4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
31 * lustre/include/lustre/lustre_user.h
33 * Lustre public user-space interface definitions.
36 #ifndef _LUSTRE_USER_H
37 #define _LUSTRE_USER_H
39 /** \defgroup lustreuser lustreuser
44 # define __USE_ISOC99 1
46 # include <stdio.h> /* snprintf() */
47 # include <stdlib.h> /* abs() */
49 # include <sys/stat.h>
52 # define __USE_XOPEN2K8 1
53 # define FILEID_LUSTRE 0x97 /* for name_to_handle_at() (and llapi_fd2fid()) */
54 #endif /* !__KERNEL__ */
57 #include <linux/limits.h>
58 #include <linux/kernel.h>
59 #include <linux/string.h>
60 #include <linux/quota.h>
61 #include <linux/types.h>
62 #include <linux/unistd.h>
63 #include <linux/lustre/lustre_fiemap.h>
64 #include <linux/lustre/lustre_ver.h>
66 #if defined(__cplusplus)
70 #ifdef __STRICT_ANSI__
71 #define typeof __typeof__
75 * This is a temporary solution of adding quota type.
76 * Should be removed as soon as system header is updated.
79 #define LL_MAXQUOTAS 3
81 #define INITQFNAMES { \
82 "user", /* USRQUOTA */ \
83 "group", /* GRPQUOTA */ \
84 "project", /* PRJQUOTA */ \
98 * We need to always use 64bit version because the structure
99 * is shared across entire cluster where 32bit and 64bit machines
102 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
103 typedef struct stat64 lstat_t;
104 #define lstat_f lstat64
105 #define fstat_f fstat64
106 #define fstatat_f fstatat64
108 typedef struct stat lstat_t;
109 #define lstat_f lstat
110 #define fstat_f fstat
111 #define fstatat_f fstatat
114 #ifndef STATX_BASIC_STATS
116 * Timestamp structure for the timestamps in struct statx.
118 * tv_sec holds the number of seconds before (negative) or after (positive)
119 * 00:00:00 1st January 1970 UTC.
121 * tv_nsec holds a number of nanoseconds (0..999,999,999) after the tv_sec time.
123 * __reserved is held in case we need a yet finer resolution.
125 struct statx_timestamp {
132 * Structures for the extended file attribute retrieval system call
135 * The caller passes a mask of what they're specifically interested in as a
136 * parameter to statx(). What statx() actually got will be indicated in
137 * st_mask upon return.
139 * For each bit in the mask argument:
141 * - if the datum is not supported:
143 * - the bit will be cleared, and
145 * - the datum will be set to an appropriate fabricated value if one is
146 * available (eg. CIFS can take a default uid and gid), otherwise
148 * - the field will be cleared;
150 * - otherwise, if explicitly requested:
152 * - the datum will be synchronised to the server if AT_STATX_FORCE_SYNC is
153 * set or if the datum is considered out of date, and
155 * - the field will be filled in and the bit will be set;
157 * - otherwise, if not requested, but available in approximate form without any
158 * effort, it will be filled in anyway, and the bit will be set upon return
159 * (it might not be up to date, however, and no attempt will be made to
160 * synchronise the internal state first);
162 * - otherwise the field and the bit will be cleared before returning.
164 * Items in STATX_BASIC_STATS may be marked unavailable on return, but they
165 * will have values installed for compatibility purposes so that stat() and
166 * co. can be emulated in userspace.
170 __u32 stx_mask; /* What results were written [uncond] */
171 __u32 stx_blksize; /* Preferred general I/O size [uncond] */
172 __u64 stx_attributes; /* Flags conveying information about the file [uncond] */
174 __u32 stx_nlink; /* Number of hard links */
175 __u32 stx_uid; /* User ID of owner */
176 __u32 stx_gid; /* Group ID of owner */
177 __u16 stx_mode; /* File mode */
180 __u64 stx_ino; /* Inode number */
181 __u64 stx_size; /* File size */
182 __u64 stx_blocks; /* Number of 512-byte blocks allocated */
183 __u64 stx_attributes_mask; /* Mask to show what's supported in stx_attributes */
185 struct statx_timestamp stx_atime; /* Last access time */
186 struct statx_timestamp stx_btime; /* File creation time */
187 struct statx_timestamp stx_ctime; /* Last attribute change time */
188 struct statx_timestamp stx_mtime; /* Last data modification time */
190 __u32 stx_rdev_major; /* Device ID of special file [if bdev/cdev] */
191 __u32 stx_rdev_minor;
192 __u32 stx_dev_major; /* ID of device containing file [uncond] */
195 __u64 __spare2[14]; /* Spare space for future expansion */
200 * Flags to be stx_mask
202 * Query request/result mask for statx() and struct statx::stx_mask.
204 * These bits should be set in the mask argument of statx() to request
205 * particular items when calling statx().
207 #define STATX_TYPE 0x00000001U /* Want/got stx_mode & S_IFMT */
208 #define STATX_MODE 0x00000002U /* Want/got stx_mode & ~S_IFMT */
209 #define STATX_NLINK 0x00000004U /* Want/got stx_nlink */
210 #define STATX_UID 0x00000008U /* Want/got stx_uid */
211 #define STATX_GID 0x00000010U /* Want/got stx_gid */
212 #define STATX_ATIME 0x00000020U /* Want/got stx_atime */
213 #define STATX_MTIME 0x00000040U /* Want/got stx_mtime */
214 #define STATX_CTIME 0x00000080U /* Want/got stx_ctime */
215 #define STATX_INO 0x00000100U /* Want/got stx_ino */
216 #define STATX_SIZE 0x00000200U /* Want/got stx_size */
217 #define STATX_BLOCKS 0x00000400U /* Want/got stx_blocks */
218 #define STATX_BASIC_STATS 0x000007ffU /* The stuff in the normal stat struct */
219 #define STATX_BTIME 0x00000800U /* Want/got stx_btime */
220 #define STATX_ALL 0x00000fffU /* All currently supported flags */
221 #define STATX__RESERVED 0x80000000U /* Reserved for future struct statx expansion */
224 * Attributes to be found in stx_attributes and masked in stx_attributes_mask.
226 * These give information about the features or the state of a file that might
227 * be of use to ordinary userspace programs such as GUIs or ls rather than
230 * Note that the flags marked [I] correspond to generic FS_IOC_FLAGS
231 * semantically. Where possible, the numerical value is picked to correspond
234 #define STATX_ATTR_COMPRESSED 0x00000004 /* [I] File is compressed by the fs */
235 #define STATX_ATTR_IMMUTABLE 0x00000010 /* [I] File is marked immutable */
236 #define STATX_ATTR_APPEND 0x00000020 /* [I] File is append-only */
237 #define STATX_ATTR_NODUMP 0x00000040 /* [I] File is not to be dumped */
238 #define STATX_ATTR_ENCRYPTED 0x00000800 /* [I] File requires key to decrypt in fs */
240 #define STATX_ATTR_AUTOMOUNT 0x00001000 /* Dir: Automount trigger */
241 /* Update attrs_array in lustreapi.h if new attributes are added. */
243 #define AT_STATX_SYNC_TYPE 0x6000 /* Type of synchronisation required from statx() */
244 #define AT_STATX_SYNC_AS_STAT 0x0000 /* - Do whatever stat() does */
245 #define AT_STATX_FORCE_SYNC 0x2000 /* - Force the attributes to be sync'd with the server */
246 #define AT_STATX_DONT_SYNC 0x4000 /* - Don't sync attributes with the server */
248 #endif /* STATX_BASIC_STATS */
250 typedef struct statx lstatx_t;
252 #define LUSTRE_EOF 0xffffffffffffffffULL
255 #define LL_SUPER_MAGIC 0x0BD00BD0
257 #define LL_IOC_GETVERSION _IOR('f', 3, long)
258 #define FSFILT_IOC_GETVERSION LL_IOC_GETVERSION /* backward compat */
259 #define LL_IOC_RESIZE_FS _IOW('f', 16, __u64)
261 /* FIEMAP flags supported by Lustre */
262 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
264 enum obd_statfs_state {
265 OS_STATFS_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
266 OS_STATFS_READONLY = 0x00000002, /**< filesystem is read-only */
267 OS_STATFS_NOPRECREATE = 0x00000004, /**< no object precreation */
268 OS_STATFS_UNUSED1 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
269 OS_STATFS_UNUSED2 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
270 OS_STATFS_ENOSPC = 0x00000020, /**< not enough free space */
271 OS_STATFS_ENOINO = 0x00000040, /**< not enough inodes */
272 OS_STATFS_SUM = 0x00000100, /**< aggregated for all tagrets */
273 OS_STATFS_NONROT = 0x00000200, /**< non-rotational device */
276 /** filesystem statistics/attributes for target device */
278 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
279 __u64 os_blocks; /* total size in #os_bsize blocks */
280 __u64 os_bfree; /* number of unused blocks */
281 __u64 os_bavail; /* blocks available for allocation */
282 __u64 os_files; /* total number of objects */
283 __u64 os_ffree; /* # objects that could be created */
284 __u8 os_fsid[40]; /* identifier for filesystem */
285 __u32 os_bsize; /* block size in bytes for os_blocks */
286 __u32 os_namelen; /* maximum length of filename in bytes*/
287 __u64 os_maxbytes; /* maximum object size in bytes */
288 __u32 os_state; /**< obd_statfs_state OS_STATFS_* */
289 __u32 os_fprecreated; /* objs available now to the caller */
290 /* used in QoS code to find preferred
292 __u32 os_granted; /* space granted for MDS */
293 __u32 os_spare3; /* Unused padding fields. Remember */
294 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
302 /** additional filesystem attributes for target device */
303 struct obd_statfs_info {
304 __u32 os_reserved_mb_low; /* reserved mb low */
305 __u32 os_reserved_mb_high; /* reserved mb high */
306 bool os_enable_pre; /* enable pre create logic */
312 * FID is a cluster-wide unique identifier of a file or an object (stripe).
313 * FIDs are never reused.
317 * FID sequence. Sequence is a unit of migration: all files (objects)
318 * with FIDs from a given sequence are stored on the same server.
319 * Lustre should support 2^64 objects, so even if each sequence
320 * has only a single object we can still enumerate 2^64 objects.
323 /* FID number within sequence. */
326 * FID version, used to distinguish different versions (in the sense
327 * of snapshots, etc.) of the same file system object. Not currently
331 } __attribute__((packed));
333 static inline bool fid_is_zero(const struct lu_fid *fid)
335 return fid->f_seq == 0 && fid->f_oid == 0;
338 /* The data name_to_handle_at() places in a struct file_handle (at f_handle) */
339 struct lustre_file_handle {
340 struct lu_fid lfh_child;
341 struct lu_fid lfh_parent;
344 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
345 * MDT-object's FID::f_ver, instead it is the OST-object index in its
346 * parent MDT-object's layout EA. */
347 #define f_stripe_idx f_ver
350 __u32 ol_stripe_size;
351 __u32 ol_stripe_count;
355 } __attribute__((packed));
357 /* The filter_fid structure has changed several times over its lifetime.
358 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
359 * stripe_index and the "self FID" (objid/seq) to be able to recover the
360 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
361 * the OST, the "trusted.lma" xattr was added to the OST objects to store
362 * the "self FID" to be consistent with the MDT on-disk format, and the
363 * filter_fid only stored the MDT inode parent FID and stripe index.
365 * In 2.10, the addition of PFL composite layouts required more information
366 * to be stored into the filter_fid in order to be able to identify which
367 * component the OST object belonged. As well, the stripe size may vary
368 * between components, so it was no longer safe to assume the stripe size
369 * or stripe_count of a file. This is also more robust for plain layouts.
371 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
372 * enough space to store both the filter_fid and LMA in the inode, so they
373 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
374 * an extra seek for every OST object access.
376 * In 2.11, FLR mirror layouts also need to store the layout version and
377 * range so that writes to old versions of the layout are not allowed.
378 * That ensures that mirrored objects are not modified by evicted clients,
379 * and ensures that the components are correctly marked stale on the MDT.
381 struct filter_fid_18_23 {
382 struct lu_fid ff_parent; /* stripe_idx in f_ver */
387 struct filter_fid_24_29 {
388 struct lu_fid ff_parent; /* stripe_idx in f_ver */
391 struct filter_fid_210 {
392 struct lu_fid ff_parent; /* stripe_idx in f_ver */
393 struct ost_layout ff_layout;
397 struct lu_fid ff_parent; /* stripe_idx in f_ver */
398 struct ost_layout ff_layout;
399 __u32 ff_layout_version;
400 __u32 ff_range; /* range of layout version that
401 * write are allowed */
402 } __attribute__((packed));
404 /* Userspace should treat lu_fid as opaque, and only use the following methods
405 * to print or parse them. Other functions (e.g. compare, swab) could be moved
406 * here from lustre_idl.h if needed. */
410 LMAC_HSM = 0x00000001,
411 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
412 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
413 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
414 * under /O/<seq>/d<x>. */
415 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
416 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
417 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
421 * Masks for all features that should be supported by a Lustre version to
422 * access a specific file.
423 * This information is stored in lustre_mdt_attrs::lma_incompat.
426 LMAI_RELEASED = 0x00000001, /* file is released */
427 LMAI_AGENT = 0x00000002, /* agent inode */
428 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
429 is on the remote MDT */
430 LMAI_STRIPED = 0x00000008, /* striped directory inode */
431 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
432 LMAI_ENCRYPT = 0x00000020, /* inode is encrypted */
433 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
434 LMAI_STRIPED | LMAI_ORPHAN | LMAI_ENCRYPT)
439 * Following struct for object attributes, that will be kept inode's EA.
440 * Introduced in 2.0 release (please see b15993, for details)
441 * Added to all objects since Lustre 2.4 as contains self FID
443 struct lustre_mdt_attrs {
445 * Bitfield for supported data in this structure. From enum lma_compat.
446 * lma_self_fid and lma_flags are always available.
450 * Per-file incompat feature list. Lustre version should support all
451 * flags set in this field. The supported feature mask is available in
455 /** FID of this inode */
456 struct lu_fid lma_self_fid;
459 #define PFID_STRIPE_IDX_BITS 16
460 #define PFID_STRIPE_COUNT_MASK ((1 << PFID_STRIPE_IDX_BITS) - 1)
462 struct lustre_ost_attrs {
463 /* Use lustre_mdt_attrs directly for now, need a common header
464 * structure if want to change lustre_mdt_attrs in future. */
465 struct lustre_mdt_attrs loa_lma;
467 /* Below five elements are for OST-object's PFID EA, the
468 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
469 * and the stripe_index (low 16 bits), the size should not exceed
470 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
471 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
472 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
473 * loa_comp_* elements are valid. */
474 struct lu_fid loa_parent_fid;
475 __u32 loa_stripe_size;
477 __u64 loa_comp_start;
482 * Prior to 2.4, the LMA structure also included SOM attributes which has since
483 * been moved to a dedicated xattr
484 * lma_flags was also removed because of lma_compat/incompat fields.
486 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
488 enum lustre_som_flags {
489 /* Unknow or no SoM data, must get size from OSTs. */
490 SOM_FL_UNKNOWN = 0x0000,
491 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
492 SOM_FL_STRICT = 0x0001,
493 /* Known stale - was right at some point in the past, but it is
494 * known (or likely) to be incorrect now (e.g. opened for write). */
495 SOM_FL_STALE = 0x0002,
496 /* Approximate, may never have been strictly correct,
497 * need to sync SOM data to achieve eventual consistency. */
498 SOM_FL_LAZY = 0x0004,
501 struct lustre_som_attrs {
503 __u16 lsa_reserved[3];
509 * OST object IDentifier.
517 struct lu_fid oi_fid;
519 } __attribute__((packed));
521 #define DOSTID "%#llx:%llu"
522 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
523 ((unsigned long long)ostid_id(oi))
525 struct ll_futimes_3 {
527 __u64 lfu_atime_nsec;
529 __u64 lfu_mtime_nsec;
531 __u64 lfu_ctime_nsec;
535 * Maximum number of mirrors currently implemented.
537 #define LUSTRE_MIRROR_COUNT_MAX 16
539 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
541 LL_LEASE_RDLCK = 0x01,
542 LL_LEASE_WRLCK = 0x02,
543 LL_LEASE_UNLCK = 0x04,
546 enum ll_lease_flags {
547 LL_LEASE_RESYNC = 0x1,
548 LL_LEASE_RESYNC_DONE = 0x2,
549 LL_LEASE_LAYOUT_MERGE = 0x4,
550 LL_LEASE_LAYOUT_SPLIT = 0x8,
551 LL_LEASE_PCC_ATTACH = 0x10,
554 #define IOC_IDS_MAX 4096
555 struct ll_ioc_lease {
562 struct ll_ioc_lease_id {
573 * The ioctl naming rules:
574 * LL_* - works on the currently opened filehandle instead of parent dir
575 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
576 * *_MDC_* - gets/sets data related to MDC
577 * *_LOV_* - gets/sets data related to OSC/LOV
578 * *FILE* - called on parent dir and passes in a filename
579 * *STRIPE* - set/get lov_user_md
580 * *INFO - set/get lov_user_mds_data
582 /* lustre_ioctl.h 101-150 */
583 /* ioctl codes 128-143 are reserved for fsverity */
584 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
585 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
586 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
587 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
588 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
589 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
590 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
591 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
592 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
593 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
594 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
595 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
596 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
597 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
598 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
599 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
600 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
601 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
602 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
603 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
604 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
605 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
606 /* lustre_ioctl.h 177-210 */
607 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
608 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
609 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
610 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
611 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
612 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
613 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
614 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
615 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
616 struct lustre_swap_layouts)
617 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
618 struct hsm_current_action)
619 /* lustre_ioctl.h 221-233 */
620 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
621 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
622 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
623 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
624 #define LL_IOC_UNLOCK_FOREIGN _IO('f', 242)
625 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
626 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
627 #define LL_IOC_GET_LEASE _IO('f', 244)
628 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
629 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
630 #define LL_IOC_MIGRATE _IOR('f', 247, int)
631 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
632 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
633 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
634 #define LL_IOC_LADVISE2 _IOW('f', 250, struct llapi_lu_ladvise2)
635 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
636 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
637 #define LL_IOC_PCC_ATTACH _IOW('f', 252, struct lu_pcc_attach)
638 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
639 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
640 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
641 #define LL_IOC_PROJECT _IOW('f', 253, struct lu_project)
643 #ifndef FS_IOC_FSGETXATTR
645 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
648 __u32 fsx_xflags; /* xflags field value (get/set) */
649 __u32 fsx_extsize; /* extsize field value (get/set)*/
650 __u32 fsx_nextents; /* nextents field value (get) */
651 __u32 fsx_projid; /* project identifier (get/set) */
652 unsigned char fsx_pad[12];
654 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
655 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
657 #ifndef FS_XFLAG_PROJINHERIT
658 #define FS_XFLAG_PROJINHERIT 0x00000200
662 #define LL_STATFS_LMV 1
663 #define LL_STATFS_LOV 2
664 #define LL_STATFS_NODELAY 4
666 #define IOC_MDC_TYPE 'i'
667 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
668 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
669 #define IOC_MDC_GETFILEINFO_V1 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
670 #define IOC_MDC_GETFILEINFO_V2 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
671 #define LL_IOC_MDC_GETINFO_V1 _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
672 #define LL_IOC_MDC_GETINFO_V2 _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
673 #define IOC_MDC_GETFILEINFO IOC_MDC_GETFILEINFO_V1
674 #define LL_IOC_MDC_GETINFO LL_IOC_MDC_GETINFO_V1
676 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
678 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
679 * files, but are unlikely to be used in practice and are not harmful if
680 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
681 * devices and are safe for use on new files. See LU-4209. */
682 /* To be compatible with old statically linked binary we keep the check for
683 * the older 0100000000 flag. This is already removed upstream. LU-812. */
684 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
686 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
688 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
689 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
690 O_LOV_DELAY_CREATE_MASK)
691 /* O_CIPHERTEXT principle is similar to O_LOV_DELAY_CREATE above,
692 * for access to encrypted files without the encryption key.
694 #define O_CIPHERTEXT (O_NOCTTY | O_NDELAY | O_DSYNC)
696 #define LL_FILE_IGNORE_LOCK 0x00000001
697 #define LL_FILE_GROUP_LOCKED 0x00000002
698 #define LL_FILE_READAHEA 0x00000004
699 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
700 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
702 #define LOV_USER_MAGIC_V1 0x0BD10BD0
703 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
704 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
705 #define LOV_USER_MAGIC_V3 0x0BD30BD0
706 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
707 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
708 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
709 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
710 #define LOV_USER_MAGIC_SEL 0x0BD80BD0
712 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
713 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
714 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
716 #define LOV_PATTERN_NONE 0x000
717 #define LOV_PATTERN_RAID0 0x001
718 #define LOV_PATTERN_RAID1 0x002
719 #define LOV_PATTERN_MDT 0x100
720 #define LOV_PATTERN_OVERSTRIPING 0x200
721 #define LOV_PATTERN_FOREIGN 0x400
722 #define LOV_PATTERN_COMPRESS 0x800
724 #define LOV_PATTERN_F_MASK 0xffff0000
725 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
726 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
727 #define LOV_PATTERN_DEFAULT 0xffffffff
729 #define LOV_OFFSET_DEFAULT ((__u16)-1)
730 #define LMV_OFFSET_DEFAULT ((__u32)-1)
732 static inline bool lov_pattern_supported(__u32 pattern)
734 __u32 pattern_base = pattern & ~LOV_PATTERN_F_RELEASED;
736 return pattern_base == LOV_PATTERN_RAID0 ||
737 pattern_base == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
738 pattern_base == LOV_PATTERN_MDT;
741 /* RELEASED and MDT patterns are not valid in many places, so rather than
742 * having many extra checks on lov_pattern_supported, we have this separate
743 * check for non-released, non-DOM components
745 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
747 return pattern == LOV_PATTERN_RAID0 ||
748 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
752 #define LOV_MAXPOOLNAME 15
753 #define LOV_POOLNAMEF "%.15s"
754 /* The poolname "ignore" is used to force a component creation without pool */
755 #define LOV_POOL_IGNORE "ignore"
756 /* The poolname "inherit" is used to force a component to inherit the pool from
757 * parent or root directory
759 #define LOV_POOL_INHERIT "inherit"
760 /* The poolname "none" is deprecated in 2.15 (same behavior as "inherit") */
761 #define LOV_POOL_NONE "none"
763 static inline bool lov_pool_is_ignored(const char *pool)
765 return pool && strncmp(pool, LOV_POOL_IGNORE, LOV_MAXPOOLNAME) == 0;
768 static inline bool lov_pool_is_inherited(const char *pool)
770 return pool && (strncmp(pool, LOV_POOL_INHERIT, LOV_MAXPOOLNAME) == 0 ||
771 strncmp(pool, LOV_POOL_NONE, LOV_MAXPOOLNAME) == 0);
774 static inline bool lov_pool_is_reserved(const char *pool)
776 return lov_pool_is_ignored(pool) || lov_pool_is_inherited(pool);
779 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
780 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
781 #define LOV_MAX_STRIPE_COUNT_OLD 160
782 /* This calculation is crafted so that input of 4096 will result in 160
783 * which in turn is equal to old maximal stripe count.
784 * XXX: In fact this is too simpified for now, what it also need is to get
785 * ea_type argument to clearly know how much space each stripe consumes.
787 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
788 * allocation that is sufficient for the current generation of systems.
790 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
791 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
792 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
793 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
795 #define XATTR_LUSTRE_PREFIX "lustre."
796 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
798 /* Please update if XATTR_LUSTRE_LOV".set" groks more flags in the future */
799 #define allowed_lustre_lov(att) (strcmp((att), XATTR_LUSTRE_LOV".add") == 0 || \
800 strcmp((att), XATTR_LUSTRE_LOV".set") == 0 || \
801 strcmp((att), XATTR_LUSTRE_LOV".set.flags") == 0 || \
802 strcmp((att), XATTR_LUSTRE_LOV".del") == 0)
804 #define lov_user_ost_data lov_user_ost_data_v1
805 struct lov_user_ost_data_v1 { /* per-stripe data structure */
806 struct ost_id l_ost_oi; /* OST object ID */
807 __u32 l_ost_gen; /* generation of this OST index */
808 __u32 l_ost_idx; /* OST index in LOV */
809 } __attribute__((packed));
811 #define lov_user_md lov_user_md_v1
812 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
813 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
814 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
815 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
816 __u32 lmm_stripe_size; /* size of stripe in bytes */
817 __u16 lmm_stripe_count; /* num stripes in use for this object */
819 __u16 lmm_stripe_offset; /* starting stripe offset in
820 * lmm_objects, use when writing */
821 __u16 lmm_layout_gen; /* layout generation number
822 * used when reading */
824 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
825 } __attribute__((packed, __may_alias__));
827 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
828 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
829 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
830 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
831 __u32 lmm_stripe_size; /* size of stripe in bytes */
832 __u16 lmm_stripe_count; /* num stripes in use for this object */
834 __u16 lmm_stripe_offset; /* starting stripe offset in
835 * lmm_objects, use when writing */
836 __u16 lmm_layout_gen; /* layout generation number
837 * used when reading */
839 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
840 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
841 } __attribute__((packed, __may_alias__));
843 struct lov_foreign_md {
844 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
845 __u32 lfm_length; /* length of lfm_value */
846 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
847 __u32 lfm_flags; /* flags, type specific */
849 } __attribute__((packed));
851 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
852 offsetof(struct lov_foreign_md, lfm_value))
854 #define foreign_size_le(lfm) \
855 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
856 offsetof(struct lov_foreign_md, lfm_value))
859 * The stripe size fields are shared for the extension size storage, however
860 * the extension size is stored in KB, not bytes.
862 #define SEL_UNIT_SIZE 1024llu
867 } __attribute__((packed));
869 #define DEXT "[%#llx, %#llx)"
870 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
872 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
873 struct lu_extent *e2)
875 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
878 static inline bool lu_extent_is_whole(struct lu_extent *e)
880 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
883 enum lov_comp_md_entry_flags {
884 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
885 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
886 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
887 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
888 LCME_FL_OFFLINE = 0x00000008, /* Not used */
889 LCME_FL_INIT = 0x00000010, /* instantiated */
890 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
891 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
892 LCME_FL_PARITY = 0x00000080, /* EC: a parity code component */
893 LCME_FL_COMPRESS = 0x00000100, /* the component should be compressed */
894 LCME_FL_PARTIAL = 0x00000200, /* some chunks in the component are
897 LCME_FL_NOCOMPR = 0x00000400, /* the component should not be
900 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
901 * won't be stored on disk
905 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
906 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
909 /* The component flags can be set by users at creation/modification time. */
910 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
913 /* The mirror flags can be set by users at creation time. */
914 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
916 /* The allowed flags obtained from the client at component creation time. */
917 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
919 /* The mirror flags sent by client */
920 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
922 /* These flags have meaning when set in a default layout and will be inherited
923 * from the default/template layout set on a directory.
925 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
928 /* lcme_id can be specified as certain flags, and the the first
929 * bit of lcme_id is used to indicate that the ID is representing
930 * certain LCME_FL_* but not a real ID. Which implies we can have
931 * at most 31 flags (see LCME_FL_XXX). */
934 LCME_ID_MAX = 0x7FFFFFFF,
935 LCME_ID_ALL = 0xFFFFFFFF,
936 LCME_ID_NOT_ID = LCME_FL_NEG
939 /* layout version equals to lcme_id, except some bits have special meanings */
940 enum layout_version_flags {
941 /* layout version reaches the high water mark to be increased to
942 * circularly reuse the smallest value
944 LU_LAYOUT_HIGEN = 0x40000000,
945 /* the highest bit is used to mark if the file is being resynced */
946 LU_LAYOUT_RESYNC = 0x80000000,
949 #define LCME_ID_MASK LCME_ID_MAX
951 struct lov_comp_md_entry_v1 {
952 __u32 lcme_id; /* unique id of component */
953 __u32 lcme_flags; /* LCME_FL_XXX */
954 /* file extent for component. If it's an EC code component, its flags
955 * contains LCME_FL_PARITY, and its extent covers the same extent of
956 * its corresponding data component.
958 struct lu_extent lcme_extent;
959 __u32 lcme_offset; /* offset of component blob,
960 start from lov_comp_md_v1 */
961 __u32 lcme_size; /* size of component blob */
962 __u32 lcme_layout_gen;
963 __u64 lcme_timestamp; /* snapshot time if applicable*/
964 __u8 lcme_dstripe_count; /* data stripe count,
967 __u8 lcme_cstripe_count; /* code stripe count,
970 __u8 lcme_compr_type; /* compress type */
971 __u8 lcme_compr_lvl:4; /* compress level */
972 __u8 lcme_compr_chunk_log_bits:4;
973 /* chunk_size = 2^(16+chunk_log_bits)
974 * i.e. power-of-two multiple of 64KiB */
975 } __attribute__((packed));
977 #define SEQ_ID_MAX 0x0000FFFF
978 #define SEQ_ID_MASK SEQ_ID_MAX
979 /* bit 30:16 of lcme_id is used to store mirror id */
980 #define MIRROR_ID_MASK 0x7FFF0000
981 #define MIRROR_ID_NEG 0x8000
982 #define MIRROR_ID_SHIFT 16
984 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
986 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
989 static inline __u16 mirror_id_of(__u32 id)
991 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
995 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
997 enum lov_comp_md_flags {
998 /* the least 4 bits are used by FLR to record file state */
1000 LCM_FL_RDONLY = 0x1,
1001 LCM_FL_WRITE_PENDING = 0x2,
1002 LCM_FL_SYNC_PENDING = 0x3,
1003 LCM_FL_PCC_RDONLY = 0x8,
1004 LCM_FL_FLR_MASK = 0xB,
1007 struct lov_comp_md_v1 {
1008 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
1009 __u32 lcm_size; /* overall size including this struct */
1010 __u32 lcm_layout_gen;
1012 __u16 lcm_entry_count;
1013 /* lcm_mirror_count stores the number of actual mirrors minus 1,
1014 * so that non-flr files will have value 0 meaning 1 mirror. */
1015 __u16 lcm_mirror_count;
1016 /* code components count, non-EC file contains 0 ec_count */
1018 __u8 lcm_padding3[1];
1019 __u16 lcm_padding1[2];
1021 struct lov_comp_md_entry_v1 lcm_entries[0];
1022 } __attribute__((packed));
1024 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
1026 if (stripes == (__u16)-1)
1029 if (lmm_magic == LOV_USER_MAGIC_V1)
1030 return sizeof(struct lov_user_md_v1) +
1031 stripes * sizeof(struct lov_user_ost_data_v1);
1032 return sizeof(struct lov_user_md_v3) +
1033 stripes * sizeof(struct lov_user_ost_data_v1);
1036 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
1037 * use this. It is unsafe to #define those values in this header as it
1038 * is possible the application has already #included <sys/stat.h>. */
1039 #define lov_user_mds_data lov_user_mds_data_v2
1040 struct lov_user_mds_data_v1 {
1041 lstat_t lmd_st; /* MDS stat struct */
1042 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
1043 } __attribute__((packed));
1045 struct lov_user_mds_data_v2 {
1046 struct lu_fid lmd_fid; /* Lustre FID */
1047 lstatx_t lmd_stx; /* MDS statx struct */
1048 __u64 lmd_flags; /* MDS stat flags */
1049 __u32 lmd_lmmsize; /* LOV EA size */
1050 __u32 lmd_padding; /* unused */
1051 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
1052 } __attribute__((packed));
1054 struct lmv_user_mds_data {
1055 struct lu_fid lum_fid;
1058 } __attribute__((packed, __may_alias__));
1060 enum lmv_hash_type {
1061 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
1062 LMV_HASH_TYPE_ALL_CHARS = 1, /* simple sum of characters */
1063 LMV_HASH_TYPE_FNV_1A_64 = 2, /* reasonable non-cryptographic hash */
1064 LMV_HASH_TYPE_CRUSH = 3, /* double-hash to optimize migration */
1065 LMV_HASH_TYPE_CRUSH2 = 4, /* CRUSH with small fixes, LU-15692 */
1067 LMV_HASH_TYPE_DEFAULT = LMV_HASH_TYPE_FNV_1A_64
1070 static __attribute__((unused)) const char *mdt_hash_name[] = {
1079 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1080 * and the higher part will be the flag to indicate the status of object,
1081 * for example the object is being migrated. And the hash function
1082 * might be interpreted differently with different flags. */
1083 #define LMV_HASH_TYPE_MASK 0x0000ffff
1085 static inline bool lmv_is_known_hash_type(__u32 type)
1087 return (type & LMV_HASH_TYPE_MASK) > LMV_HASH_TYPE_UNKNOWN &&
1088 (type & LMV_HASH_TYPE_MASK) < LMV_HASH_TYPE_MAX;
1091 /* This flag indicates that overstriping (>1 stripe per MDT) is desired */
1092 #define LMV_HASH_FLAG_OVERSTRIPED 0x01000000
1093 /* fixed layout, such directories won't split automatically */
1094 /* NB, update LMV_HASH_FLAG_KNOWN when adding new flag */
1095 #define LMV_HASH_FLAG_FIXED 0x02000000
1096 #define LMV_HASH_FLAG_MERGE 0x04000000
1097 #define LMV_HASH_FLAG_SPLIT 0x08000000
1099 /* The striped directory has ever lost its master LMV EA, then LFSCK
1100 * re-generated it. This flag is used to indicate such case. It is an
1102 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
1104 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1105 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1107 #define LMV_HASH_FLAG_LAYOUT_CHANGE \
1108 (LMV_HASH_FLAG_MIGRATION | LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MERGE)
1110 #define LMV_HASH_FLAG_KNOWN 0xbe000000
1112 /* migration failure may leave hash type as
1113 * LMV_HASH_TYPE_UNKNOWN|LMV_HASH_FLAG_BAD_TYPE, which should be treated as
1114 * sane, so such directory can be accessed (resume migration or unlink).
1116 static inline bool lmv_is_sane_hash_type(__u32 type)
1118 return lmv_is_known_hash_type(type) ||
1119 type == (LMV_HASH_TYPE_UNKNOWN | LMV_HASH_FLAG_BAD_TYPE);
1122 /* both SPLIT and MIGRATION are set for directory split */
1123 static inline bool lmv_hash_is_splitting(__u32 hash)
1125 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1126 (LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MIGRATION);
1129 /* both MERGE and MIGRATION are set for directory merge */
1130 static inline bool lmv_hash_is_merging(__u32 hash)
1132 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1133 (LMV_HASH_FLAG_MERGE | LMV_HASH_FLAG_MIGRATION);
1136 /* only MIGRATION is set for directory migration */
1137 static inline bool lmv_hash_is_migrating(__u32 hash)
1139 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) == LMV_HASH_FLAG_MIGRATION;
1142 static inline bool lmv_hash_is_restriping(__u32 hash)
1144 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash);
1147 static inline bool lmv_hash_is_layout_changing(__u32 hash)
1149 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash) ||
1150 lmv_hash_is_migrating(hash);
1153 struct lustre_foreign_type {
1155 const char *lft_name;
1159 * LOV/LMV foreign types
1161 enum lustre_foreign_types {
1162 LU_FOREIGN_TYPE_NONE = 0,
1163 /* HSM copytool lhsm_posix */
1164 LU_FOREIGN_TYPE_POSIX = 1,
1165 /* Used for PCC-RW. PCCRW components are local to a single archive. */
1166 LU_FOREIGN_TYPE_PCCRW = 2,
1167 /* Used for PCC-RO. PCCRO components may be shared between archives. */
1168 LU_FOREIGN_TYPE_PCCRO = 3,
1170 LU_FOREIGN_TYPE_S3 = 4,
1172 LU_FOREIGN_TYPE_SYMLINK = 0xda05,
1173 /* must be the max/last one */
1174 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
1177 extern struct lustre_foreign_type lu_foreign_types[];
1179 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
1180 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
1181 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
1182 #define lmv_user_md lmv_user_md_v1
1183 struct lmv_user_md_v1 {
1184 __u32 lum_magic; /* must be the first field */
1185 __u32 lum_stripe_count; /* dirstripe count */
1186 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
1187 __u32 lum_hash_type; /* Dir stripe policy */
1188 __u32 lum_type; /* LMV type: default */
1189 __u8 lum_max_inherit; /* inherit depth of default LMV */
1190 __u8 lum_max_inherit_rr; /* inherit depth of default LMV to round-robin mkdir */
1194 char lum_pool_name[LOV_MAXPOOLNAME + 1];
1195 struct lmv_user_mds_data lum_objects[0];
1196 } __attribute__((packed));
1198 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
1200 if (size <= sizeof(struct lmv_user_md))
1203 size -= sizeof(struct lmv_user_md);
1204 return (size + sizeof(struct lmv_user_mds_data) - 1) /
1205 sizeof(struct lmv_user_mds_data);
1209 * NB, historically default layout didn't set type, but use XATTR name to differ
1210 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
1211 * and still use the same method.
1214 LMV_TYPE_DEFAULT = 0x0000,
1215 /* fetch raw default LMV set on directory inode */
1216 LMV_TYPE_RAW = 0x0001,
1219 /* lum_max_inherit will be decreased by 1 after each inheritance if it's not
1220 * LMV_INHERIT_UNLIMITED or > LMV_INHERIT_MAX.
1223 /* for historical reason, 0 means unlimited inheritance */
1224 LMV_INHERIT_UNLIMITED = 0,
1225 /* unlimited lum_max_inherit by default for plain stripe (0 or 1) */
1226 LMV_INHERIT_DEFAULT_PLAIN = LMV_INHERIT_UNLIMITED,
1227 /* not inherit any more */
1228 LMV_INHERIT_END = 1,
1229 /* for multiple stripes, the default lum_max_inherit is 3 */
1230 LMV_INHERIT_DEFAULT_STRIPED = 3,
1231 /* max inherit depth */
1232 LMV_INHERIT_MAX = 250,
1233 /* [251, 254] are reserved */
1234 /* not set, or when inherit depth goes beyond end, */
1235 LMV_INHERIT_NONE = 255,
1239 /* not set, or when inherit_rr depth goes beyond end, */
1240 LMV_INHERIT_RR_NONE = 0,
1241 /* disable lum_max_inherit_rr by default */
1242 LMV_INHERIT_RR_DEFAULT = LMV_INHERIT_RR_NONE,
1243 /* not inherit any more */
1244 LMV_INHERIT_RR_END = 1,
1245 /* default inherit_rr of ROOT */
1246 LMV_INHERIT_RR_ROOT = 3,
1247 /* max inherit depth */
1248 LMV_INHERIT_RR_MAX = 250,
1249 /* [251, 254] are reserved */
1250 /* unlimited inheritance */
1251 LMV_INHERIT_RR_UNLIMITED = 255,
1254 static inline int lmv_user_md_size(int stripes, int lmm_magic)
1256 int size = sizeof(struct lmv_user_md);
1258 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
1259 size += stripes * sizeof(struct lmv_user_mds_data);
1264 struct ll_recreate_obj {
1270 __u64 id; /* holds object id */
1271 __u32 generation; /* holds object generation */
1272 __u32 f_type; /* holds object type or stripe idx when passing it to
1273 * OST for saving into EA. */
1278 char uuid[UUID_MAX];
1281 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
1282 const struct obd_uuid *u2)
1284 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
1287 static inline int obd_uuid_empty(struct obd_uuid *uuid)
1289 return uuid->uuid[0] == '\0';
1292 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
1294 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
1295 uuid->uuid[sizeof(*uuid) - 1] = '\0';
1298 /* For printf's only, make sure uuid is terminated */
1299 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
1304 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
1305 /* Obviously not safe, but for printfs, no real harm done...
1306 we're always null-terminated, even in a race. */
1307 static char temp[sizeof(*uuid->uuid)];
1309 memcpy(temp, uuid->uuid, sizeof(*uuid->uuid) - 1);
1310 temp[sizeof(*uuid->uuid) - 1] = '\0';
1314 return (char *)(uuid->uuid);
1317 #define LUSTRE_MAXFSNAME 8
1318 #define LUSTRE_MAXINSTANCE 16
1320 /* Extract fsname from uuid (or target name) of a target
1321 e.g. (myfs-OST0007_UUID -> myfs)
1322 see also deuuidify. */
1323 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1327 strncpy(buf, uuid, buflen - 1);
1328 buf[buflen - 1] = '\0';
1329 p = strrchr(buf, '-');
1334 /* printf display format for Lustre FIDs
1335 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1336 #define FID_NOBRACE_LEN 40
1337 #define FID_LEN (FID_NOBRACE_LEN + 2)
1338 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1339 #define DFID "[" DFID_NOBRACE "]"
1340 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1342 /* scanf input parse format for fids in DFID_NOBRACE format
1343 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1344 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1345 #define SFID "0x%llx:0x%x:0x%x"
1346 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1347 #define PLOGID(logid) (unsigned long long)(logid)->lgl_oi.oi.oi_seq, (__u32)(logid)->lgl_oi.oi.oi_id, 0
1349 /********* Quotas **********/
1351 /* From linux/fs/quota/quota.c */
1352 static inline __u64 toqb(__kernel_size_t space)
1354 return (space + QIF_DQBLKSIZE - 1) >> QIF_DQBLKSIZE_BITS;
1357 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1358 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1359 #define Q_GETOINFO 0x800102 /* get obd quota info */
1360 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1361 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1363 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1364 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1365 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1366 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1367 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1368 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1369 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1370 /* lustre-specific control commands */
1371 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1372 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1373 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1374 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1375 #define LUSTRE_Q_GETQUOTAPOOL 0x80000f /* get user pool quota */
1376 #define LUSTRE_Q_SETQUOTAPOOL 0x800010 /* set user pool quota */
1377 #define LUSTRE_Q_GETINFOPOOL 0x800011 /* get pool quota info */
1378 #define LUSTRE_Q_SETINFOPOOL 0x800012 /* set pool quota info */
1379 #define LUSTRE_Q_GETDEFAULT_POOL 0x800013 /* get default pool quota*/
1380 #define LUSTRE_Q_SETDEFAULT_POOL 0x800014 /* set default pool quota */
1381 #define LUSTRE_Q_DELETEQID 0x800015 /* delete quota ID */
1382 #define LUSTRE_Q_RESETQID 0x800016 /* reset quota ID */
1383 /* In the current Lustre implementation, the grace time is either the time
1384 * or the timestamp to be used after some quota ID exceeds the soft limt,
1385 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1387 #define LQUOTA_GRACE_BITS 48
1388 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1389 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1390 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1391 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1392 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1394 /* special grace time, only notify the user when its quota is over soft limit
1395 * but doesn't block new writes until the hard limit is reached. */
1396 #define NOTIFY_GRACE "notify"
1397 #define NOTIFY_GRACE_TIME LQUOTA_GRACE_MASK
1399 /* different quota flags */
1401 /* the default quota flag, the corresponding quota ID will use the default
1402 * quota setting, the hardlimit and softlimit of its quota record in the global
1403 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1404 * and high 16 bits will contain this flag (see above comment).
1406 #define LQUOTA_FLAG_DEFAULT 0x0001
1407 #define LQUOTA_FLAG_DELETED 0x0002
1408 #define LQUOTA_FLAG_RESET 0x0004
1410 #define LUSTRE_Q_CMD_IS_POOL(cmd) \
1411 (cmd == LUSTRE_Q_GETQUOTAPOOL || \
1412 cmd == LUSTRE_Q_SETQUOTAPOOL || \
1413 cmd == LUSTRE_Q_SETINFOPOOL || \
1414 cmd == LUSTRE_Q_GETINFOPOOL || \
1415 cmd == LUSTRE_Q_SETDEFAULT_POOL || \
1416 cmd == LUSTRE_Q_GETDEFAULT_POOL)
1418 #define ALLQUOTA 255 /* set all quota */
1419 static inline const char *qtype_name(int qtype)
1432 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1435 #define N_PERMS_MAX 64
1437 struct perm_downcall_data {
1443 struct identity_downcall_data {
1450 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1451 __u32 idd_groups[0];
1454 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
1455 /* old interface struct is deprecated in 2.14 */
1456 #define SEPOL_DOWNCALL_MAGIC_OLD 0x8b8bb842
1457 struct sepol_downcall_data_old {
1459 __s64 sdd_sepol_mtime;
1460 __u16 sdd_sepol_len;
1465 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb843
1466 struct sepol_downcall_data {
1468 __u16 sdd_sepol_len;
1470 __s64 sdd_sepol_mtime;
1474 #ifdef NEED_QUOTA_DEFS
1476 #define QIF_BLIMITS 1
1478 #define QIF_ILIMITS 4
1479 #define QIF_INODES 8
1480 #define QIF_BTIME 16
1481 #define QIF_ITIME 32
1482 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1483 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1484 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1485 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1488 #endif /* !__KERNEL__ */
1490 /* lustre volatile file support
1491 * file name header: ".^L^S^T^R:volatile"
1493 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1494 #define LUSTRE_VOLATILE_HDR_LEN 14
1496 enum lustre_quota_version {
1500 /* XXX: same as if_dqinfo struct in kernel */
1508 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1510 __u64 dqb_bhardlimit; /* kbytes unit */
1511 __u64 dqb_bsoftlimit; /* kbytes unit */
1512 __u64 dqb_curspace; /* bytes unit */
1513 __u64 dqb_ihardlimit;
1514 __u64 dqb_isoftlimit;
1515 __u64 dqb_curinodes;
1529 struct if_quotactl {
1536 struct obd_dqinfo qc_dqinfo;
1537 struct obd_dqblk qc_dqblk;
1539 struct obd_uuid obd_uuid;
1543 /* swap layout flags */
1544 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1545 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1546 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1547 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1548 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1550 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1551 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1552 struct lustre_swap_layouts {
1560 /** Bit-mask of valid attributes */
1561 /* The LA_* flags are written to disk as part of the ChangeLog records
1562 * so they are part of the on-disk and network protocol, and cannot be changed.
1563 * Only the first 12 bits are currently saved.
1566 LA_ATIME = 1 << 0, /* 0x00001 */
1567 LA_MTIME = 1 << 1, /* 0x00002 */
1568 LA_CTIME = 1 << 2, /* 0x00004 */
1569 LA_SIZE = 1 << 3, /* 0x00008 */
1570 LA_MODE = 1 << 4, /* 0x00010 */
1571 LA_UID = 1 << 5, /* 0x00020 */
1572 LA_GID = 1 << 6, /* 0x00040 */
1573 LA_BLOCKS = 1 << 7, /* 0x00080 */
1574 LA_TYPE = 1 << 8, /* 0x00100 */
1575 LA_FLAGS = 1 << 9, /* 0x00200 */
1576 LA_NLINK = 1 << 10, /* 0x00400 */
1577 LA_RDEV = 1 << 11, /* 0x00800 */
1578 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1579 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1580 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1581 LA_PROJID = 1 << 15, /* 0x08000 */
1582 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1583 LA_LSIZE = 1 << 17, /* 0x20000 */
1584 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1585 LA_BTIME = 1 << 19, /* 0x80000 */
1587 * Attributes must be transmitted to OST objects
1589 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1592 #define MDS_FMODE_READ 00000001
1593 #define MDS_FMODE_WRITE 00000002
1595 #define MDS_FMODE_CLOSED 00000000
1596 #define MDS_FMODE_EXEC 00000004
1597 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1598 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1599 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1601 #define MDS_OPEN_CREATED 00000010
1602 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1604 #define MDS_OPEN_CREAT 00000100
1605 #define MDS_OPEN_EXCL 00000200
1606 #define MDS_OPEN_TRUNC 00001000
1607 #define MDS_OPEN_APPEND 00002000
1608 #define MDS_OPEN_SYNC 00010000
1609 #define MDS_OPEN_DIRECTORY 00200000
1611 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1612 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1613 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1614 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1615 * We do not support JOIN FILE
1616 * anymore, reserve this flags
1617 * just for preventing such bit
1620 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1621 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1622 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1623 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1624 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1626 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1628 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1629 * delegation, succeed if it's not
1630 * being opened with conflict mode.
1632 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1634 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1635 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1636 * for newly created file */
1637 #define MDS_OP_WITH_FID 020000000000000ULL /* operation carried out by FID */
1638 #define MDS_OPEN_DEFAULT_LMV 040000000000000ULL /* open fetches default LMV,
1639 * or mkdir with default LMV
1642 /* lustre internal open flags, which should not be set from user space */
1643 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1644 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1645 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1646 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1647 MDS_OPEN_PCC | MDS_OP_WITH_FID | \
1648 MDS_OPEN_DEFAULT_LMV)
1651 /********* Changelogs **********/
1652 /** Changelog record types */
1653 enum changelog_rec_type {
1656 CL_CREATE = 1, /* namespace */
1657 CL_MKDIR = 2, /* namespace */
1658 CL_HARDLINK = 3, /* namespace */
1659 CL_SOFTLINK = 4, /* namespace */
1660 CL_MKNOD = 5, /* namespace */
1661 CL_UNLINK = 6, /* namespace */
1662 CL_RMDIR = 7, /* namespace */
1663 CL_RENAME = 8, /* namespace */
1664 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1665 CL_OPEN = 10, /* not currently used */
1666 CL_CLOSE = 11, /* may be written to log only with mtime change */
1667 CL_LAYOUT = 12, /* file layout/striping modified */
1671 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1672 CL_HSM = 16, /* HSM specific events, see flags */
1673 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1677 CL_FLRW = 21, /* FLR: file was firstly written */
1678 CL_RESYNC = 22, /* FLR: file was resync-ed */
1680 CL_DN_OPEN = 24, /* denied open */
1684 static inline const char *changelog_type2str(int type) {
1685 static const char *const changelog_str[] = {
1686 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1687 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1688 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1689 "FLRW", "RESYNC","GXATR", "NOPEN",
1692 if (type >= 0 && type < CL_LAST)
1693 return changelog_str[type];
1697 /* 12 bits of per-record data can be stored in the bottom of the flags */
1698 #define CLF_FLAGSHIFT 12
1699 enum changelog_rec_flags {
1700 CLF_VERSION = 0x1000,
1701 CLF_RENAME = 0x2000,
1703 CLF_EXTRA_FLAGS = 0x8000,
1704 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1706 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1707 CLF_VERMASK = ~CLF_FLAGMASK,
1711 /* Anything under the flagmask may be per-type (if desired) */
1712 /* Flags for unlink */
1713 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1714 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1715 /* HSM cleaning needed */
1716 /* Flags for rename */
1717 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1719 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1720 * has an archive in backend */
1723 /* 12b used (from high weight to low weight):
1728 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1729 #define CLF_HSM_ERR_H 6
1730 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1731 #define CLF_HSM_EVENT_H 9
1732 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1733 #define CLF_HSM_FLAG_H 11
1734 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1735 #define CLF_HSM_SPARE_H 15
1736 #define CLF_HSM_LAST 15
1738 /* Remove bits higher than _h, then extract the value
1739 * between _h and _l by shifting lower weigth to bit 0. */
1740 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1741 >> (CLF_HSM_LAST - _h + _l))
1743 #define CLF_HSM_SUCCESS 0x00
1744 #define CLF_HSM_MAXERROR 0x7E
1745 #define CLF_HSM_ERROVERFLOW 0x7F
1747 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1749 /* 3 bits field => 8 values allowed */
1761 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1763 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1767 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1770 *clf_flags = (enum changelog_rec_flags)
1771 (*clf_flags | (he << CLF_HSM_EVENT_L));
1774 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1776 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1779 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1782 *clf_flags = (enum changelog_rec_flags)
1783 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1786 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1788 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1791 static inline int hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1794 /* In case a negative error is given */
1797 if (error > CLF_HSM_MAXERROR)
1798 error = CLF_HSM_ERROVERFLOW;
1800 *clf_flags = (enum changelog_rec_flags)
1801 (*clf_flags | (error << CLF_HSM_ERR_L));
1803 return error == CLF_HSM_ERROVERFLOW ? -EOVERFLOW : 0;
1806 enum changelog_rec_extra_flags {
1808 CLFE_UIDGID = 0x0001,
1811 CLFE_XATTR = 0x0008,
1812 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1815 enum changelog_send_flag {
1816 /* Use changelog follow mode: llapi_changelog_recv() will not stop at
1817 * the end of records and wait for new records to be generated.
1819 CHANGELOG_FLAG_FOLLOW = 0x01,
1820 /* Deprecated since Lustre 2.10 */
1821 CHANGELOG_FLAG_BLOCK = 0x02,
1822 /* Pack jobid into the changelog records if available. */
1823 CHANGELOG_FLAG_JOBID = 0x04,
1824 /* Pack additional flag bits into the changelog record */
1825 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1828 enum changelog_send_extra_flag {
1829 /* Pack uid/gid into the changelog record */
1830 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1831 /* Pack nid into the changelog record */
1832 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1833 /* Pack open mode into the changelog record */
1834 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1835 /* Pack xattr name into the changelog record */
1836 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1839 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1840 changelog_rec_offset(CLF_SUPPORTED, \
1843 /* 31 usable bytes string + null terminator. */
1844 #define LUSTRE_JOBID_SIZE 32
1846 /* This is the minimal changelog record. It can contain extensions
1847 * such as rename fields or process jobid. Its exact content is described
1848 * by the cr_flags and cr_extra_flags.
1850 * Extensions are packed in the same order as their corresponding flags,
1851 * then in the same order as their corresponding extra flags.
1853 struct changelog_rec {
1855 __u16 cr_flags; /**< \a changelog_rec_flags */
1856 __u32 cr_type; /**< \a changelog_rec_type */
1857 __u64 cr_index; /**< changelog record number */
1858 __u64 cr_prev; /**< last index for this target fid */
1861 struct lu_fid cr_tfid; /**< target fid */
1862 __u32 cr_markerflags; /**< CL_MARK flags */
1864 struct lu_fid cr_pfid; /**< parent fid */
1865 } __attribute__ ((packed));
1867 /* Changelog extension for RENAME. */
1868 struct changelog_ext_rename {
1869 struct lu_fid cr_sfid; /**< source fid, or zero */
1870 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1873 /* Changelog extension to include JOBID. */
1874 struct changelog_ext_jobid {
1875 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1878 /* Changelog extension to include additional flags. */
1879 struct changelog_ext_extra_flags {
1880 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1883 /* Changelog extra extension to include UID/GID. */
1884 struct changelog_ext_uidgid {
1889 /* Changelog extra extension to include NID. */
1890 struct changelog_ext_nid {
1891 /* have __u64 instead of lnet_nid_t type for use by client api */
1893 /* for use when IPv6 support is added */
1898 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1899 struct changelog_ext_openmode {
1903 /* Changelog extra extension to include xattr */
1904 struct changelog_ext_xattr {
1905 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1908 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1909 const struct changelog_rec *rec);
1911 static inline __kernel_size_t changelog_rec_offset(enum changelog_rec_flags crf,
1912 enum changelog_rec_extra_flags cref)
1914 __kernel_size_t size = sizeof(struct changelog_rec);
1916 if (crf & CLF_RENAME)
1917 size += sizeof(struct changelog_ext_rename);
1919 if (crf & CLF_JOBID)
1920 size += sizeof(struct changelog_ext_jobid);
1922 if (crf & CLF_EXTRA_FLAGS) {
1923 size += sizeof(struct changelog_ext_extra_flags);
1924 if (cref & CLFE_UIDGID)
1925 size += sizeof(struct changelog_ext_uidgid);
1926 if (cref & CLFE_NID)
1927 size += sizeof(struct changelog_ext_nid);
1928 if (cref & CLFE_OPEN)
1929 size += sizeof(struct changelog_ext_openmode);
1930 if (cref & CLFE_XATTR)
1931 size += sizeof(struct changelog_ext_xattr);
1937 static inline __kernel_size_t changelog_rec_size(const struct changelog_rec *rec)
1939 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1941 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1942 cref = (enum changelog_rec_extra_flags)
1943 changelog_rec_extra_flags(rec)->cr_extra_flags;
1945 return changelog_rec_offset(
1946 (enum changelog_rec_flags)rec->cr_flags, cref);
1949 static inline __kernel_size_t changelog_rec_varsize(const struct changelog_rec *rec)
1951 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1955 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1957 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1958 (rec->cr_flags & CLF_VERSION);
1960 return (struct changelog_ext_rename *)((char *)rec +
1961 changelog_rec_offset(crf,
1965 /* The jobid follows the rename extension, if present */
1967 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1969 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1970 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1972 return (struct changelog_ext_jobid *)((char *)rec +
1973 changelog_rec_offset(crf,
1977 /* The additional flags follow the rename and jobid extensions, if present */
1979 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1980 const struct changelog_rec *rec)
1982 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1983 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1985 return (struct changelog_ext_extra_flags *)((char *)rec +
1986 changelog_rec_offset(crf,
1990 /* The uid/gid is the first extra extension */
1992 struct changelog_ext_uidgid *changelog_rec_uidgid(
1993 const struct changelog_rec *rec)
1995 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1997 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1999 return (struct changelog_ext_uidgid *)((char *)rec +
2000 changelog_rec_offset(crf,
2004 /* The nid is the second extra extension */
2006 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
2008 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2010 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2011 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2013 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2014 cref = (enum changelog_rec_extra_flags)
2015 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2018 return (struct changelog_ext_nid *)((char *)rec +
2019 changelog_rec_offset(crf, cref));
2022 /* The OPEN mode is the third extra extension */
2024 struct changelog_ext_openmode *changelog_rec_openmode(
2025 const struct changelog_rec *rec)
2027 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2029 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2030 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2032 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
2033 cref = (enum changelog_rec_extra_flags)
2034 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2035 (CLFE_UIDGID | CLFE_NID));
2038 return (struct changelog_ext_openmode *)((char *)rec +
2039 changelog_rec_offset(crf, cref));
2042 /* The xattr name is the fourth extra extension */
2044 struct changelog_ext_xattr *changelog_rec_xattr(
2045 const struct changelog_rec *rec)
2047 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2049 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2050 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2052 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2053 cref = (enum changelog_rec_extra_flags)
2054 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2055 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
2057 return (struct changelog_ext_xattr *)((char *)rec +
2058 changelog_rec_offset(crf, cref));
2061 /* The name follows the rename, jobid and extra flags extns, if present */
2062 static inline char *changelog_rec_name(const struct changelog_rec *rec)
2064 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2066 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2067 cref = (enum changelog_rec_extra_flags)
2068 changelog_rec_extra_flags(rec)->cr_extra_flags;
2070 return (char *)rec + changelog_rec_offset(
2071 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
2072 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
2075 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
2077 char *str = changelog_rec_name(rec);
2079 while (*str != '\0')
2084 static inline __kernel_size_t changelog_rec_snamelen(const struct changelog_rec *rec)
2086 return strlen(changelog_rec_sname(rec));
2090 * Remap a record to the desired format as specified by the crf flags.
2091 * The record must be big enough to contain the final remapped version.
2092 * Superfluous extension fields are removed and missing ones are added
2093 * and zeroed. The flags of the record are updated accordingly.
2095 * The jobid and rename extensions can be added to a record, to match the
2096 * format an application expects, typically. In this case, the newly added
2097 * fields will be zeroed.
2098 * The Jobid field can be removed, to guarantee compatibility with older
2099 * clients that don't expect this field in the records they process.
2101 * The following assumptions are being made:
2102 * - CLF_RENAME will not be removed
2103 * - CLF_JOBID will not be added without CLF_RENAME being added too
2104 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
2106 * @param[in,out] rec The record to remap.
2107 * @param[in] crf_wanted Flags describing the desired extensions.
2108 * @param[in] cref_want Flags describing the desired extra extensions.
2110 static inline void changelog_remap_rec(struct changelog_rec *rec,
2111 enum changelog_rec_flags crf_wanted,
2112 enum changelog_rec_extra_flags cref_want)
2114 char *xattr_mov = NULL;
2115 char *omd_mov = NULL;
2116 char *nid_mov = NULL;
2117 char *uidgid_mov = NULL;
2121 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2123 crf_wanted = (enum changelog_rec_flags)
2124 (crf_wanted & CLF_SUPPORTED);
2125 cref_want = (enum changelog_rec_extra_flags)
2126 (cref_want & CLFE_SUPPORTED);
2128 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
2129 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
2130 (rec->cr_flags & CLF_EXTRA_FLAGS &&
2131 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2137 /* First move the variable-length name field */
2138 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
2139 changelog_rec_name(rec), rec->cr_namelen);
2141 /* Locations of extensions in the remapped record */
2142 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
2143 xattr_mov = (char *)rec +
2144 changelog_rec_offset(
2145 (enum changelog_rec_flags)
2146 (crf_wanted & CLF_SUPPORTED),
2147 (enum changelog_rec_extra_flags)
2148 (cref_want & ~CLFE_XATTR));
2149 omd_mov = (char *)rec +
2150 changelog_rec_offset(
2151 (enum changelog_rec_flags)
2152 (crf_wanted & CLF_SUPPORTED),
2153 (enum changelog_rec_extra_flags)
2154 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
2155 nid_mov = (char *)rec +
2156 changelog_rec_offset(
2157 (enum changelog_rec_flags)
2158 (crf_wanted & CLF_SUPPORTED),
2159 (enum changelog_rec_extra_flags)
2161 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
2162 uidgid_mov = (char *)rec +
2163 changelog_rec_offset(
2164 (enum changelog_rec_flags)
2165 (crf_wanted & CLF_SUPPORTED),
2166 (enum changelog_rec_extra_flags)
2167 (cref_want & ~(CLFE_UIDGID |
2171 cref = (enum changelog_rec_extra_flags)
2172 changelog_rec_extra_flags(rec)->cr_extra_flags;
2175 ef_mov = (char *)rec +
2176 changelog_rec_offset(
2177 (enum changelog_rec_flags)
2178 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
2179 jid_mov = (char *)rec +
2180 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2181 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
2183 rnm_mov = (char *)rec +
2184 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2190 /* Move the extension fields to the desired positions */
2191 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2192 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
2193 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
2194 memmove(xattr_mov, changelog_rec_xattr(rec),
2195 sizeof(struct changelog_ext_xattr));
2197 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
2198 memmove(omd_mov, changelog_rec_openmode(rec),
2199 sizeof(struct changelog_ext_openmode));
2201 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
2202 memmove(nid_mov, changelog_rec_nid(rec),
2203 sizeof(struct changelog_ext_nid));
2205 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
2206 memmove(uidgid_mov, changelog_rec_uidgid(rec),
2207 sizeof(struct changelog_ext_uidgid));
2209 memmove(ef_mov, changelog_rec_extra_flags(rec),
2210 sizeof(struct changelog_ext_extra_flags));
2213 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
2214 memmove(jid_mov, changelog_rec_jobid(rec),
2215 sizeof(struct changelog_ext_jobid));
2217 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
2218 memmove(rnm_mov, changelog_rec_rename(rec),
2219 sizeof(struct changelog_ext_rename));
2221 /* Clear newly added fields */
2222 if (xattr_mov && (cref_want & CLFE_XATTR) &&
2223 !(cref & CLFE_XATTR))
2224 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
2226 if (omd_mov && (cref_want & CLFE_OPEN) &&
2227 !(cref & CLFE_OPEN))
2228 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
2230 if (nid_mov && (cref_want & CLFE_NID) &&
2232 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
2234 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
2235 !(cref & CLFE_UIDGID))
2236 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
2238 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2239 !(rec->cr_flags & CLF_EXTRA_FLAGS))
2240 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
2242 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
2243 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
2245 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
2246 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
2248 /* Update the record's flags accordingly */
2249 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
2250 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2251 changelog_rec_extra_flags(rec)->cr_extra_flags =
2252 changelog_rec_extra_flags(rec)->cr_extra_flags |
2256 enum changelog_message_type {
2257 CL_RECORD = 10, /* message is a changelog_rec */
2258 CL_EOF = 11, /* at end of current changelog */
2261 /********* Misc **********/
2263 struct ioc_data_version {
2265 __u32 idv_layout_version; /* FLR: layout version for OST objects */
2266 __u32 idv_flags; /* enum ioc_data_version_flags */
2269 enum ioc_data_version_flags {
2270 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
2271 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
2272 LL_DV_SZ_UPDATE = (1 << 2), /* Update the file size on the client */
2276 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
2279 #define dot_lustre_name ".lustre"
2280 #define dot_fscrypt_name ".fscrypt"
2283 /********* HSM **********/
2285 /** HSM per-file state
2286 * See HSM_FLAGS below.
2289 HS_NONE = 0x00000000,
2290 HS_EXISTS = 0x00000001,
2291 HS_DIRTY = 0x00000002,
2292 HS_RELEASED = 0x00000004,
2293 HS_ARCHIVED = 0x00000008,
2294 HS_NORELEASE = 0x00000010,
2295 HS_NOARCHIVE = 0x00000020,
2296 HS_LOST = 0x00000040,
2297 HS_PCCRW = 0x00000080,
2298 HS_PCCRO = 0x00000100,
2301 /* HSM user-setable flags. */
2302 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
2304 /* Other HSM flags. */
2305 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
2308 * All HSM-related possible flags that could be applied to a file.
2309 * This should be kept in sync with hsm_states.
2311 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
2314 * HSM request progress state
2316 enum hsm_progress_states {
2323 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
2326 case HPS_WAITING: return "waiting";
2327 case HPS_RUNNING: return "running";
2328 case HPS_DONE: return "done";
2329 default: return "unknown";
2336 } __attribute__((packed));
2339 * Current HSM states of a Lustre file.
2341 * This structure purpose is to be sent to user-space mainly. It describes the
2342 * current HSM flags and in-progress action.
2344 struct hsm_user_state {
2345 /** Current HSM states, from enum hsm_states. */
2347 __u32 hus_archive_id;
2348 /** The current undergoing action, if there is one */
2349 __u32 hus_in_progress_state;
2350 __u32 hus_in_progress_action;
2351 struct hsm_extent hus_in_progress_location;
2352 char hus_extended_info[];
2355 struct hsm_state_set_ioc {
2356 struct lu_fid hssi_fid;
2358 __u64 hssi_clearmask;
2362 * This structure describes the current in-progress action for a file.
2363 * it is retuned to user space and send over the wire
2365 struct hsm_current_action {
2366 /** The current undergoing action, if there is one */
2367 /* state is one of hsm_progress_states */
2369 /* action is one of hsm_user_action */
2371 struct hsm_extent hca_location;
2374 /***** HSM user requests ******/
2375 /* User-generated (lfs/ioctl) request types */
2376 enum hsm_user_action {
2377 HUA_NONE = 1, /* no action (noop) */
2378 HUA_ARCHIVE = 10, /* copy to hsm */
2379 HUA_RESTORE = 11, /* prestage */
2380 HUA_RELEASE = 12, /* drop ost objects */
2381 HUA_REMOVE = 13, /* remove from archive */
2382 HUA_CANCEL = 14 /* cancel a request */
2385 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2388 case HUA_NONE: return "NOOP";
2389 case HUA_ARCHIVE: return "ARCHIVE";
2390 case HUA_RESTORE: return "RESTORE";
2391 case HUA_RELEASE: return "RELEASE";
2392 case HUA_REMOVE: return "REMOVE";
2393 case HUA_CANCEL: return "CANCEL";
2394 default: return "UNKNOWN";
2399 * List of hr_flags (bit field)
2401 #define HSM_FORCE_ACTION 0x0001
2402 /* used by CT, cannot be set by user */
2403 #define HSM_GHOST_COPY 0x0002
2406 * Contains all the fixed part of struct hsm_user_request.
2409 struct hsm_request {
2410 __u32 hr_action; /* enum hsm_user_action */
2411 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2412 __u64 hr_flags; /* request flags */
2413 __u32 hr_itemcount; /* item count in hur_user_item vector */
2417 struct hsm_user_item {
2418 struct lu_fid hui_fid;
2419 struct hsm_extent hui_extent;
2420 } __attribute__((packed));
2422 struct hsm_user_request {
2423 struct hsm_request hur_request;
2424 struct hsm_user_item hur_user_item[0];
2425 /* extra data blob at end of struct (after all
2426 * hur_user_items), only use helpers to access it
2428 } __attribute__((packed));
2430 /** Return pointer to data field in a hsm user request */
2431 static inline void *hur_data(struct hsm_user_request *hur)
2433 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2437 * Compute the current length of the provided hsm_user_request. This returns -1
2438 * instead of an errno because __kernel_ssize_t is defined to be only
2441 * return -1 on bounds check error.
2443 static inline __kernel_size_t hur_len(struct hsm_user_request *hur)
2447 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2448 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2449 (__u64)hur->hur_request.hr_itemcount *
2450 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2452 if ((__kernel_ssize_t)size < 0)
2458 /****** HSM RPCs to copytool *****/
2459 /* Message types the copytool may receive */
2460 enum hsm_message_type {
2461 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2464 /* Actions the copytool may be instructed to take for a given action_item */
2465 enum hsm_copytool_action {
2466 HSMA_NONE = 10, /* no action */
2467 HSMA_ARCHIVE = 20, /* arbitrary offset */
2473 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2476 case HSMA_NONE: return "NOOP";
2477 case HSMA_ARCHIVE: return "ARCHIVE";
2478 case HSMA_RESTORE: return "RESTORE";
2479 case HSMA_REMOVE: return "REMOVE";
2480 case HSMA_CANCEL: return "CANCEL";
2481 default: return "UNKNOWN";
2485 /* Copytool item action description */
2486 struct hsm_action_item {
2487 __u32 hai_len; /* valid size of this struct */
2488 __u32 hai_action; /* hsm_copytool_action, but use known size */
2489 struct lu_fid hai_fid; /* Lustre FID to operate on */
2490 struct lu_fid hai_dfid; /* fid used for data access */
2491 struct hsm_extent hai_extent; /* byte range to operate on */
2492 __u64 hai_cookie; /* action cookie from coordinator */
2493 __u64 hai_gid; /* grouplock id */
2494 char hai_data[0]; /* variable length */
2495 } __attribute__((packed));
2498 * helper function which print in hexa the first bytes of
2501 * \param hai [IN] record to print
2502 * \param buffer [IN,OUT] buffer to write the hex string to
2503 * \param len [IN] max buffer length
2507 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2508 char *buffer, __kernel_size_t len)
2515 data_len = hai->hai_len - sizeof(*hai);
2516 for (i = 0; (i < data_len) && (len > 2); i++) {
2517 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2528 /* Register even when the CDT is shutdown or being initialized */
2529 HAL_CDT_FORCE = 1 << 1,
2532 /* Copytool action list */
2533 #define HAL_VERSION 1
2534 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2535 struct hsm_action_list {
2537 __u32 hal_count; /* number of hai's to follow */
2538 __u64 hal_compound_id; /* returned by coordinator, ignored */
2540 __u32 hal_archive_id; /* which archive backend */
2542 char hal_fsname[0]; /* null-terminated */
2543 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2544 boundaries. See hai_zero */
2545 } __attribute__((packed));
2547 /* Return pointer to first hai in action list */
2548 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2550 __kernel_size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2552 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2555 /* Return pointer to next hai */
2556 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2558 __kernel_size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2560 return (struct hsm_action_item *)((char *)hai + offset);
2563 /* Return size of an hsm_action_list */
2564 static inline __kernel_size_t hal_size(struct hsm_action_list *hal)
2568 struct hsm_action_item *hai;
2570 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2571 hai = hai_first(hal);
2572 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2573 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2579 * describe the attributes to be set on imported file
2581 struct hsm_user_import {
2590 __u32 hui_archive_id;
2593 /* Copytool progress reporting */
2594 #define HP_FLAG_COMPLETED 0x01
2595 #define HP_FLAG_RETRY 0x02
2597 struct hsm_progress {
2598 struct lu_fid hp_fid;
2600 struct hsm_extent hp_extent;
2602 __u16 hp_errval; /* positive val */
2607 __u64 hc_data_version;
2609 __u16 hc_errval; /* positive val */
2611 struct hsm_action_item hc_hai;
2614 enum lu_ladvise_type {
2615 LU_LADVISE_INVALID = 0,
2616 LU_LADVISE_WILLREAD = 1,
2617 LU_LADVISE_DONTNEED = 2,
2618 LU_LADVISE_LOCKNOEXPAND = 3,
2619 LU_LADVISE_LOCKAHEAD = 4,
2620 /* Ahead operations for open|create|stat|read|write. */
2621 LU_LADVISE_AHEAD = 5,
2625 #define LU_LADVISE_NAMES { \
2626 [LU_LADVISE_WILLREAD] = "willread", \
2627 [LU_LADVISE_DONTNEED] = "dontneed", \
2628 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2629 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2630 [LU_LADVISE_AHEAD] = "ahead", \
2633 /* This is the userspace argument for ladvise. It is currently the same as
2634 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2635 * may need info which is only used locally. */
2636 struct llapi_lu_ladvise {
2637 __u16 lla_advice; /* advice type */
2638 __u16 lla_value1; /* values for different advice types */
2640 __u64 lla_start; /* first byte of extent for advice */
2641 __u64 lla_end; /* last byte of extent for advice */
2646 struct llapi_lu_ladvise2 {
2647 __u16 lla_advice; /* advice type */
2648 __u16 lla_value1; /* values for different advice types */
2659 char lla_buf[NAME_MAX + 1];
2663 /* I/O call sequences in a batch access. */
2664 enum lu_access_flags {
2665 ACCESS_FL_NONE = 0x0,
2666 ACCESS_FL_STAT = 0x01,
2667 ACCESS_FL_OPEN = 0x02,
2668 ACCESS_FL_CREAT = 0x04,
2669 ACCESS_FL_READ = 0x08,
2670 ACCESS_FL_WRITE = 0x10,
2671 ACCESS_FL_OC = ACCESS_FL_OPEN | ACCESS_FL_CREAT,
2672 ACCESS_FL_SOR = ACCESS_FL_STAT | ACCESS_FL_OPEN | ACCESS_FL_READ,
2673 ACCESS_FL_OCW = ACCESS_FL_OPEN | ACCESS_FL_CREAT | ACCESS_FL_WRITE,
2676 enum lu_ahead_mode {
2677 LU_AH_MODE_NONE = 0,
2679 * The batch access pattern obeys certain naming rules, such as mdtest
2680 * with the file naming format mdtest.$rank.$i.
2682 LU_AH_NAME_INDEX = 1,
2684 * Provide a file name list as input to do batch accesses with
2685 * irregular file name format.
2687 LU_AH_NAME_ARRAY = 2,
2688 /* Prefetching in readdir() order under a directory. */
2689 LU_AH_NAME_READDIR = 3,
2693 #define lla_ahead_mode lla_value1
2694 #define lla_access_flags lla_value2
2695 #define lla_batch_max lla_value3
2696 #define lla_fname lla_buf
2699 LF_ASYNC = 0x00000001,
2700 LF_UNSET = 0x00000002,
2703 #define LADVISE_MAGIC 0x1ADF1CE0
2704 /* Masks of valid flags for each advice */
2705 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2706 /* Flags valid for all advices not explicitly specified */
2707 #define LF_DEFAULT_MASK LF_ASYNC
2709 #define LF_MASK (LF_ASYNC | LF_UNSET)
2711 #define lla_lockahead_mode lla_value1
2712 #define lla_peradvice_flags lla_value2
2713 #define lla_lockahead_result lla_value3
2715 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2716 * is used on the wire. It is defined separately as we may need info which is
2717 * only used locally. */
2718 struct llapi_ladvise_hdr {
2719 __u32 lah_magic; /* LADVISE_MAGIC */
2720 __u32 lah_count; /* number of advices */
2721 __u64 lah_flags; /* from enum ladvise_flag */
2722 __u32 lah_value1; /* unused */
2723 __u32 lah_value2; /* unused */
2724 __u64 lah_value3; /* unused */
2725 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2728 #define LAH_COUNT_MAX (1024)
2733 SK_CRYPT_AES256_CTR = 1,
2734 SK_CRYPT_INVALID = __UINT16_MAX__
2741 SK_HMAC_INVALID = __UINT16_MAX__
2744 struct sk_crypt_type {
2745 const char *sct_name;
2749 struct sk_hmac_type {
2750 const char *sht_name;
2754 enum lock_mode_user {
2760 #define LOCK_MODE_NAMES { \
2761 [MODE_READ_USER] = "READ",\
2762 [MODE_WRITE_USER] = "WRITE"\
2765 enum lockahead_results {
2766 LLA_RESULT_SENT = 0,
2767 LLA_RESULT_DIFFERENT,
2771 enum lu_heat_flag_bit {
2772 LU_HEAT_FLAG_BIT_INVALID = 0,
2773 LU_HEAT_FLAG_BIT_OFF,
2774 LU_HEAT_FLAG_BIT_CLEAR,
2778 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2779 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2782 enum obd_heat_type {
2783 OBD_HEAT_READSAMPLE = 0,
2784 OBD_HEAT_WRITESAMPLE = 1,
2785 OBD_HEAT_READBYTE = 2,
2786 OBD_HEAT_WRITEBYTE = 3,
2790 #define LU_HEAT_NAMES { \
2791 [OBD_HEAT_READSAMPLE] = "readsample", \
2792 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2793 [OBD_HEAT_READBYTE] = "readbyte", \
2794 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2805 LU_PCC_READWRITE = 0x01,
2806 LU_PCC_READONLY = 0x02,
2807 LU_PCC_TYPE_MASK = LU_PCC_READWRITE | LU_PCC_READONLY,
2808 LU_PCC_FL_ASYNC = 0x10,
2812 static inline const char *pcc_type2string(enum lu_pcc_type type)
2814 switch (type & LU_PCC_TYPE_MASK) {
2817 case LU_PCC_READWRITE:
2819 case LU_PCC_READONLY:
2826 struct lu_pcc_attach {
2827 __u32 pcca_type; /* PCC type */
2828 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2831 enum lu_pcc_detach_opts {
2832 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2833 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2836 struct lu_pcc_detach_fid {
2837 /* fid of the file to detach */
2838 struct lu_fid pccd_fid;
2842 struct lu_pcc_detach {
2846 enum lu_pcc_state_flags {
2847 PCC_STATE_FL_NONE = 0x0,
2848 /* The inode attr is cached locally */
2849 PCC_STATE_FL_ATTR_VALID = 0x01,
2850 /* The file is being attached into PCC */
2851 PCC_STATE_FL_ATTACHING = 0x02,
2852 /* The PCC copy is unlinked */
2853 PCC_STATE_FL_UNLINKED = 0x04,
2856 struct lu_pcc_state {
2857 __u32 pccs_type; /* enum lu_pcc_type */
2858 __u32 pccs_open_count;
2859 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2861 char pccs_path[PATH_MAX];
2864 enum lu_project_type {
2865 LU_PROJECT_NONE = 0,
2872 __u32 project_type; /* enum lu_project_type */
2874 __u32 project_xflags;
2875 __u32 project_reserved;
2876 char project_name[NAME_MAX + 1];
2881 /* make header's size equal lu_fid */
2884 struct lu_fid fa_fids[0];
2886 #define OBD_MAX_FIDS_IN_ARRAY 4096
2888 /* more types could be defined upon need for more complex
2889 * format to be used in foreign symlink LOV/LMV EAs, like
2890 * one to describe a delimiter string and occurence number
2891 * of delimited sub-string, ...
2893 enum ll_foreign_symlink_upcall_item_type {
2899 /* may need to be modified to allow for more format items to be defined, and
2900 * like for ll_foreign_symlink_upcall_item_type enum
2902 struct ll_foreign_symlink_upcall_item {
2912 /* internal storage of constant string */
2914 /* upcall stores constant string in a raw */
2921 #define POSLEN_ITEM_SZ (offsetof(struct ll_foreign_symlink_upcall_item, len) + \
2922 sizeof(((struct ll_foreign_symlink_upcall_item *)0)->len))
2923 #define STRING_ITEM_SZ(sz) ( \
2924 offsetof(struct ll_foreign_symlink_upcall_item, bytestring) + \
2925 (sz + sizeof(__u32) - 1) / sizeof(__u32) * sizeof(__u32))
2927 /* presently limited to not cause max stack frame size to be reached
2928 * because of temporary automatic array of
2929 * "struct ll_foreign_symlink_upcall_item" presently used in
2930 * foreign_symlink_upcall_info_store()
2932 #define MAX_NB_UPCALL_ITEMS 32
2934 #if defined(__cplusplus)
2938 /** @} lustreuser */
2940 #endif /* _LUSTRE_USER_H */