struct lmv_user_mds_data);
1115 * NB, historically default layout didn't set type, but use XATTR name to differ
1116 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
1117 * and still use the same method.
1120 LMV_TYPE_DEFAULT = 0x0000,
1123 static inline int lmv_user_md_size(int stripes, int lmm_magic)
1125 int size = sizeof(struct lmv_user_md);
1127 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
1128 size += stripes * sizeof(struct lmv_user_mds_data);
1133 struct ll_recreate_obj {
1139 __u64 id; /* holds object id */
1140 __u32 generation; /* holds object generation */
1141 __u32 f_type; /* holds object type or stripe idx when passing it to
1142 * OST for saving into EA. */
1147 char uuid[UUID_MAX];
1150 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
1151 const struct obd_uuid *u2)
1153 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
1156 static inline int obd_uuid_empty(struct obd_uuid *uuid)
1158 return uuid->uuid[0] == '\0';
1161 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
1163 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
1164 uuid->uuid[sizeof(*uuid) - 1] = '\0';
1167 /* For printf's only, make sure uuid is terminated */
1168 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
1173 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
1174 /* Obviously not safe, but for printfs, no real harm done...
1175 we're always null-terminated, even in a race. */
1176 static char temp[sizeof(*uuid)];
1177 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
1178 temp[sizeof(*uuid) - 1] = '\0';
1181 return (char *)(uuid->uuid);
1184 #define LUSTRE_MAXFSNAME 8
1185 #define LUSTRE_MAXINSTANCE 16
1187 /* Extract fsname from uuid (or target name) of a target
1188 e.g. (myfs-OST0007_UUID -> myfs)
1189 see also deuuidify. */
1190 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1194 strncpy(buf, uuid, buflen - 1);
1195 buf[buflen - 1] = '\0';
1196 p = strrchr(buf, '-');
1201 /* printf display format for Lustre FIDs
1202 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1203 #define FID_NOBRACE_LEN 40
1204 #define FID_LEN (FID_NOBRACE_LEN + 2)
1205 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1206 #define DFID "[" DFID_NOBRACE "]"
1207 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1209 /* scanf input parse format for fids in DFID_NOBRACE format
1210 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1211 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1212 #define SFID "0x%llx:0x%x:0x%x"
1213 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1215 /********* Quotas **********/
1217 #define LUSTRE_QUOTABLOCK_BITS 10
1218 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1220 static inline __u64 lustre_stoqb(__kernel_size_t space)
1222 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1225 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1226 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1227 #define Q_GETOINFO 0x800102 /* get obd quota info */
1228 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1229 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1231 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1232 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1233 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1234 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1235 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1236 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1237 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1238 /* lustre-specific control commands */
1239 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1240 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1241 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1242 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1243 #define LUSTRE_Q_GETQUOTAPOOL 0x80000f /* get user pool quota */
1244 #define LUSTRE_Q_SETQUOTAPOOL 0x800010 /* set user pool quota */
1245 #define LUSTRE_Q_GETINFOPOOL 0x800011 /* get pool quota info */
1246 #define LUSTRE_Q_SETINFOPOOL 0x800012 /* set pool quota info */
1247 /* In the current Lustre implementation, the grace time is either the time
1248 * or the timestamp to be used after some quota ID exceeds the soft limt,
1249 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1251 #define LQUOTA_GRACE_BITS 48
1252 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1253 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1254 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1255 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1256 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1258 /* special grace time, only notify the user when its quota is over soft limit
1259 * but doesn't block new writes until the hard limit is reached. */
1260 #define NOTIFY_GRACE "notify"
1261 #define NOTIFY_GRACE_TIME LQUOTA_GRACE_MASK
1263 /* different quota flags */
1265 /* the default quota flag, the corresponding quota ID will use the default
1266 * quota setting, the hardlimit and softlimit of its quota record in the global
1267 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1268 * and high 16 bits will contain this flag (see above comment).
1270 #define LQUOTA_FLAG_DEFAULT 0x0001
1272 #define LUSTRE_Q_CMD_IS_POOL(cmd) \
1273 (cmd == LUSTRE_Q_GETQUOTAPOOL || \
1274 cmd == LUSTRE_Q_SETQUOTAPOOL || \
1275 cmd == LUSTRE_Q_SETINFOPOOL || \
1276 cmd == LUSTRE_Q_GETINFOPOOL)
1278 #define ALLQUOTA 255 /* set all quota */
1279 static inline const char *qtype_name(int qtype)
1292 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1293 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1296 #define N_PERMS_MAX 64
1298 struct perm_downcall_data {
1304 struct identity_downcall_data {
1311 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1312 __u32 idd_groups[0];
1315 struct sepol_downcall_data {
1317 __s64 sdd_sepol_mtime;
1318 __u16 sdd_sepol_len;
1322 #ifdef NEED_QUOTA_DEFS
1324 #define QIF_BLIMITS 1
1326 #define QIF_ILIMITS 4
1327 #define QIF_INODES 8
1328 #define QIF_BTIME 16
1329 #define QIF_ITIME 32
1330 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1331 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1332 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1333 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1336 #endif /* !__KERNEL__ */
1338 /* lustre volatile file support
1339 * file name header: ".^L^S^T^R:volatile"
1341 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1342 #define LUSTRE_VOLATILE_HDR_LEN 14
1344 enum lustre_quota_version {
1348 /* XXX: same as if_dqinfo struct in kernel */
1356 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1358 __u64 dqb_bhardlimit;
1359 __u64 dqb_bsoftlimit;
1361 __u64 dqb_ihardlimit;
1362 __u64 dqb_isoftlimit;
1363 __u64 dqb_curinodes;
1377 struct if_quotactl {
1384 struct obd_dqinfo qc_dqinfo;
1385 struct obd_dqblk qc_dqblk;
1387 struct obd_uuid obd_uuid;
1388 char qc_poolname[0];
1391 /* swap layout flags */
1392 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1393 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1394 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1395 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1396 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1398 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1399 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1400 struct lustre_swap_layouts {
1408 /** Bit-mask of valid attributes */
1409 /* The LA_* flags are written to disk as part of the ChangeLog records
1410 * so they are part of the on-disk and network protocol, and cannot be changed.
1411 * Only the first 12 bits are currently saved.
1414 LA_ATIME = 1 << 0, /* 0x00001 */
1415 LA_MTIME = 1 << 1, /* 0x00002 */
1416 LA_CTIME = 1 << 2, /* 0x00004 */
1417 LA_SIZE = 1 << 3, /* 0x00008 */
1418 LA_MODE = 1 << 4, /* 0x00010 */
1419 LA_UID = 1 << 5, /* 0x00020 */
1420 LA_GID = 1 << 6, /* 0x00040 */
1421 LA_BLOCKS = 1 << 7, /* 0x00080 */
1422 LA_TYPE = 1 << 8, /* 0x00100 */
1423 LA_FLAGS = 1 << 9, /* 0x00200 */
1424 LA_NLINK = 1 << 10, /* 0x00400 */
1425 LA_RDEV = 1 << 11, /* 0x00800 */
1426 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1427 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1428 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1429 LA_PROJID = 1 << 15, /* 0x08000 */
1430 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1431 LA_LSIZE = 1 << 17, /* 0x20000 */
1432 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1433 LA_BTIME = 1 << 19, /* 0x80000 */
1435 * Attributes must be transmitted to OST objects
1437 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1440 #define MDS_FMODE_READ 00000001
1441 #define MDS_FMODE_WRITE 00000002
1443 #define MDS_FMODE_CLOSED 00000000
1444 #define MDS_FMODE_EXEC 00000004
1445 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1446 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1447 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1449 #define MDS_OPEN_CREATED 00000010
1450 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1452 #define MDS_OPEN_CREAT 00000100
1453 #define MDS_OPEN_EXCL 00000200
1454 #define MDS_OPEN_TRUNC 00001000
1455 #define MDS_OPEN_APPEND 00002000
1456 #define MDS_OPEN_SYNC 00010000
1457 #define MDS_OPEN_DIRECTORY 00200000
1459 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1460 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1461 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1462 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1463 * We do not support JOIN FILE
1464 * anymore, reserve this flags
1465 * just for preventing such bit
1468 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1469 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1470 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1471 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1472 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1474 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1476 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1477 * delegation, succeed if it's not
1478 * being opened with conflict mode.
1480 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1482 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1483 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1484 * for newly created file */
1486 /* lustre internal open flags, which should not be set from user space */
1487 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1488 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1489 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1490 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1494 /********* Changelogs **********/
1495 /** Changelog record types */
1496 enum changelog_rec_type {
1499 CL_CREATE = 1, /* namespace */
1500 CL_MKDIR = 2, /* namespace */
1501 CL_HARDLINK = 3, /* namespace */
1502 CL_SOFTLINK = 4, /* namespace */
1503 CL_MKNOD = 5, /* namespace */
1504 CL_UNLINK = 6, /* namespace */
1505 CL_RMDIR = 7, /* namespace */
1506 CL_RENAME = 8, /* namespace */
1507 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1508 CL_OPEN = 10, /* not currently used */
1509 CL_CLOSE = 11, /* may be written to log only with mtime change */
1510 CL_LAYOUT = 12, /* file layout/striping modified */
1514 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1515 CL_HSM = 16, /* HSM specific events, see flags */
1516 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1520 CL_FLRW = 21, /* FLR: file was firstly written */
1521 CL_RESYNC = 22, /* FLR: file was resync-ed */
1523 CL_DN_OPEN = 24, /* denied open */
1527 static inline const char *changelog_type2str(int type) {
1528 static const char *changelog_str[] = {
1529 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1530 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1531 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1532 "FLRW", "RESYNC","GXATR", "NOPEN",
1535 if (type >= 0 && type < CL_LAST)
1536 return changelog_str[type];
1540 /* 12 bits of per-record data can be stored in the bottom of the flags */
1541 #define CLF_FLAGSHIFT 12
1542 enum changelog_rec_flags {
1543 CLF_VERSION = 0x1000,
1544 CLF_RENAME = 0x2000,
1546 CLF_EXTRA_FLAGS = 0x8000,
1547 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1549 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1550 CLF_VERMASK = ~CLF_FLAGMASK,
1554 /* Anything under the flagmask may be per-type (if desired) */
1555 /* Flags for unlink */
1556 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1557 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1558 /* HSM cleaning needed */
1559 /* Flags for rename */
1560 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1562 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1563 * has an archive in backend */
1566 /* 12b used (from high weight to low weight):
1571 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1572 #define CLF_HSM_ERR_H 6
1573 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1574 #define CLF_HSM_EVENT_H 9
1575 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1576 #define CLF_HSM_FLAG_H 11
1577 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1578 #define CLF_HSM_SPARE_H 15
1579 #define CLF_HSM_LAST 15
1581 /* Remove bits higher than _h, then extract the value
1582 * between _h and _l by shifting lower weigth to bit 0. */
1583 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1584 >> (CLF_HSM_LAST - _h + _l))
1586 #define CLF_HSM_SUCCESS 0x00
1587 #define CLF_HSM_MAXERROR 0x7E
1588 #define CLF_HSM_ERROVERFLOW 0x7F
1590 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1592 /* 3 bits field => 8 values allowed */
1604 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1606 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1610 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1613 *clf_flags = (enum changelog_rec_flags)
1614 (*clf_flags | (he << CLF_HSM_EVENT_L));
1617 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1619 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1622 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1625 *clf_flags = (enum changelog_rec_flags)
1626 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1629 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1631 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1634 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1637 *clf_flags = (enum changelog_rec_flags)
1638 (*clf_flags | (error << CLF_HSM_ERR_L));
1641 enum changelog_rec_extra_flags {
1643 CLFE_UIDGID = 0x0001,
1646 CLFE_XATTR = 0x0008,
1647 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1650 enum changelog_send_flag {
1651 /* Not yet implemented */
1652 CHANGELOG_FLAG_FOLLOW = 0x01,
1653 /* Blocking IO makes sense in case of slow user parsing of the records,
1654 * but it also prevents us from cleaning up if the records are not
1656 CHANGELOG_FLAG_BLOCK = 0x02,
1657 /* Pack jobid into the changelog records if available. */
1658 CHANGELOG_FLAG_JOBID = 0x04,
1659 /* Pack additional flag bits into the changelog record */
1660 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1663 enum changelog_send_extra_flag {
1664 /* Pack uid/gid into the changelog record */
1665 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1666 /* Pack nid into the changelog record */
1667 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1668 /* Pack open mode into the changelog record */
1669 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1670 /* Pack xattr name into the changelog record */
1671 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1674 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1675 changelog_rec_offset(CLF_SUPPORTED, \
1678 /* 31 usable bytes string + null terminator. */
1679 #define LUSTRE_JOBID_SIZE 32
1681 /* This is the minimal changelog record. It can contain extensions
1682 * such as rename fields or process jobid. Its exact content is described
1683 * by the cr_flags and cr_extra_flags.
1685 * Extensions are packed in the same order as their corresponding flags,
1686 * then in the same order as their corresponding extra flags.
1688 struct changelog_rec {
1690 __u16 cr_flags; /**< \a changelog_rec_flags */
1691 __u32 cr_type; /**< \a changelog_rec_type */
1692 __u64 cr_index; /**< changelog record number */
1693 __u64 cr_prev; /**< last index for this target fid */
1696 struct lu_fid cr_tfid; /**< target fid */
1697 __u32 cr_markerflags; /**< CL_MARK flags */
1699 struct lu_fid cr_pfid; /**< parent fid */
1700 } __attribute__ ((packed));
1702 /* Changelog extension for RENAME. */
1703 struct changelog_ext_rename {
1704 struct lu_fid cr_sfid; /**< source fid, or zero */
1705 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1708 /* Changelog extension to include JOBID. */
1709 struct changelog_ext_jobid {
1710 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1713 /* Changelog extension to include additional flags. */
1714 struct changelog_ext_extra_flags {
1715 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1718 /* Changelog extra extension to include UID/GID. */
1719 struct changelog_ext_uidgid {
1724 /* Changelog extra extension to include NID. */
1725 struct changelog_ext_nid {
1726 /* have __u64 instead of lnet_nid_t type for use by client api */
1728 /* for use when IPv6 support is added */
1733 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1734 struct changelog_ext_openmode {
1738 /* Changelog extra extension to include xattr */
1739 struct changelog_ext_xattr {
1740 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1743 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1744 const struct changelog_rec *rec);
1746 static inline __kernel_size_t changelog_rec_offset(enum changelog_rec_flags crf,
1747 enum changelog_rec_extra_flags cref)
1749 __kernel_size_t size = sizeof(struct changelog_rec);
1751 if (crf & CLF_RENAME)
1752 size += sizeof(struct changelog_ext_rename);
1754 if (crf & CLF_JOBID)
1755 size += sizeof(struct changelog_ext_jobid);
1757 if (crf & CLF_EXTRA_FLAGS) {
1758 size += sizeof(struct changelog_ext_extra_flags);
1759 if (cref & CLFE_UIDGID)
1760 size += sizeof(struct changelog_ext_uidgid);
1761 if (cref & CLFE_NID)
1762 size += sizeof(struct changelog_ext_nid);
1763 if (cref & CLFE_OPEN)
1764 size += sizeof(struct changelog_ext_openmode);
1765 if (cref & CLFE_XATTR)
1766 size += sizeof(struct changelog_ext_xattr);
1772 static inline __kernel_size_t changelog_rec_size(const struct changelog_rec *rec)
1774 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1776 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1777 cref = (enum changelog_rec_extra_flags)
1778 changelog_rec_extra_flags(rec)->cr_extra_flags;
1780 return changelog_rec_offset(
1781 (enum changelog_rec_flags)rec->cr_flags, cref);
1784 static inline __kernel_size_t changelog_rec_varsize(const struct changelog_rec *rec)
1786 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1790 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1792 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1793 (rec->cr_flags & CLF_VERSION);
1795 return (struct changelog_ext_rename *)((char *)rec +
1796 changelog_rec_offset(crf,
1800 /* The jobid follows the rename extension, if present */
1802 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1804 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1805 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1807 return (struct changelog_ext_jobid *)((char *)rec +
1808 changelog_rec_offset(crf,
1812 /* The additional flags follow the rename and jobid extensions, if present */
1814 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1815 const struct changelog_rec *rec)
1817 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1818 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1820 return (struct changelog_ext_extra_flags *)((char *)rec +
1821 changelog_rec_offset(crf,
1825 /* The uid/gid is the first extra extension */
1827 struct changelog_ext_uidgid *changelog_rec_uidgid(
1828 const struct changelog_rec *rec)
1830 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1832 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1834 return (struct changelog_ext_uidgid *)((char *)rec +
1835 changelog_rec_offset(crf,
1839 /* The nid is the second extra extension */
1841 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1843 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1845 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1846 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1848 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1849 cref = (enum changelog_rec_extra_flags)
1850 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1853 return (struct changelog_ext_nid *)((char *)rec +
1854 changelog_rec_offset(crf, cref));
1857 /* The OPEN mode is the third extra extension */
1859 struct changelog_ext_openmode *changelog_rec_openmode(
1860 const struct changelog_rec *rec)
1862 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1864 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1865 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1867 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1868 cref = (enum changelog_rec_extra_flags)
1869 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1870 (CLFE_UIDGID | CLFE_NID));
1873 return (struct changelog_ext_openmode *)((char *)rec +
1874 changelog_rec_offset(crf, cref));
1877 /* The xattr name is the fourth extra extension */
1879 struct changelog_ext_xattr *changelog_rec_xattr(
1880 const struct changelog_rec *rec)
1882 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1884 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1885 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1887 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1888 cref = (enum changelog_rec_extra_flags)
1889 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1890 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
1892 return (struct changelog_ext_xattr *)((char *)rec +
1893 changelog_rec_offset(crf, cref));
1896 /* The name follows the rename, jobid and extra flags extns, if present */
1897 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1899 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1901 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1902 cref = (enum changelog_rec_extra_flags)
1903 changelog_rec_extra_flags(rec)->cr_extra_flags;
1905 return (char *)rec + changelog_rec_offset(
1906 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
1907 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
1910 static inline __kernel_size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1912 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1915 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1917 char *cr_name = changelog_rec_name(rec);
1919 return cr_name + strlen(cr_name) + 1;
1923 * Remap a record to the desired format as specified by the crf flags.
1924 * The record must be big enough to contain the final remapped version.
1925 * Superfluous extension fields are removed and missing ones are added
1926 * and zeroed. The flags of the record are updated accordingly.
1928 * The jobid and rename extensions can be added to a record, to match the
1929 * format an application expects, typically. In this case, the newly added
1930 * fields will be zeroed.
1931 * The Jobid field can be removed, to guarantee compatibility with older
1932 * clients that don't expect this field in the records they process.
1934 * The following assumptions are being made:
1935 * - CLF_RENAME will not be removed
1936 * - CLF_JOBID will not be added without CLF_RENAME being added too
1937 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1939 * @param[in,out] rec The record to remap.
1940 * @param[in] crf_wanted Flags describing the desired extensions.
1941 * @param[in] cref_want Flags describing the desired extra extensions.
1943 static inline void changelog_remap_rec(struct changelog_rec *rec,
1944 enum changelog_rec_flags crf_wanted,
1945 enum changelog_rec_extra_flags cref_want)
1947 char *xattr_mov = NULL;
1948 char *omd_mov = NULL;
1949 char *nid_mov = NULL;
1950 char *uidgid_mov = NULL;
1954 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1956 crf_wanted = (enum changelog_rec_flags)
1957 (crf_wanted & CLF_SUPPORTED);
1958 cref_want = (enum changelog_rec_extra_flags)
1959 (cref_want & CLFE_SUPPORTED);
1961 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1962 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1963 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1964 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1970 /* First move the variable-length name field */
1971 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1972 changelog_rec_name(rec), rec->cr_namelen);
1974 /* Locations of extensions in the remapped record */
1975 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1976 xattr_mov = (char *)rec +
1977 changelog_rec_offset(
1978 (enum changelog_rec_flags)
1979 (crf_wanted & CLF_SUPPORTED),
1980 (enum changelog_rec_extra_flags)
1981 (cref_want & ~CLFE_XATTR));
1982 omd_mov = (char *)rec +
1983 changelog_rec_offset(
1984 (enum changelog_rec_flags)
1985 (crf_wanted & CLF_SUPPORTED),
1986 (enum changelog_rec_extra_flags)
1987 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
1988 nid_mov = (char *)rec +
1989 changelog_rec_offset(
1990 (enum changelog_rec_flags)
1991 (crf_wanted & CLF_SUPPORTED),
1992 (enum changelog_rec_extra_flags)
1994 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
1995 uidgid_mov = (char *)rec +
1996 changelog_rec_offset(
1997 (enum changelog_rec_flags)
1998 (crf_wanted & CLF_SUPPORTED),
1999 (enum changelog_rec_extra_flags)
2000 (cref_want & ~(CLFE_UIDGID |
2004 cref = (enum changelog_rec_extra_flags)
2005 changelog_rec_extra_flags(rec)->cr_extra_flags;
2008 ef_mov = (char *)rec +
2009 changelog_rec_offset(
2010 (enum changelog_rec_flags)
2011 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
2012 jid_mov = (char *)rec +
2013 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2014 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
2016 rnm_mov = (char *)rec +
2017 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2023 /* Move the extension fields to the desired positions */
2024 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2025 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
2026 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
2027 memmove(xattr_mov, changelog_rec_xattr(rec),
2028 sizeof(struct changelog_ext_xattr));
2030 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
2031 memmove(omd_mov, changelog_rec_openmode(rec),
2032 sizeof(struct changelog_ext_openmode));
2034 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
2035 memmove(nid_mov, changelog_rec_nid(rec),
2036 sizeof(struct changelog_ext_nid));
2038 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
2039 memmove(uidgid_mov, changelog_rec_uidgid(rec),
2040 sizeof(struct changelog_ext_uidgid));
2042 memmove(ef_mov, changelog_rec_extra_flags(rec),
2043 sizeof(struct changelog_ext_extra_flags));
2046 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
2047 memmove(jid_mov, changelog_rec_jobid(rec),
2048 sizeof(struct changelog_ext_jobid));
2050 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
2051 memmove(rnm_mov, changelog_rec_rename(rec),
2052 sizeof(struct changelog_ext_rename));
2054 /* Clear newly added fields */
2055 if (xattr_mov && (cref_want & CLFE_XATTR) &&
2056 !(cref & CLFE_XATTR))
2057 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
2059 if (omd_mov && (cref_want & CLFE_OPEN) &&
2060 !(cref & CLFE_OPEN))
2061 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
2063 if (nid_mov && (cref_want & CLFE_NID) &&
2065 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
2067 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
2068 !(cref & CLFE_UIDGID))
2069 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
2071 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2072 !(rec->cr_flags & CLF_EXTRA_FLAGS))
2073 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
2075 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
2076 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
2078 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
2079 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
2081 /* Update the record's flags accordingly */
2082 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
2083 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2084 changelog_rec_extra_flags(rec)->cr_extra_flags =
2085 changelog_rec_extra_flags(rec)->cr_extra_flags |
2089 enum changelog_message_type {
2090 CL_RECORD = 10, /* message is a changelog_rec */
2091 CL_EOF = 11, /* at end of current changelog */
2094 /********* Misc **********/
2096 struct ioc_data_version {
2098 __u32 idv_layout_version; /* FLR: layout version for OST objects */
2099 __u32 idv_flags; /* enum ioc_data_version_flags */
2102 enum ioc_data_version_flags {
2103 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
2104 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
2108 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
2111 #define dot_lustre_name ".lustre"
2114 /********* HSM **********/
2116 /** HSM per-file state
2117 * See HSM_FLAGS below.
2120 HS_NONE = 0x00000000,
2121 HS_EXISTS = 0x00000001,
2122 HS_DIRTY = 0x00000002,
2123 HS_RELEASED = 0x00000004,
2124 HS_ARCHIVED = 0x00000008,
2125 HS_NORELEASE = 0x00000010,
2126 HS_NOARCHIVE = 0x00000020,
2127 HS_LOST = 0x00000040,
2130 /* HSM user-setable flags. */
2131 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
2133 /* Other HSM flags. */
2134 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
2137 * All HSM-related possible flags that could be applied to a file.
2138 * This should be kept in sync with hsm_states.
2140 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
2143 * HSM request progress state
2145 enum hsm_progress_states {
2152 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
2155 case HPS_WAITING: return "waiting";
2156 case HPS_RUNNING: return "running";
2157 case HPS_DONE: return "done";
2158 default: return "unknown";
2165 } __attribute__((packed));
2168 * Current HSM states of a Lustre file.
2170 * This structure purpose is to be sent to user-space mainly. It describes the
2171 * current HSM flags and in-progress action.
2173 struct hsm_user_state {
2174 /** Current HSM states, from enum hsm_states. */
2176 __u32 hus_archive_id;
2177 /** The current undergoing action, if there is one */
2178 __u32 hus_in_progress_state;
2179 __u32 hus_in_progress_action;
2180 struct hsm_extent hus_in_progress_location;
2181 char hus_extended_info[];
2184 struct hsm_state_set_ioc {
2185 struct lu_fid hssi_fid;
2187 __u64 hssi_clearmask;
2191 * This structure describes the current in-progress action for a file.
2192 * it is retuned to user space and send over the wire
2194 struct hsm_current_action {
2195 /** The current undergoing action, if there is one */
2196 /* state is one of hsm_progress_states */
2198 /* action is one of hsm_user_action */
2200 struct hsm_extent hca_location;
2203 /***** HSM user requests ******/
2204 /* User-generated (lfs/ioctl) request types */
2205 enum hsm_user_action {
2206 HUA_NONE = 1, /* no action (noop) */
2207 HUA_ARCHIVE = 10, /* copy to hsm */
2208 HUA_RESTORE = 11, /* prestage */
2209 HUA_RELEASE = 12, /* drop ost objects */
2210 HUA_REMOVE = 13, /* remove from archive */
2211 HUA_CANCEL = 14 /* cancel a request */
2214 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2217 case HUA_NONE: return "NOOP";
2218 case HUA_ARCHIVE: return "ARCHIVE";
2219 case HUA_RESTORE: return "RESTORE";
2220 case HUA_RELEASE: return "RELEASE";
2221 case HUA_REMOVE: return "REMOVE";
2222 case HUA_CANCEL: return "CANCEL";
2223 default: return "UNKNOWN";
2228 * List of hr_flags (bit field)
2230 #define HSM_FORCE_ACTION 0x0001
2231 /* used by CT, cannot be set by user */
2232 #define HSM_GHOST_COPY 0x0002
2235 * Contains all the fixed part of struct hsm_user_request.
2238 struct hsm_request {
2239 __u32 hr_action; /* enum hsm_user_action */
2240 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2241 __u64 hr_flags; /* request flags */
2242 __u32 hr_itemcount; /* item count in hur_user_item vector */
2246 struct hsm_user_item {
2247 struct lu_fid hui_fid;
2248 struct hsm_extent hui_extent;
2249 } __attribute__((packed));
2251 struct hsm_user_request {
2252 struct hsm_request hur_request;
2253 struct hsm_user_item hur_user_item[0];
2254 /* extra data blob at end of struct (after all
2255 * hur_user_items), only use helpers to access it
2257 } __attribute__((packed));
2259 /** Return pointer to data field in a hsm user request */
2260 static inline void *hur_data(struct hsm_user_request *hur)
2262 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2266 * Compute the current length of the provided hsm_user_request. This returns -1
2267 * instead of an errno because __kernel_ssize_t is defined to be only
2270 * return -1 on bounds check error.
2272 static inline __kernel_size_t hur_len(struct hsm_user_request *hur)
2276 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2277 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2278 (__u64)hur->hur_request.hr_itemcount *
2279 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2281 if ((__kernel_ssize_t)size < 0)
2287 /****** HSM RPCs to copytool *****/
2288 /* Message types the copytool may receive */
2289 enum hsm_message_type {
2290 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2293 /* Actions the copytool may be instructed to take for a given action_item */
2294 enum hsm_copytool_action {
2295 HSMA_NONE = 10, /* no action */
2296 HSMA_ARCHIVE = 20, /* arbitrary offset */
2302 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2305 case HSMA_NONE: return "NOOP";
2306 case HSMA_ARCHIVE: return "ARCHIVE";
2307 case HSMA_RESTORE: return "RESTORE";
2308 case HSMA_REMOVE: return "REMOVE";
2309 case HSMA_CANCEL: return "CANCEL";
2310 default: return "UNKNOWN";
2314 /* Copytool item action description */
2315 struct hsm_action_item {
2316 __u32 hai_len; /* valid size of this struct */
2317 __u32 hai_action; /* hsm_copytool_action, but use known size */
2318 struct lu_fid hai_fid; /* Lustre FID to operate on */
2319 struct lu_fid hai_dfid; /* fid used for data access */
2320 struct hsm_extent hai_extent; /* byte range to operate on */
2321 __u64 hai_cookie; /* action cookie from coordinator */
2322 __u64 hai_gid; /* grouplock id */
2323 char hai_data[0]; /* variable length */
2324 } __attribute__((packed));
2327 * helper function which print in hexa the first bytes of
2330 * \param hai [IN] record to print
2331 * \param buffer [IN,OUT] buffer to write the hex string to
2332 * \param len [IN] max buffer length
2336 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2337 char *buffer, __kernel_size_t len)
2344 data_len = hai->hai_len - sizeof(*hai);
2345 for (i = 0; (i < data_len) && (len > 2); i++) {
2346 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2356 /* Copytool action list */
2357 #define HAL_VERSION 1
2358 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2359 struct hsm_action_list {
2361 __u32 hal_count; /* number of hai's to follow */
2362 __u64 hal_compound_id; /* returned by coordinator, ignored */
2364 __u32 hal_archive_id; /* which archive backend */
2366 char hal_fsname[0]; /* null-terminated */
2367 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2368 boundaries. See hai_zero */
2369 } __attribute__((packed));
2371 /* Return pointer to first hai in action list */
2372 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2374 __kernel_size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2376 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2379 /* Return pointer to next hai */
2380 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2382 __kernel_size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2384 return (struct hsm_action_item *)((char *)hai + offset);
2387 /* Return size of an hsm_action_list */
2388 static inline __kernel_size_t hal_size(struct hsm_action_list *hal)
2392 struct hsm_action_item *hai;
2394 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2395 hai = hai_first(hal);
2396 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2397 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2403 * describe the attributes to be set on imported file
2405 struct hsm_user_import {
2414 __u32 hui_archive_id;
2417 /* Copytool progress reporting */
2418 #define HP_FLAG_COMPLETED 0x01
2419 #define HP_FLAG_RETRY 0x02
2421 struct hsm_progress {
2422 struct lu_fid hp_fid;
2424 struct hsm_extent hp_extent;
2426 __u16 hp_errval; /* positive val */
2431 __u64 hc_data_version;
2433 __u16 hc_errval; /* positive val */
2435 struct hsm_action_item hc_hai;
2439 enum llapi_json_types {
2440 LLAPI_JSON_INTEGER = 1,
2446 struct llapi_json_item {
2455 struct llapi_json_item *lji_next;
2458 struct llapi_json_item_list {
2459 int ljil_item_count;
2460 struct llapi_json_item *ljil_items;
2463 enum lu_ladvise_type {
2464 LU_LADVISE_INVALID = 0,
2465 LU_LADVISE_WILLREAD = 1,
2466 LU_LADVISE_DONTNEED = 2,
2467 LU_LADVISE_LOCKNOEXPAND = 3,
2468 LU_LADVISE_LOCKAHEAD = 4,
2472 #define LU_LADVISE_NAMES { \
2473 [LU_LADVISE_WILLREAD] = "willread", \
2474 [LU_LADVISE_DONTNEED] = "dontneed", \
2475 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2476 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2479 /* This is the userspace argument for ladvise. It is currently the same as
2480 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2481 * may need info which is only used locally. */
2482 struct llapi_lu_ladvise {
2483 __u16 lla_advice; /* advice type */
2484 __u16 lla_value1; /* values for different advice types */
2486 __u64 lla_start; /* first byte of extent for advice */
2487 __u64 lla_end; /* last byte of extent for advice */
2493 LF_ASYNC = 0x00000001,
2494 LF_UNSET = 0x00000002,
2497 #define LADVISE_MAGIC 0x1ADF1CE0
2498 /* Masks of valid flags for each advice */
2499 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2500 /* Flags valid for all advices not explicitly specified */
2501 #define LF_DEFAULT_MASK LF_ASYNC
2503 #define LF_MASK (LF_ASYNC | LF_UNSET)
2505 #define lla_lockahead_mode lla_value1
2506 #define lla_peradvice_flags lla_value2
2507 #define lla_lockahead_result lla_value3
2509 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2510 * is used on the wire. It is defined separately as we may need info which is
2511 * only used locally. */
2512 struct llapi_ladvise_hdr {
2513 __u32 lah_magic; /* LADVISE_MAGIC */
2514 __u32 lah_count; /* number of advices */
2515 __u64 lah_flags; /* from enum ladvise_flag */
2516 __u32 lah_value1; /* unused */
2517 __u32 lah_value2; /* unused */
2518 __u64 lah_value3; /* unused */
2519 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2522 #define LAH_COUNT_MAX (1024)
2526 SK_CRYPT_INVALID = -1,
2528 SK_CRYPT_AES256_CTR = 1,
2532 SK_HMAC_INVALID = -1,
2538 struct sk_crypt_type {
2539 const char *sct_name;
2543 struct sk_hmac_type {
2544 const char *sht_name;
2548 enum lock_mode_user {
2554 #define LOCK_MODE_NAMES { \
2555 [MODE_READ_USER] = "READ",\
2556 [MODE_WRITE_USER] = "WRITE"\
2559 enum lockahead_results {
2560 LLA_RESULT_SENT = 0,
2561 LLA_RESULT_DIFFERENT,
2565 enum lu_heat_flag_bit {
2566 LU_HEAT_FLAG_BIT_INVALID = 0,
2567 LU_HEAT_FLAG_BIT_OFF,
2568 LU_HEAT_FLAG_BIT_CLEAR,
2572 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2573 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2576 enum obd_heat_type {
2577 OBD_HEAT_READSAMPLE = 0,
2578 OBD_HEAT_WRITESAMPLE = 1,
2579 OBD_HEAT_READBYTE = 2,
2580 OBD_HEAT_WRITEBYTE = 3,
2584 #define LU_HEAT_NAMES { \
2585 [OBD_HEAT_READSAMPLE] = "readsample", \
2586 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2587 [OBD_HEAT_READBYTE] = "readbyte", \
2588 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2603 static inline const char *pcc_type2string(enum lu_pcc_type type)
2608 case LU_PCC_READWRITE:
2615 struct lu_pcc_attach {
2616 __u32 pcca_type; /* PCC type */
2617 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2620 enum lu_pcc_detach_opts {
2621 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2622 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2625 struct lu_pcc_detach_fid {
2626 /* fid of the file to detach */
2627 struct lu_fid pccd_fid;
2631 struct lu_pcc_detach {
2635 enum lu_pcc_state_flags {
2636 PCC_STATE_FL_NONE = 0x0,
2637 /* The inode attr is cached locally */
2638 PCC_STATE_FL_ATTR_VALID = 0x01,
2639 /* The file is being attached into PCC */
2640 PCC_STATE_FL_ATTACHING = 0x02,
2643 struct lu_pcc_state {
2644 __u32 pccs_type; /* enum lu_pcc_type */
2645 __u32 pccs_open_count;
2646 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2648 char pccs_path[PATH_MAX];
2653 /* make header's size equal lu_fid */
2656 struct lu_fid fa_fids[0];
2658 #define OBD_MAX_FIDS_IN_ARRAY 4096
2660 #if defined(__cplusplus)
2664 /** @} lustreuser */
2666 #endif /* _LUSTRE_USER_H */