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 /* combine exclusive patterns as a bad pattern */
725 #define LOV_PATTERN_BAD (LOV_PATTERN_RAID1 | LOV_PATTERN_MDT | \
728 #define LOV_PATTERN_F_MASK 0xffff0000
729 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
730 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
731 #define LOV_PATTERN_DEFAULT 0xffffffff
733 #define LOV_OFFSET_DEFAULT ((__u16)-1)
734 #define LMV_OFFSET_DEFAULT ((__u32)-1)
736 static inline bool lov_pattern_supported(__u32 pattern)
738 __u32 pattern_base = pattern & ~LOV_PATTERN_F_RELEASED;
740 return pattern_base == LOV_PATTERN_RAID0 ||
741 pattern_base == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
742 pattern_base == LOV_PATTERN_MDT;
745 /* RELEASED and MDT patterns are not valid in many places, so rather than
746 * having many extra checks on lov_pattern_supported, we have this separate
747 * check for non-released, non-readonly, non-DOM components
749 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
751 return pattern == LOV_PATTERN_RAID0 ||
752 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
756 #define LOV_MAXPOOLNAME 15
757 #define LOV_POOLNAMEF "%.15s"
758 /* The poolname "ignore" is used to force a component creation without pool */
759 #define LOV_POOL_IGNORE "ignore"
760 /* The poolname "inherit" is used to force a component to inherit the pool from
761 * parent or root directory
763 #define LOV_POOL_INHERIT "inherit"
764 /* The poolname "none" is deprecated in 2.15 (same behavior as "inherit") */
765 #define LOV_POOL_NONE "none"
767 static inline bool lov_pool_is_ignored(const char *pool)
769 return pool && strncmp(pool, LOV_POOL_IGNORE, LOV_MAXPOOLNAME) == 0;
772 static inline bool lov_pool_is_inherited(const char *pool)
774 return pool && (strncmp(pool, LOV_POOL_INHERIT, LOV_MAXPOOLNAME) == 0 ||
775 strncmp(pool, LOV_POOL_NONE, LOV_MAXPOOLNAME) == 0);
778 static inline bool lov_pool_is_reserved(const char *pool)
780 return lov_pool_is_ignored(pool) || lov_pool_is_inherited(pool);
783 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
784 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
785 #define LOV_MAX_STRIPE_COUNT_OLD 160
786 /* This calculation is crafted so that input of 4096 will result in 160
787 * which in turn is equal to old maximal stripe count.
788 * XXX: In fact this is too simpified for now, what it also need is to get
789 * ea_type argument to clearly know how much space each stripe consumes.
791 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
792 * allocation that is sufficient for the current generation of systems.
794 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
795 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
796 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
797 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
799 #define XATTR_LUSTRE_PREFIX "lustre."
800 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
802 /* Please update if XATTR_LUSTRE_LOV".set" groks more flags in the future */
803 #define allowed_lustre_lov(att) (strcmp((att), XATTR_LUSTRE_LOV".add") == 0 || \
804 strcmp((att), XATTR_LUSTRE_LOV".set") == 0 || \
805 strcmp((att), XATTR_LUSTRE_LOV".set.flags") == 0 || \
806 strcmp((att), XATTR_LUSTRE_LOV".del") == 0)
808 #define lov_user_ost_data lov_user_ost_data_v1
809 struct lov_user_ost_data_v1 { /* per-stripe data structure */
810 struct ost_id l_ost_oi; /* OST object ID */
811 __u32 l_ost_gen; /* generation of this OST index */
812 __u32 l_ost_idx; /* OST index in LOV */
813 } __attribute__((packed));
815 #define lov_user_md lov_user_md_v1
816 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
817 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
818 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
819 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
820 __u32 lmm_stripe_size; /* size of stripe in bytes */
821 __u16 lmm_stripe_count; /* num stripes in use for this object */
823 __u16 lmm_stripe_offset; /* starting stripe offset in
824 * lmm_objects, use when writing */
825 __u16 lmm_layout_gen; /* layout generation number
826 * used when reading */
828 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
829 } __attribute__((packed, __may_alias__));
831 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
832 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
833 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
834 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
835 __u32 lmm_stripe_size; /* size of stripe in bytes */
836 __u16 lmm_stripe_count; /* num stripes in use for this object */
838 __u16 lmm_stripe_offset; /* starting stripe offset in
839 * lmm_objects, use when writing */
840 __u16 lmm_layout_gen; /* layout generation number
841 * used when reading */
843 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
844 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
845 } __attribute__((packed, __may_alias__));
847 struct lov_foreign_md {
848 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
849 __u32 lfm_length; /* length of lfm_value */
850 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
851 __u32 lfm_flags; /* flags, type specific */
853 } __attribute__((packed));
855 #define lov_foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
856 offsetof(struct lov_foreign_md, lfm_value))
858 #define lov_foreign_size_le(lfm) \
859 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
860 offsetof(struct lov_foreign_md, lfm_value))
863 * The stripe size fields are shared for the extension size storage, however
864 * the extension size is stored in KB, not bytes.
866 #define SEL_UNIT_SIZE 1024llu
871 } __attribute__((packed));
873 #define DEXT "[%#llx, %#llx)"
874 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
876 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
877 struct lu_extent *e2)
879 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
882 static inline bool lu_extent_is_whole(struct lu_extent *e)
884 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
887 enum lov_comp_md_entry_flags {
888 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
889 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
890 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
891 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
892 LCME_FL_OFFLINE = 0x00000008, /* Not used */
893 LCME_FL_INIT = 0x00000010, /* instantiated */
894 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
895 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
896 LCME_FL_PARITY = 0x00000080, /* EC: a parity code component */
897 LCME_FL_COMPRESS = 0x00000100, /* the component should be compressed */
898 LCME_FL_PARTIAL = 0x00000200, /* some chunks in the component are
901 LCME_FL_NOCOMPR = 0x00000400, /* the component should not be
904 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
905 * won't be stored on disk
909 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
910 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
913 /* The component flags can be set by users at creation/modification time. */
914 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
917 /* The mirror flags can be set by users at creation time. */
918 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
920 /* The allowed flags obtained from the client at component creation time. */
921 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
923 /* The mirror flags sent by client */
924 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
926 /* These flags have meaning when set in a default layout and will be inherited
927 * from the default/template layout set on a directory.
929 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
932 /* lcme_id can be specified as certain flags, and the the first
933 * bit of lcme_id is used to indicate that the ID is representing
934 * certain LCME_FL_* but not a real ID. Which implies we can have
935 * at most 31 flags (see LCME_FL_XXX). */
938 LCME_ID_MAX = 0x7FFFFFFF,
939 LCME_ID_ALL = 0xFFFFFFFF,
940 LCME_ID_NOT_ID = LCME_FL_NEG
943 /* layout version equals to lcme_id, except some bits have special meanings */
944 enum layout_version_flags {
945 /* layout version reaches the high water mark to be increased to
946 * circularly reuse the smallest value
948 LU_LAYOUT_HIGEN = 0x40000000,
949 /* the highest bit is used to mark if the file is being resynced */
950 LU_LAYOUT_RESYNC = 0x80000000,
953 #define LCME_ID_MASK LCME_ID_MAX
955 struct lov_comp_md_entry_v1 {
956 __u32 lcme_id; /* unique id of component */
957 __u32 lcme_flags; /* LCME_FL_XXX */
958 /* file extent for component. If it's an EC code component, its flags
959 * contains LCME_FL_PARITY, and its extent covers the same extent of
960 * its corresponding data component.
962 struct lu_extent lcme_extent;
963 __u32 lcme_offset; /* offset of component blob,
964 start from lov_comp_md_v1 */
965 __u32 lcme_size; /* size of component blob */
966 __u32 lcme_layout_gen;
967 __u64 lcme_timestamp; /* snapshot time if applicable*/
968 __u8 lcme_dstripe_count; /* data stripe count,
971 __u8 lcme_cstripe_count; /* code stripe count,
974 __u8 lcme_compr_type; /* compress type */
975 __u8 lcme_compr_lvl:4; /* compress level */
976 __u8 lcme_compr_chunk_log_bits:4;
977 /* chunk_size = 2^(16+chunk_log_bits)
978 * i.e. power-of-two multiple of 64KiB */
979 } __attribute__((packed));
981 #define SEQ_ID_MAX 0x0000FFFF
982 #define SEQ_ID_MASK SEQ_ID_MAX
983 /* bit 30:16 of lcme_id is used to store mirror id */
984 #define MIRROR_ID_MASK 0x7FFF0000
985 #define MIRROR_ID_NEG 0x8000
986 #define MIRROR_ID_SHIFT 16
988 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
990 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
993 static inline __u16 mirror_id_of(__u32 id)
995 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
999 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
1001 enum lov_comp_md_flags {
1002 /* the least 4 bits are used by FLR to record file state */
1004 LCM_FL_RDONLY = 0x1,
1005 LCM_FL_WRITE_PENDING = 0x2,
1006 LCM_FL_SYNC_PENDING = 0x3,
1007 LCM_FL_PCC_RDONLY = 0x8,
1008 LCM_FL_FLR_MASK = 0xB,
1011 struct lov_comp_md_v1 {
1012 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
1013 __u32 lcm_size; /* overall size including this struct */
1014 __u32 lcm_layout_gen;
1016 __u16 lcm_entry_count;
1017 /* lcm_mirror_count stores the number of actual mirrors minus 1,
1018 * so that non-flr files will have value 0 meaning 1 mirror. */
1019 __u16 lcm_mirror_count;
1020 /* code components count, non-EC file contains 0 ec_count */
1022 __u8 lcm_padding3[1];
1023 __u16 lcm_padding1[2];
1025 struct lov_comp_md_entry_v1 lcm_entries[0];
1026 } __attribute__((packed));
1028 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
1030 if (stripes == LOV_ALL_STRIPES)
1033 if (lmm_magic == LOV_USER_MAGIC_V1)
1034 return sizeof(struct lov_user_md_v1) +
1035 stripes * sizeof(struct lov_user_ost_data_v1);
1036 return sizeof(struct lov_user_md_v3) +
1037 stripes * sizeof(struct lov_user_ost_data_v1);
1040 static inline __u32 lov_foreign_md_size(__u32 length)
1042 return length + offsetof(struct lov_foreign_md, lfm_value);
1045 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
1046 * use this. It is unsafe to #define those values in this header as it
1047 * is possible the application has already #included <sys/stat.h>. */
1048 #define lov_user_mds_data lov_user_mds_data_v2
1049 struct lov_user_mds_data_v1 {
1050 lstat_t lmd_st; /* MDS stat struct */
1051 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
1052 } __attribute__((packed));
1054 struct lov_user_mds_data_v2 {
1055 struct lu_fid lmd_fid; /* Lustre FID */
1056 lstatx_t lmd_stx; /* MDS statx struct */
1057 __u64 lmd_flags; /* MDS stat flags */
1058 __u32 lmd_lmmsize; /* LOV EA size */
1059 __u32 lmd_padding; /* unused */
1060 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
1061 } __attribute__((packed));
1063 struct lmv_user_mds_data {
1064 struct lu_fid lum_fid;
1067 } __attribute__((packed, __may_alias__));
1069 enum lmv_hash_type {
1070 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
1071 LMV_HASH_TYPE_ALL_CHARS = 1, /* simple sum of characters */
1072 LMV_HASH_TYPE_FNV_1A_64 = 2, /* reasonable non-cryptographic hash */
1073 LMV_HASH_TYPE_CRUSH = 3, /* double-hash to optimize migration */
1074 LMV_HASH_TYPE_CRUSH2 = 4, /* CRUSH with small fixes, LU-15692 */
1076 LMV_HASH_TYPE_DEFAULT = LMV_HASH_TYPE_FNV_1A_64
1079 static __attribute__((unused)) const char *mdt_hash_name[] = {
1088 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1089 * and the higher part will be the flag to indicate the status of object,
1090 * for example the object is being migrated. And the hash function
1091 * might be interpreted differently with different flags. */
1092 #define LMV_HASH_TYPE_MASK 0x0000ffff
1094 static inline bool lmv_is_known_hash_type(__u32 type)
1096 return (type & LMV_HASH_TYPE_MASK) > LMV_HASH_TYPE_UNKNOWN &&
1097 (type & LMV_HASH_TYPE_MASK) < LMV_HASH_TYPE_MAX;
1100 /* This flag indicates that overstriping (>1 stripe per MDT) is desired */
1101 #define LMV_HASH_FLAG_OVERSTRIPED 0x01000000
1102 /* fixed layout, such directories won't split automatically */
1103 /* NB, update LMV_HASH_FLAG_KNOWN when adding new flag */
1104 #define LMV_HASH_FLAG_FIXED 0x02000000
1105 #define LMV_HASH_FLAG_MERGE 0x04000000
1106 #define LMV_HASH_FLAG_SPLIT 0x08000000
1108 /* The striped directory has ever lost its master LMV EA, then LFSCK
1109 * re-generated it. This flag is used to indicate such case. It is an
1111 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
1113 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1114 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1116 #define LMV_HASH_FLAG_LAYOUT_CHANGE \
1117 (LMV_HASH_FLAG_MIGRATION | LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MERGE)
1119 #define LMV_HASH_FLAG_KNOWN 0xbe000000
1121 /* migration failure may leave hash type as
1122 * LMV_HASH_TYPE_UNKNOWN|LMV_HASH_FLAG_BAD_TYPE, which should be treated as
1123 * sane, so such directory can be accessed (resume migration or unlink).
1125 static inline bool lmv_is_sane_hash_type(__u32 type)
1127 return lmv_is_known_hash_type(type) ||
1128 type == (LMV_HASH_TYPE_UNKNOWN | LMV_HASH_FLAG_BAD_TYPE);
1131 /* both SPLIT and MIGRATION are set for directory split */
1132 static inline bool lmv_hash_is_splitting(__u32 hash)
1134 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1135 (LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MIGRATION);
1138 /* both MERGE and MIGRATION are set for directory merge */
1139 static inline bool lmv_hash_is_merging(__u32 hash)
1141 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1142 (LMV_HASH_FLAG_MERGE | LMV_HASH_FLAG_MIGRATION);
1145 /* only MIGRATION is set for directory migration */
1146 static inline bool lmv_hash_is_migrating(__u32 hash)
1148 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) == LMV_HASH_FLAG_MIGRATION;
1151 static inline bool lmv_hash_is_restriping(__u32 hash)
1153 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash);
1156 static inline bool lmv_hash_is_layout_changing(__u32 hash)
1158 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash) ||
1159 lmv_hash_is_migrating(hash);
1162 struct lustre_foreign_type {
1164 const char *lft_name;
1168 * LOV/LMV foreign types
1170 enum lustre_foreign_types {
1171 LU_FOREIGN_TYPE_NONE = 0,
1172 /* HSM copytool lhsm_posix */
1173 LU_FOREIGN_TYPE_POSIX = 1,
1174 /* Used for PCC-RW. PCCRW components are local to a single archive. */
1175 LU_FOREIGN_TYPE_PCCRW = 2,
1176 /* Used for PCC-RO. PCCRO components may be shared between archives. */
1177 LU_FOREIGN_TYPE_PCCRO = 3,
1179 LU_FOREIGN_TYPE_S3 = 4,
1181 LU_FOREIGN_TYPE_SYMLINK = 0xda05,
1182 /* must be the max/last one */
1183 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
1186 extern struct lustre_foreign_type lu_foreign_types[];
1188 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
1189 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
1190 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
1191 #define lmv_user_md lmv_user_md_v1
1192 struct lmv_user_md_v1 {
1193 __u32 lum_magic; /* must be the first field */
1194 __u32 lum_stripe_count; /* dirstripe count */
1195 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
1196 __u32 lum_hash_type; /* Dir stripe policy */
1197 __u32 lum_type; /* LMV type: default */
1198 __u8 lum_max_inherit; /* inherit depth of default LMV */
1199 __u8 lum_max_inherit_rr; /* inherit depth of default LMV to round-robin mkdir */
1203 char lum_pool_name[LOV_MAXPOOLNAME + 1];
1204 struct lmv_user_mds_data lum_objects[0];
1205 } __attribute__((packed));
1207 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
1209 if (size <= sizeof(struct lmv_user_md))
1212 size -= sizeof(struct lmv_user_md);
1213 return (size + sizeof(struct lmv_user_mds_data) - 1) /
1214 sizeof(struct lmv_user_mds_data);
1218 * NB, historically default layout didn't set type, but use XATTR name to differ
1219 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
1220 * and still use the same method.
1223 LMV_TYPE_DEFAULT = 0x0000,
1224 /* fetch raw default LMV set on directory inode */
1225 LMV_TYPE_RAW = 0x0001,
1228 /* lum_max_inherit will be decreased by 1 after each inheritance if it's not
1229 * LMV_INHERIT_UNLIMITED or > LMV_INHERIT_MAX.
1232 /* for historical reason, 0 means unlimited inheritance */
1233 LMV_INHERIT_UNLIMITED = 0,
1234 /* unlimited lum_max_inherit by default for plain stripe (0 or 1) */
1235 LMV_INHERIT_DEFAULT_PLAIN = LMV_INHERIT_UNLIMITED,
1236 /* not inherit any more */
1237 LMV_INHERIT_END = 1,
1238 /* for multiple stripes, the default lum_max_inherit is 3 */
1239 LMV_INHERIT_DEFAULT_STRIPED = 3,
1240 /* max inherit depth */
1241 LMV_INHERIT_MAX = 250,
1242 /* [251, 254] are reserved */
1243 /* not set, or when inherit depth goes beyond end, */
1244 LMV_INHERIT_NONE = 255,
1248 /* not set, or when inherit_rr depth goes beyond end, */
1249 LMV_INHERIT_RR_NONE = 0,
1250 /* disable lum_max_inherit_rr by default */
1251 LMV_INHERIT_RR_DEFAULT = LMV_INHERIT_RR_NONE,
1252 /* not inherit any more */
1253 LMV_INHERIT_RR_END = 1,
1254 /* default inherit_rr of ROOT */
1255 LMV_INHERIT_RR_ROOT = 3,
1256 /* max inherit depth */
1257 LMV_INHERIT_RR_MAX = 250,
1258 /* [251, 254] are reserved */
1259 /* unlimited inheritance */
1260 LMV_INHERIT_RR_UNLIMITED = 255,
1263 static inline int lmv_user_md_size(int stripes, int lmm_magic)
1265 int size = sizeof(struct lmv_user_md);
1267 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
1268 size += stripes * sizeof(struct lmv_user_mds_data);
1273 struct ll_recreate_obj {
1279 __u64 id; /* holds object id */
1280 __u32 generation; /* holds object generation */
1281 __u32 f_type; /* holds object type or stripe idx when passing it to
1282 * OST for saving into EA. */
1287 char uuid[UUID_MAX];
1290 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
1291 const struct obd_uuid *u2)
1293 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
1296 static inline int obd_uuid_empty(struct obd_uuid *uuid)
1298 return uuid->uuid[0] == '\0';
1301 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
1303 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
1304 uuid->uuid[sizeof(*uuid) - 1] = '\0';
1307 /* For printf's only, make sure uuid is terminated */
1308 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
1313 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
1314 /* Obviously not safe, but for printfs, no real harm done...
1315 we're always null-terminated, even in a race. */
1316 static char temp[sizeof(*uuid->uuid)];
1318 memcpy(temp, uuid->uuid, sizeof(*uuid->uuid) - 1);
1319 temp[sizeof(*uuid->uuid) - 1] = '\0';
1323 return (char *)(uuid->uuid);
1326 #define LUSTRE_MAXFSNAME 8
1327 #define LUSTRE_MAXINSTANCE 16
1329 /* Extract fsname from uuid (or target name) of a target
1330 e.g. (myfs-OST0007_UUID -> myfs)
1331 see also deuuidify. */
1332 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1336 strncpy(buf, uuid, buflen - 1);
1337 buf[buflen - 1] = '\0';
1338 p = strrchr(buf, '-');
1343 /* printf display format for Lustre FIDs
1344 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1345 #define FID_NOBRACE_LEN 40
1346 #define FID_LEN (FID_NOBRACE_LEN + 2)
1347 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1348 #define DFID "[" DFID_NOBRACE "]"
1349 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1351 /* scanf input parse format for fids in DFID_NOBRACE format
1352 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1353 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1354 #define SFID "0x%llx:0x%x:0x%x"
1355 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1356 #define PLOGID(logid) (unsigned long long)(logid)->lgl_oi.oi.oi_seq, (__u32)(logid)->lgl_oi.oi.oi_id, 0
1358 /********* Quotas **********/
1360 /* From linux/fs/quota/quota.c */
1361 static inline __u64 toqb(__kernel_size_t space)
1363 return (space + QIF_DQBLKSIZE - 1) >> QIF_DQBLKSIZE_BITS;
1366 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1367 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1368 #define Q_GETOINFO 0x800102 /* get obd quota info */
1369 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1370 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1372 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1373 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1374 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1375 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1376 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1377 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1378 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1379 /* lustre-specific control commands */
1380 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1381 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1382 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1383 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1384 #define LUSTRE_Q_GETQUOTAPOOL 0x80000f /* get user pool quota */
1385 #define LUSTRE_Q_SETQUOTAPOOL 0x800010 /* set user pool quota */
1386 #define LUSTRE_Q_GETINFOPOOL 0x800011 /* get pool quota info */
1387 #define LUSTRE_Q_SETINFOPOOL 0x800012 /* set pool quota info */
1388 #define LUSTRE_Q_GETDEFAULT_POOL 0x800013 /* get default pool quota*/
1389 #define LUSTRE_Q_SETDEFAULT_POOL 0x800014 /* set default pool quota */
1390 #define LUSTRE_Q_DELETEQID 0x800015 /* delete quota ID */
1391 #define LUSTRE_Q_RESETQID 0x800016 /* reset quota ID */
1392 /* In the current Lustre implementation, the grace time is either the time
1393 * or the timestamp to be used after some quota ID exceeds the soft limt,
1394 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1396 #define LQUOTA_GRACE_BITS 48
1397 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1398 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1399 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1400 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1401 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1403 /* special grace time, only notify the user when its quota is over soft limit
1404 * but doesn't block new writes until the hard limit is reached. */
1405 #define NOTIFY_GRACE "notify"
1406 #define NOTIFY_GRACE_TIME LQUOTA_GRACE_MASK
1408 /* different quota flags */
1410 /* the default quota flag, the corresponding quota ID will use the default
1411 * quota setting, the hardlimit and softlimit of its quota record in the global
1412 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1413 * and high 16 bits will contain this flag (see above comment).
1415 #define LQUOTA_FLAG_DEFAULT 0x0001
1416 #define LQUOTA_FLAG_DELETED 0x0002
1417 #define LQUOTA_FLAG_RESET 0x0004
1419 #define LUSTRE_Q_CMD_IS_POOL(cmd) \
1420 (cmd == LUSTRE_Q_GETQUOTAPOOL || \
1421 cmd == LUSTRE_Q_SETQUOTAPOOL || \
1422 cmd == LUSTRE_Q_SETINFOPOOL || \
1423 cmd == LUSTRE_Q_GETINFOPOOL || \
1424 cmd == LUSTRE_Q_SETDEFAULT_POOL || \
1425 cmd == LUSTRE_Q_GETDEFAULT_POOL)
1427 #define ALLQUOTA 255 /* set all quota */
1428 static inline const char *qtype_name(int qtype)
1441 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1444 #define N_PERMS_MAX 64
1446 struct perm_downcall_data {
1452 struct identity_downcall_data {
1459 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1460 __u32 idd_groups[0];
1463 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
1464 /* old interface struct is deprecated in 2.14 */
1465 #define SEPOL_DOWNCALL_MAGIC_OLD 0x8b8bb842
1466 struct sepol_downcall_data_old {
1468 __s64 sdd_sepol_mtime;
1469 __u16 sdd_sepol_len;
1474 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb843
1475 struct sepol_downcall_data {
1477 __u16 sdd_sepol_len;
1479 __s64 sdd_sepol_mtime;
1483 #ifdef NEED_QUOTA_DEFS
1485 #define QIF_BLIMITS 1
1487 #define QIF_ILIMITS 4
1488 #define QIF_INODES 8
1489 #define QIF_BTIME 16
1490 #define QIF_ITIME 32
1491 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1492 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1493 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1494 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1497 #endif /* !__KERNEL__ */
1499 /* lustre volatile file support
1500 * file name header: ".^L^S^T^R:volatile"
1502 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1503 #define LUSTRE_VOLATILE_HDR_LEN 14
1505 enum lustre_quota_version {
1509 /* XXX: same as if_dqinfo struct in kernel */
1517 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1519 __u64 dqb_bhardlimit; /* kbytes unit */
1520 __u64 dqb_bsoftlimit; /* kbytes unit */
1521 __u64 dqb_curspace; /* bytes unit */
1522 __u64 dqb_ihardlimit;
1523 __u64 dqb_isoftlimit;
1524 __u64 dqb_curinodes;
1538 struct if_quotactl {
1545 struct obd_dqinfo qc_dqinfo;
1546 struct obd_dqblk qc_dqblk;
1548 struct obd_uuid obd_uuid;
1552 /* swap layout flags */
1553 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1554 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1555 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1556 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1557 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1559 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1560 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1561 struct lustre_swap_layouts {
1569 /** Bit-mask of valid attributes */
1570 /* The LA_* flags are written to disk as part of the ChangeLog records
1571 * so they are part of the on-disk and network protocol, and cannot be changed.
1572 * Only the first 12 bits are currently saved.
1575 LA_ATIME = 1 << 0, /* 0x00001 */
1576 LA_MTIME = 1 << 1, /* 0x00002 */
1577 LA_CTIME = 1 << 2, /* 0x00004 */
1578 LA_SIZE = 1 << 3, /* 0x00008 */
1579 LA_MODE = 1 << 4, /* 0x00010 */
1580 LA_UID = 1 << 5, /* 0x00020 */
1581 LA_GID = 1 << 6, /* 0x00040 */
1582 LA_BLOCKS = 1 << 7, /* 0x00080 */
1583 LA_TYPE = 1 << 8, /* 0x00100 */
1584 LA_FLAGS = 1 << 9, /* 0x00200 */
1585 LA_NLINK = 1 << 10, /* 0x00400 */
1586 LA_RDEV = 1 << 11, /* 0x00800 */
1587 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1588 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1589 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1590 LA_PROJID = 1 << 15, /* 0x08000 */
1591 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1592 LA_LSIZE = 1 << 17, /* 0x20000 */
1593 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1594 LA_BTIME = 1 << 19, /* 0x80000 */
1596 * Attributes must be transmitted to OST objects
1598 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1601 #define MDS_FMODE_READ 00000001
1602 #define MDS_FMODE_WRITE 00000002
1604 #define MDS_FMODE_CLOSED 00000000
1605 #define MDS_FMODE_EXEC 00000004
1606 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1607 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1608 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1610 #define MDS_OPEN_CREATED 00000010
1611 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1613 #define MDS_OPEN_CREAT 00000100
1614 #define MDS_OPEN_EXCL 00000200
1615 #define MDS_OPEN_TRUNC 00001000
1616 #define MDS_OPEN_APPEND 00002000
1617 #define MDS_OPEN_SYNC 00010000
1618 #define MDS_OPEN_DIRECTORY 00200000
1620 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1621 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1622 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1623 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1624 * We do not support JOIN FILE
1625 * anymore, reserve this flags
1626 * just for preventing such bit
1629 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1630 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1631 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1632 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1633 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1635 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1637 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1638 * delegation, succeed if it's not
1639 * being opened with conflict mode.
1641 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1643 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1644 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1645 * for newly created file */
1646 #define MDS_OP_WITH_FID 020000000000000ULL /* operation carried out by FID */
1647 #define MDS_OPEN_DEFAULT_LMV 040000000000000ULL /* open fetches default LMV,
1648 * or mkdir with default LMV
1651 /* lustre internal open flags, which should not be set from user space */
1652 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1653 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1654 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1655 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1656 MDS_OPEN_PCC | MDS_OP_WITH_FID | \
1657 MDS_OPEN_DEFAULT_LMV)
1660 /********* Changelogs **********/
1661 /** Changelog record types */
1662 enum changelog_rec_type {
1665 CL_CREATE = 1, /* namespace */
1666 CL_MKDIR = 2, /* namespace */
1667 CL_HARDLINK = 3, /* namespace */
1668 CL_SOFTLINK = 4, /* namespace */
1669 CL_MKNOD = 5, /* namespace */
1670 CL_UNLINK = 6, /* namespace */
1671 CL_RMDIR = 7, /* namespace */
1672 CL_RENAME = 8, /* namespace */
1673 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1674 CL_OPEN = 10, /* not currently used */
1675 CL_CLOSE = 11, /* may be written to log only with mtime change */
1676 CL_LAYOUT = 12, /* file layout/striping modified */
1680 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1681 CL_HSM = 16, /* HSM specific events, see flags */
1682 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1686 CL_FLRW = 21, /* FLR: file was firstly written */
1687 CL_RESYNC = 22, /* FLR: file was resync-ed */
1689 CL_DN_OPEN = 24, /* denied open */
1693 static inline const char *changelog_type2str(int type) {
1694 static const char *const changelog_str[] = {
1695 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1696 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1697 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1698 "FLRW", "RESYNC","GXATR", "NOPEN",
1701 if (type >= 0 && type < CL_LAST)
1702 return changelog_str[type];
1706 /* 12 bits of per-record data can be stored in the bottom of the flags */
1707 #define CLF_FLAGSHIFT 12
1708 enum changelog_rec_flags {
1709 CLF_VERSION = 0x1000,
1710 CLF_RENAME = 0x2000,
1712 CLF_EXTRA_FLAGS = 0x8000,
1713 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1715 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1716 CLF_VERMASK = ~CLF_FLAGMASK,
1720 /* Anything under the flagmask may be per-type (if desired) */
1721 /* Flags for unlink */
1722 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1723 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1724 /* HSM cleaning needed */
1725 /* Flags for rename */
1726 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1728 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1729 * has an archive in backend */
1732 /* 12b used (from high weight to low weight):
1737 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1738 #define CLF_HSM_ERR_H 6
1739 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1740 #define CLF_HSM_EVENT_H 9
1741 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1742 #define CLF_HSM_FLAG_H 11
1743 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1744 #define CLF_HSM_SPARE_H 15
1745 #define CLF_HSM_LAST 15
1747 /* Remove bits higher than _h, then extract the value
1748 * between _h and _l by shifting lower weigth to bit 0. */
1749 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1750 >> (CLF_HSM_LAST - _h + _l))
1752 #define CLF_HSM_SUCCESS 0x00
1753 #define CLF_HSM_MAXERROR 0x7E
1754 #define CLF_HSM_ERROVERFLOW 0x7F
1756 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1758 /* 3 bits field => 8 values allowed */
1770 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1772 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1776 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1779 *clf_flags = (enum changelog_rec_flags)
1780 (*clf_flags | (he << CLF_HSM_EVENT_L));
1783 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1785 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1788 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1791 *clf_flags = (enum changelog_rec_flags)
1792 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1795 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1797 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1800 static inline int hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1803 /* In case a negative error is given */
1806 if (error > CLF_HSM_MAXERROR)
1807 error = CLF_HSM_ERROVERFLOW;
1809 *clf_flags = (enum changelog_rec_flags)
1810 (*clf_flags | (error << CLF_HSM_ERR_L));
1812 return error == CLF_HSM_ERROVERFLOW ? -EOVERFLOW : 0;
1815 enum changelog_rec_extra_flags {
1817 CLFE_UIDGID = 0x0001,
1820 CLFE_XATTR = 0x0008,
1821 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1824 enum changelog_send_flag {
1825 /* Use changelog follow mode: llapi_changelog_recv() will not stop at
1826 * the end of records and wait for new records to be generated.
1828 CHANGELOG_FLAG_FOLLOW = 0x01,
1829 /* Deprecated since Lustre 2.10 */
1830 CHANGELOG_FLAG_BLOCK = 0x02,
1831 /* Pack jobid into the changelog records if available. */
1832 CHANGELOG_FLAG_JOBID = 0x04,
1833 /* Pack additional flag bits into the changelog record */
1834 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1837 enum changelog_send_extra_flag {
1838 /* Pack uid/gid into the changelog record */
1839 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1840 /* Pack nid into the changelog record */
1841 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1842 /* Pack open mode into the changelog record */
1843 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1844 /* Pack xattr name into the changelog record */
1845 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1848 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1849 changelog_rec_offset(CLF_SUPPORTED, \
1852 /* 31 usable bytes string + null terminator. */
1853 #define LUSTRE_JOBID_SIZE 32
1855 /* This is the minimal changelog record. It can contain extensions
1856 * such as rename fields or process jobid. Its exact content is described
1857 * by the cr_flags and cr_extra_flags.
1859 * Extensions are packed in the same order as their corresponding flags,
1860 * then in the same order as their corresponding extra flags.
1862 struct changelog_rec {
1864 __u16 cr_flags; /**< \a changelog_rec_flags */
1865 __u32 cr_type; /**< \a changelog_rec_type */
1866 __u64 cr_index; /**< changelog record number */
1867 __u64 cr_prev; /**< last index for this target fid */
1870 struct lu_fid cr_tfid; /**< target fid */
1871 __u32 cr_markerflags; /**< CL_MARK flags */
1873 struct lu_fid cr_pfid; /**< parent fid */
1874 } __attribute__ ((packed));
1876 /* Changelog extension for RENAME. */
1877 struct changelog_ext_rename {
1878 struct lu_fid cr_sfid; /**< source fid, or zero */
1879 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1882 /* Changelog extension to include JOBID. */
1883 struct changelog_ext_jobid {
1884 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1887 /* Changelog extension to include additional flags. */
1888 struct changelog_ext_extra_flags {
1889 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1892 /* Changelog extra extension to include UID/GID. */
1893 struct changelog_ext_uidgid {
1898 /* Changelog extra extension to include NID. */
1899 struct changelog_ext_nid {
1900 /* have __u64 instead of lnet_nid_t type for use by client api */
1902 /* for use when IPv6 support is added */
1907 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1908 struct changelog_ext_openmode {
1912 /* Changelog extra extension to include xattr */
1913 struct changelog_ext_xattr {
1914 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1917 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1918 const struct changelog_rec *rec);
1920 static inline __kernel_size_t changelog_rec_offset(enum changelog_rec_flags crf,
1921 enum changelog_rec_extra_flags cref)
1923 __kernel_size_t size = sizeof(struct changelog_rec);
1925 if (crf & CLF_RENAME)
1926 size += sizeof(struct changelog_ext_rename);
1928 if (crf & CLF_JOBID)
1929 size += sizeof(struct changelog_ext_jobid);
1931 if (crf & CLF_EXTRA_FLAGS) {
1932 size += sizeof(struct changelog_ext_extra_flags);
1933 if (cref & CLFE_UIDGID)
1934 size += sizeof(struct changelog_ext_uidgid);
1935 if (cref & CLFE_NID)
1936 size += sizeof(struct changelog_ext_nid);
1937 if (cref & CLFE_OPEN)
1938 size += sizeof(struct changelog_ext_openmode);
1939 if (cref & CLFE_XATTR)
1940 size += sizeof(struct changelog_ext_xattr);
1946 static inline __kernel_size_t changelog_rec_size(const struct changelog_rec *rec)
1948 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1950 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1951 cref = (enum changelog_rec_extra_flags)
1952 changelog_rec_extra_flags(rec)->cr_extra_flags;
1954 return changelog_rec_offset(
1955 (enum changelog_rec_flags)rec->cr_flags, cref);
1958 static inline __kernel_size_t changelog_rec_varsize(const struct changelog_rec *rec)
1960 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1964 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1966 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1967 (rec->cr_flags & CLF_VERSION);
1969 return (struct changelog_ext_rename *)((char *)rec +
1970 changelog_rec_offset(crf,
1974 /* The jobid follows the rename extension, if present */
1976 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1978 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1979 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1981 return (struct changelog_ext_jobid *)((char *)rec +
1982 changelog_rec_offset(crf,
1986 /* The additional flags follow the rename and jobid extensions, if present */
1988 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1989 const struct changelog_rec *rec)
1991 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1992 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1994 return (struct changelog_ext_extra_flags *)((char *)rec +
1995 changelog_rec_offset(crf,
1999 /* The uid/gid is the first extra extension */
2001 struct changelog_ext_uidgid *changelog_rec_uidgid(
2002 const struct changelog_rec *rec)
2004 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2006 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2008 return (struct changelog_ext_uidgid *)((char *)rec +
2009 changelog_rec_offset(crf,
2013 /* The nid is the second extra extension */
2015 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
2017 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2019 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2020 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2022 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2023 cref = (enum changelog_rec_extra_flags)
2024 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2027 return (struct changelog_ext_nid *)((char *)rec +
2028 changelog_rec_offset(crf, cref));
2031 /* The OPEN mode is the third extra extension */
2033 struct changelog_ext_openmode *changelog_rec_openmode(
2034 const struct changelog_rec *rec)
2036 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2038 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2039 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2041 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
2042 cref = (enum changelog_rec_extra_flags)
2043 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2044 (CLFE_UIDGID | CLFE_NID));
2047 return (struct changelog_ext_openmode *)((char *)rec +
2048 changelog_rec_offset(crf, cref));
2051 /* The xattr name is the fourth extra extension */
2053 struct changelog_ext_xattr *changelog_rec_xattr(
2054 const struct changelog_rec *rec)
2056 enum changelog_rec_flags crf = (enum changelog_rec_flags)
2058 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
2059 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2061 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2062 cref = (enum changelog_rec_extra_flags)
2063 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2064 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
2066 return (struct changelog_ext_xattr *)((char *)rec +
2067 changelog_rec_offset(crf, cref));
2070 /* The name follows the rename, jobid and extra flags extns, if present */
2071 static inline char *changelog_rec_name(const struct changelog_rec *rec)
2073 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2075 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2076 cref = (enum changelog_rec_extra_flags)
2077 changelog_rec_extra_flags(rec)->cr_extra_flags;
2079 return (char *)rec + changelog_rec_offset(
2080 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
2081 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
2084 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
2086 char *str = changelog_rec_name(rec);
2088 while (*str != '\0')
2093 static inline __kernel_size_t changelog_rec_snamelen(const struct changelog_rec *rec)
2095 return strlen(changelog_rec_sname(rec));
2099 * Remap a record to the desired format as specified by the crf flags.
2100 * The record must be big enough to contain the final remapped version.
2101 * Superfluous extension fields are removed and missing ones are added
2102 * and zeroed. The flags of the record are updated accordingly.
2104 * The jobid and rename extensions can be added to a record, to match the
2105 * format an application expects, typically. In this case, the newly added
2106 * fields will be zeroed.
2107 * The Jobid field can be removed, to guarantee compatibility with older
2108 * clients that don't expect this field in the records they process.
2110 * The following assumptions are being made:
2111 * - CLF_RENAME will not be removed
2112 * - CLF_JOBID will not be added without CLF_RENAME being added too
2113 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
2115 * @param[in,out] rec The record to remap.
2116 * @param[in] crf_wanted Flags describing the desired extensions.
2117 * @param[in] cref_want Flags describing the desired extra extensions.
2119 static inline void changelog_remap_rec(struct changelog_rec *rec,
2120 enum changelog_rec_flags crf_wanted,
2121 enum changelog_rec_extra_flags cref_want)
2123 char *xattr_mov = NULL;
2124 char *omd_mov = NULL;
2125 char *nid_mov = NULL;
2126 char *uidgid_mov = NULL;
2130 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2132 crf_wanted = (enum changelog_rec_flags)
2133 (crf_wanted & CLF_SUPPORTED);
2134 cref_want = (enum changelog_rec_extra_flags)
2135 (cref_want & CLFE_SUPPORTED);
2137 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
2138 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
2139 (rec->cr_flags & CLF_EXTRA_FLAGS &&
2140 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2146 /* First move the variable-length name field */
2147 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
2148 changelog_rec_name(rec), rec->cr_namelen);
2150 /* Locations of extensions in the remapped record */
2151 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
2152 xattr_mov = (char *)rec +
2153 changelog_rec_offset(
2154 (enum changelog_rec_flags)
2155 (crf_wanted & CLF_SUPPORTED),
2156 (enum changelog_rec_extra_flags)
2157 (cref_want & ~CLFE_XATTR));
2158 omd_mov = (char *)rec +
2159 changelog_rec_offset(
2160 (enum changelog_rec_flags)
2161 (crf_wanted & CLF_SUPPORTED),
2162 (enum changelog_rec_extra_flags)
2163 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
2164 nid_mov = (char *)rec +
2165 changelog_rec_offset(
2166 (enum changelog_rec_flags)
2167 (crf_wanted & CLF_SUPPORTED),
2168 (enum changelog_rec_extra_flags)
2170 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
2171 uidgid_mov = (char *)rec +
2172 changelog_rec_offset(
2173 (enum changelog_rec_flags)
2174 (crf_wanted & CLF_SUPPORTED),
2175 (enum changelog_rec_extra_flags)
2176 (cref_want & ~(CLFE_UIDGID |
2180 cref = (enum changelog_rec_extra_flags)
2181 changelog_rec_extra_flags(rec)->cr_extra_flags;
2184 ef_mov = (char *)rec +
2185 changelog_rec_offset(
2186 (enum changelog_rec_flags)
2187 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
2188 jid_mov = (char *)rec +
2189 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2190 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
2192 rnm_mov = (char *)rec +
2193 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2199 /* Move the extension fields to the desired positions */
2200 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2201 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
2202 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
2203 memmove(xattr_mov, changelog_rec_xattr(rec),
2204 sizeof(struct changelog_ext_xattr));
2206 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
2207 memmove(omd_mov, changelog_rec_openmode(rec),
2208 sizeof(struct changelog_ext_openmode));
2210 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
2211 memmove(nid_mov, changelog_rec_nid(rec),
2212 sizeof(struct changelog_ext_nid));
2214 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
2215 memmove(uidgid_mov, changelog_rec_uidgid(rec),
2216 sizeof(struct changelog_ext_uidgid));
2218 memmove(ef_mov, changelog_rec_extra_flags(rec),
2219 sizeof(struct changelog_ext_extra_flags));
2222 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
2223 memmove(jid_mov, changelog_rec_jobid(rec),
2224 sizeof(struct changelog_ext_jobid));
2226 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
2227 memmove(rnm_mov, changelog_rec_rename(rec),
2228 sizeof(struct changelog_ext_rename));
2230 /* Clear newly added fields */
2231 if (xattr_mov && (cref_want & CLFE_XATTR) &&
2232 !(cref & CLFE_XATTR))
2233 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
2235 if (omd_mov && (cref_want & CLFE_OPEN) &&
2236 !(cref & CLFE_OPEN))
2237 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
2239 if (nid_mov && (cref_want & CLFE_NID) &&
2241 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
2243 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
2244 !(cref & CLFE_UIDGID))
2245 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
2247 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2248 !(rec->cr_flags & CLF_EXTRA_FLAGS))
2249 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
2251 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
2252 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
2254 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
2255 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
2257 /* Update the record's flags accordingly */
2258 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
2259 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2260 changelog_rec_extra_flags(rec)->cr_extra_flags =
2261 changelog_rec_extra_flags(rec)->cr_extra_flags |
2265 enum changelog_message_type {
2266 CL_RECORD = 10, /* message is a changelog_rec */
2267 CL_EOF = 11, /* at end of current changelog */
2270 /********* Misc **********/
2272 struct ioc_data_version {
2274 __u32 idv_layout_version; /* FLR: layout version for OST objects */
2275 __u32 idv_flags; /* enum ioc_data_version_flags */
2278 enum ioc_data_version_flags {
2279 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
2280 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
2281 LL_DV_SZ_UPDATE = (1 << 2), /* Update the file size on the client */
2285 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
2288 #define dot_lustre_name ".lustre"
2289 #define dot_fscrypt_name ".fscrypt"
2292 /********* HSM **********/
2296 struct lov_hsm_base {
2297 /* HSM archive ID */
2298 __u64 lhb_archive_id;
2299 /* Data version associated with the last archiving, if any. */
2300 __u64 lhb_archive_ver;
2301 /* Identifier within HSM backend */
2302 char lhb_uuid[UUID_MAX];
2306 * HSM layout is a kind of FOREIGN layout.
2309 /* LOV_MAGIC_FOREIGN */
2311 /* To make HSM layout compatible with lov_foreign_md, this @length
2312 * includes everything after @lhm_flags: sizeof(lhm_archive_id) +
2313 * sizeof(lhm_archive_ver) + lenght of lhm_archive_uuid.
2316 /* HSM type, see LU_FOREIGN_TYPE_(POSIX, S3, PCCRW, PCCRO}. */
2318 /* HSM flags, see enum hsm_states */
2321 * Data structure members above are compatible with @lov_foreign_md.
2322 * The following members are private to HSM layout.
2324 struct lov_hsm_base lhm_hsm;
2325 } __attribute__((packed));
2327 #define lhm_archive_id lhm_hsm.lhb_archive_id
2328 #define lhm_archive_ver lhm_hsm.lhb_archive_ver
2329 #define lhm_archive_uuid lhm_hsm.lhb_uuid
2331 static inline bool lov_hsm_type_supported(__u32 type)
2333 return type == LU_FOREIGN_TYPE_POSIX || type == LU_FOREIGN_TYPE_PCCRW ||
2334 type == LU_FOREIGN_TYPE_PCCRO || type == LU_FOREIGN_TYPE_S3;
2337 static inline bool lov_foreign_type_supported(__u32 type)
2339 return lov_hsm_type_supported(type) || type == LU_FOREIGN_TYPE_SYMLINK;
2343 * HSM per-file state
2344 * See HSM_FLAGS below.
2347 HS_NONE = 0x00000000,
2348 HS_EXISTS = 0x00000001,
2349 HS_DIRTY = 0x00000002,
2350 HS_RELEASED = 0x00000004,
2351 HS_ARCHIVED = 0x00000008,
2352 HS_NORELEASE = 0x00000010,
2353 HS_NOARCHIVE = 0x00000020,
2354 HS_LOST = 0x00000040,
2355 HS_PCCRW = 0x00000080,
2356 HS_PCCRO = 0x00000100,
2359 /* HSM user-setable flags. */
2360 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
2362 /* Other HSM flags. */
2363 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED | \
2364 HS_PCCRW | HS_PCCRO)
2367 * All HSM-related possible flags that could be applied to a file.
2368 * This should be kept in sync with hsm_states.
2370 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
2373 * HSM request progress state
2375 enum hsm_progress_states {
2382 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
2385 case HPS_WAITING: return "waiting";
2386 case HPS_RUNNING: return "running";
2387 case HPS_DONE: return "done";
2388 default: return "unknown";
2395 } __attribute__((packed));
2398 * Current HSM states of a Lustre file.
2400 * This structure purpose is to be sent to user-space mainly. It describes the
2401 * current HSM flags and in-progress action.
2403 struct hsm_user_state {
2404 /** Current HSM states, from enum hsm_states. */
2406 __u32 hus_archive_id;
2407 /** The current undergoing action, if there is one */
2408 __u32 hus_in_progress_state;
2409 __u32 hus_in_progress_action;
2410 struct hsm_extent hus_in_progress_location;
2411 char hus_extended_info[];
2414 struct hsm_state_set_ioc {
2415 struct lu_fid hssi_fid;
2417 __u64 hssi_clearmask;
2421 * This structure describes the current in-progress action for a file.
2422 * it is retuned to user space and send over the wire
2424 struct hsm_current_action {
2425 /** The current undergoing action, if there is one */
2426 /* state is one of hsm_progress_states */
2428 /* action is one of hsm_user_action */
2430 struct hsm_extent hca_location;
2433 /***** HSM user requests ******/
2434 /* User-generated (lfs/ioctl) request types */
2435 enum hsm_user_action {
2436 HUA_NONE = 1, /* no action (noop) */
2437 HUA_ARCHIVE = 10, /* copy to hsm */
2438 HUA_RESTORE = 11, /* prestage */
2439 HUA_RELEASE = 12, /* drop ost objects */
2440 HUA_REMOVE = 13, /* remove from archive */
2441 HUA_CANCEL = 14 /* cancel a request */
2444 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2447 case HUA_NONE: return "NOOP";
2448 case HUA_ARCHIVE: return "ARCHIVE";
2449 case HUA_RESTORE: return "RESTORE";
2450 case HUA_RELEASE: return "RELEASE";
2451 case HUA_REMOVE: return "REMOVE";
2452 case HUA_CANCEL: return "CANCEL";
2453 default: return "UNKNOWN";
2458 * List of hr_flags (bit field)
2460 #define HSM_FORCE_ACTION 0x0001
2461 /* used by CT, cannot be set by user */
2462 #define HSM_GHOST_COPY 0x0002
2465 * Contains all the fixed part of struct hsm_user_request.
2468 struct hsm_request {
2469 __u32 hr_action; /* enum hsm_user_action */
2470 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2471 __u64 hr_flags; /* request flags */
2472 __u32 hr_itemcount; /* item count in hur_user_item vector */
2476 struct hsm_user_item {
2477 struct lu_fid hui_fid;
2478 struct hsm_extent hui_extent;
2479 } __attribute__((packed));
2481 struct hsm_user_request {
2482 struct hsm_request hur_request;
2483 struct hsm_user_item hur_user_item[0];
2484 /* extra data blob at end of struct (after all
2485 * hur_user_items), only use helpers to access it
2487 } __attribute__((packed));
2489 /** Return pointer to data field in a hsm user request */
2490 static inline void *hur_data(struct hsm_user_request *hur)
2492 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2496 * Compute the current length of the provided hsm_user_request. This returns -1
2497 * instead of an errno because __kernel_ssize_t is defined to be only
2500 * return -1 on bounds check error.
2502 static inline __kernel_size_t hur_len(struct hsm_user_request *hur)
2506 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2507 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2508 (__u64)hur->hur_request.hr_itemcount *
2509 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2511 if ((__kernel_ssize_t)size < 0)
2517 /****** HSM RPCs to copytool *****/
2518 /* Message types the copytool may receive */
2519 enum hsm_message_type {
2520 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2523 /* Actions the copytool may be instructed to take for a given action_item */
2524 enum hsm_copytool_action {
2525 HSMA_NONE = 10, /* no action */
2526 HSMA_ARCHIVE = 20, /* arbitrary offset */
2532 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2535 case HSMA_NONE: return "NOOP";
2536 case HSMA_ARCHIVE: return "ARCHIVE";
2537 case HSMA_RESTORE: return "RESTORE";
2538 case HSMA_REMOVE: return "REMOVE";
2539 case HSMA_CANCEL: return "CANCEL";
2540 default: return "UNKNOWN";
2544 /* Copytool item action description */
2545 struct hsm_action_item {
2546 __u32 hai_len; /* valid size of this struct */
2547 __u32 hai_action; /* hsm_copytool_action, but use known size */
2548 struct lu_fid hai_fid; /* Lustre FID to operate on */
2549 struct lu_fid hai_dfid; /* fid used for data access */
2550 struct hsm_extent hai_extent; /* byte range to operate on */
2551 __u64 hai_cookie; /* action cookie from coordinator */
2552 __u64 hai_gid; /* grouplock id */
2553 char hai_data[0]; /* variable length */
2554 } __attribute__((packed));
2557 * helper function which print in hexa the first bytes of
2560 * \param hai [IN] record to print
2561 * \param buffer [IN,OUT] buffer to write the hex string to
2562 * \param len [IN] max buffer length
2566 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2567 char *buffer, __kernel_size_t len)
2574 data_len = hai->hai_len - sizeof(*hai);
2575 for (i = 0; (i < data_len) && (len > 2); i++) {
2576 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2587 /* Register even when the CDT is shutdown or being initialized */
2588 HAL_CDT_FORCE = 1 << 1,
2591 /* Copytool action list */
2592 #define HAL_VERSION 1
2593 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2594 struct hsm_action_list {
2596 __u32 hal_count; /* number of hai's to follow */
2597 __u64 hal_compound_id; /* returned by coordinator, ignored */
2599 __u32 hal_archive_id; /* which archive backend */
2601 char hal_fsname[0]; /* null-terminated */
2602 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2603 boundaries. See hai_zero */
2604 } __attribute__((packed));
2606 /* Return pointer to first hai in action list */
2607 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2609 __kernel_size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2611 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2614 /* Return pointer to next hai */
2615 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2617 __kernel_size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2619 return (struct hsm_action_item *)((char *)hai + offset);
2622 /* Return size of an hsm_action_list */
2623 static inline __kernel_size_t hal_size(struct hsm_action_list *hal)
2627 struct hsm_action_item *hai;
2629 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2630 hai = hai_first(hal);
2631 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2632 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2638 * describe the attributes to be set on imported file
2640 struct hsm_user_import {
2649 __u32 hui_archive_id;
2652 /* Copytool progress reporting */
2653 #define HP_FLAG_COMPLETED 0x01
2654 #define HP_FLAG_RETRY 0x02
2656 struct hsm_progress {
2657 struct lu_fid hp_fid;
2659 struct hsm_extent hp_extent;
2661 __u16 hp_errval; /* positive val */
2666 __u64 hc_data_version;
2668 __u16 hc_errval; /* positive val */
2670 struct hsm_action_item hc_hai;
2673 enum lu_ladvise_type {
2674 LU_LADVISE_INVALID = 0,
2675 LU_LADVISE_WILLREAD = 1,
2676 LU_LADVISE_DONTNEED = 2,
2677 LU_LADVISE_LOCKNOEXPAND = 3,
2678 LU_LADVISE_LOCKAHEAD = 4,
2679 /* Ahead operations for open|create|stat|read|write. */
2680 LU_LADVISE_AHEAD = 5,
2684 #define LU_LADVISE_NAMES { \
2685 [LU_LADVISE_WILLREAD] = "willread", \
2686 [LU_LADVISE_DONTNEED] = "dontneed", \
2687 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2688 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2689 [LU_LADVISE_AHEAD] = "ahead", \
2692 /* This is the userspace argument for ladvise. It is currently the same as
2693 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2694 * may need info which is only used locally. */
2695 struct llapi_lu_ladvise {
2696 __u16 lla_advice; /* advice type */
2697 __u16 lla_value1; /* values for different advice types */
2699 __u64 lla_start; /* first byte of extent for advice */
2700 __u64 lla_end; /* last byte of extent for advice */
2705 struct llapi_lu_ladvise2 {
2706 __u16 lla_advice; /* advice type */
2707 __u16 lla_value1; /* values for different advice types */
2718 char lla_buf[NAME_MAX + 1];
2722 /* I/O call sequences in a batch access. */
2723 enum lu_access_flags {
2724 ACCESS_FL_NONE = 0x0,
2725 ACCESS_FL_STAT = 0x01,
2726 ACCESS_FL_OPEN = 0x02,
2727 ACCESS_FL_CREAT = 0x04,
2728 ACCESS_FL_READ = 0x08,
2729 ACCESS_FL_WRITE = 0x10,
2730 ACCESS_FL_OC = ACCESS_FL_OPEN | ACCESS_FL_CREAT,
2731 ACCESS_FL_SOR = ACCESS_FL_STAT | ACCESS_FL_OPEN | ACCESS_FL_READ,
2732 ACCESS_FL_OCW = ACCESS_FL_OPEN | ACCESS_FL_CREAT | ACCESS_FL_WRITE,
2735 enum lu_ahead_mode {
2736 LU_AH_MODE_NONE = 0,
2738 * The batch access pattern obeys certain naming rules, such as mdtest
2739 * with the file naming format mdtest.$rank.$i.
2741 LU_AH_NAME_INDEX = 1,
2743 * Provide a file name list as input to do batch accesses with
2744 * irregular file name format.
2746 LU_AH_NAME_ARRAY = 2,
2747 /* Prefetching in readdir() order under a directory. */
2748 LU_AH_NAME_READDIR = 3,
2752 #define lla_ahead_mode lla_value1
2753 #define lla_access_flags lla_value2
2754 #define lla_batch_max lla_value3
2755 #define lla_fname lla_buf
2758 LF_ASYNC = 0x00000001,
2759 LF_UNSET = 0x00000002,
2762 #define LADVISE_MAGIC 0x1ADF1CE0
2763 /* Masks of valid flags for each advice */
2764 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2765 /* Flags valid for all advices not explicitly specified */
2766 #define LF_DEFAULT_MASK LF_ASYNC
2768 #define LF_MASK (LF_ASYNC | LF_UNSET)
2770 #define lla_lockahead_mode lla_value1
2771 #define lla_peradvice_flags lla_value2
2772 #define lla_lockahead_result lla_value3
2774 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2775 * is used on the wire. It is defined separately as we may need info which is
2776 * only used locally. */
2777 struct llapi_ladvise_hdr {
2778 __u32 lah_magic; /* LADVISE_MAGIC */
2779 __u32 lah_count; /* number of advices */
2780 __u64 lah_flags; /* from enum ladvise_flag */
2781 __u32 lah_value1; /* unused */
2782 __u32 lah_value2; /* unused */
2783 __u64 lah_value3; /* unused */
2784 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2787 #define LAH_COUNT_MAX (1024)
2792 SK_CRYPT_AES256_CTR = 1,
2793 SK_CRYPT_INVALID = __UINT16_MAX__
2800 SK_HMAC_INVALID = __UINT16_MAX__
2803 struct sk_crypt_type {
2804 const char *sct_name;
2808 struct sk_hmac_type {
2809 const char *sht_name;
2813 enum lock_mode_user {
2819 #define LOCK_MODE_NAMES { \
2820 [MODE_READ_USER] = "READ",\
2821 [MODE_WRITE_USER] = "WRITE"\
2824 enum lockahead_results {
2825 LLA_RESULT_SENT = 0,
2826 LLA_RESULT_DIFFERENT,
2830 enum lu_heat_flag_bit {
2831 LU_HEAT_FLAG_BIT_INVALID = 0,
2832 LU_HEAT_FLAG_BIT_OFF,
2833 LU_HEAT_FLAG_BIT_CLEAR,
2837 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2838 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2841 enum obd_heat_type {
2842 OBD_HEAT_READSAMPLE = 0,
2843 OBD_HEAT_WRITESAMPLE = 1,
2844 OBD_HEAT_READBYTE = 2,
2845 OBD_HEAT_WRITEBYTE = 3,
2849 #define LU_HEAT_NAMES { \
2850 [OBD_HEAT_READSAMPLE] = "readsample", \
2851 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2852 [OBD_HEAT_READBYTE] = "readbyte", \
2853 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2864 LU_PCC_READWRITE = 0x01,
2865 LU_PCC_READONLY = 0x02,
2866 LU_PCC_TYPE_MASK = LU_PCC_READWRITE | LU_PCC_READONLY,
2867 LU_PCC_FL_ASYNC = 0x10,
2871 static inline const char *pcc_type2string(enum lu_pcc_type type)
2873 switch (type & LU_PCC_TYPE_MASK) {
2876 case LU_PCC_READWRITE:
2878 case LU_PCC_READONLY:
2885 struct lu_pcc_attach {
2886 __u32 pcca_type; /* PCC type */
2887 __u32 pcca_id; /* Attach ID */
2890 enum lu_pcc_detach_opts {
2891 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2892 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2895 struct lu_pcc_detach_fid {
2896 /* fid of the file to detach */
2897 struct lu_fid pccd_fid;
2901 struct lu_pcc_detach {
2905 enum lu_pcc_state_flags {
2906 PCC_STATE_FL_NONE = 0x0,
2907 /* The inode attr is cached locally */
2908 PCC_STATE_FL_ATTR_VALID = 0x01,
2909 /* The file is being attached into PCC */
2910 PCC_STATE_FL_ATTACHING = 0x02,
2911 /* The PCC copy is unlinked */
2912 PCC_STATE_FL_UNLINKED = 0x04,
2915 struct lu_pcc_state {
2916 __u32 pccs_type; /* enum lu_pcc_type */
2917 __u32 pccs_open_count;
2918 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2920 char pccs_path[PATH_MAX];
2923 enum lu_project_type {
2924 LU_PROJECT_NONE = 0,
2931 __u32 project_type; /* enum lu_project_type */
2933 __u32 project_xflags;
2934 __u32 project_reserved;
2935 char project_name[NAME_MAX + 1];
2940 /* make header's size equal lu_fid */
2943 struct lu_fid fa_fids[0];
2945 #define OBD_MAX_FIDS_IN_ARRAY 4096
2947 /* more types could be defined upon need for more complex
2948 * format to be used in foreign symlink LOV/LMV EAs, like
2949 * one to describe a delimiter string and occurence number
2950 * of delimited sub-string, ...
2952 enum ll_foreign_symlink_upcall_item_type {
2958 /* may need to be modified to allow for more format items to be defined, and
2959 * like for ll_foreign_symlink_upcall_item_type enum
2961 struct ll_foreign_symlink_upcall_item {
2971 /* internal storage of constant string */
2973 /* upcall stores constant string in a raw */
2980 #define POSLEN_ITEM_SZ (offsetof(struct ll_foreign_symlink_upcall_item, len) + \
2981 sizeof(((struct ll_foreign_symlink_upcall_item *)0)->len))
2982 #define STRING_ITEM_SZ(sz) ( \
2983 offsetof(struct ll_foreign_symlink_upcall_item, bytestring) + \
2984 (sz + sizeof(__u32) - 1) / sizeof(__u32) * sizeof(__u32))
2986 /* presently limited to not cause max stack frame size to be reached
2987 * because of temporary automatic array of
2988 * "struct ll_foreign_symlink_upcall_item" presently used in
2989 * foreign_symlink_upcall_info_store()
2991 #define MAX_NB_UPCALL_ITEMS 32
2993 #if defined(__cplusplus)
2997 /** @} lustreuser */
2999 #endif /* _LUSTRE_USER_H */