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 */
242 #define AT_STATX_SYNC_TYPE 0x6000 /* Type of synchronisation required from statx() */
243 #define AT_STATX_SYNC_AS_STAT 0x0000 /* - Do whatever stat() does */
244 #define AT_STATX_FORCE_SYNC 0x2000 /* - Force the attributes to be sync'd with the server */
245 #define AT_STATX_DONT_SYNC 0x4000 /* - Don't sync attributes with the server */
247 #endif /* STATX_BASIC_STATS */
249 typedef struct statx lstatx_t;
251 #define LUSTRE_EOF 0xffffffffffffffffULL
254 #define LL_SUPER_MAGIC 0x0BD00BD0
256 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
258 /* FIEMAP flags supported by Lustre */
259 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
261 enum obd_statfs_state {
262 OS_STATFS_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
263 OS_STATFS_READONLY = 0x00000002, /**< filesystem is read-only */
264 OS_STATFS_NOPRECREATE = 0x00000004, /**< no object precreation */
265 OS_STATFS_UNUSED1 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
266 OS_STATFS_UNUSED2 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
267 OS_STATFS_ENOSPC = 0x00000020, /**< not enough free space */
268 OS_STATFS_ENOINO = 0x00000040, /**< not enough inodes */
269 OS_STATFS_SUM = 0x00000100, /**< aggregated for all tagrets */
270 OS_STATFS_NONROT = 0x00000200, /**< non-rotational device */
273 /** filesystem statistics/attributes for target device */
275 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
276 __u64 os_blocks; /* total size in #os_bsize blocks */
277 __u64 os_bfree; /* number of unused blocks */
278 __u64 os_bavail; /* blocks available for allocation */
279 __u64 os_files; /* total number of objects */
280 __u64 os_ffree; /* # objects that could be created */
281 __u8 os_fsid[40]; /* identifier for filesystem */
282 __u32 os_bsize; /* block size in bytes for os_blocks */
283 __u32 os_namelen; /* maximum length of filename in bytes*/
284 __u64 os_maxbytes; /* maximum object size in bytes */
285 __u32 os_state; /**< obd_statfs_state OS_STATFS_* */
286 __u32 os_fprecreated; /* objs available now to the caller */
287 /* used in QoS code to find preferred
289 __u32 os_granted; /* space granted for MDS */
290 __u32 os_spare3; /* Unused padding fields. Remember */
291 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
299 /** additional filesystem attributes for target device */
300 struct obd_statfs_info {
301 __u32 os_reserved_mb_low; /* reserved mb low */
302 __u32 os_reserved_mb_high; /* reserved mb high */
303 bool os_enable_pre; /* enable pre create logic */
309 * FID is a cluster-wide unique identifier of a file or an object (stripe).
310 * FIDs are never reused.
314 * FID sequence. Sequence is a unit of migration: all files (objects)
315 * with FIDs from a given sequence are stored on the same server.
316 * Lustre should support 2^64 objects, so even if each sequence
317 * has only a single object we can still enumerate 2^64 objects.
320 /* FID number within sequence. */
323 * FID version, used to distinguish different versions (in the sense
324 * of snapshots, etc.) of the same file system object. Not currently
328 } __attribute__((packed));
330 static inline bool fid_is_zero(const struct lu_fid *fid)
332 return fid->f_seq == 0 && fid->f_oid == 0;
335 /* The data name_to_handle_at() places in a struct file_handle (at f_handle) */
336 struct lustre_file_handle {
337 struct lu_fid lfh_child;
338 struct lu_fid lfh_parent;
341 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
342 * MDT-object's FID::f_ver, instead it is the OST-object index in its
343 * parent MDT-object's layout EA. */
344 #define f_stripe_idx f_ver
347 __u32 ol_stripe_size;
348 __u32 ol_stripe_count;
352 } __attribute__((packed));
354 /* The filter_fid structure has changed several times over its lifetime.
355 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
356 * stripe_index and the "self FID" (objid/seq) to be able to recover the
357 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
358 * the OST, the "trusted.lma" xattr was added to the OST objects to store
359 * the "self FID" to be consistent with the MDT on-disk format, and the
360 * filter_fid only stored the MDT inode parent FID and stripe index.
362 * In 2.10, the addition of PFL composite layouts required more information
363 * to be stored into the filter_fid in order to be able to identify which
364 * component the OST object belonged. As well, the stripe size may vary
365 * between components, so it was no longer safe to assume the stripe size
366 * or stripe_count of a file. This is also more robust for plain layouts.
368 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
369 * enough space to store both the filter_fid and LMA in the inode, so they
370 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
371 * an extra seek for every OST object access.
373 * In 2.11, FLR mirror layouts also need to store the layout version and
374 * range so that writes to old versions of the layout are not allowed.
375 * That ensures that mirrored objects are not modified by evicted clients,
376 * and ensures that the components are correctly marked stale on the MDT.
378 struct filter_fid_18_23 {
379 struct lu_fid ff_parent; /* stripe_idx in f_ver */
384 struct filter_fid_24_29 {
385 struct lu_fid ff_parent; /* stripe_idx in f_ver */
388 struct filter_fid_210 {
389 struct lu_fid ff_parent; /* stripe_idx in f_ver */
390 struct ost_layout ff_layout;
394 struct lu_fid ff_parent; /* stripe_idx in f_ver */
395 struct ost_layout ff_layout;
396 __u32 ff_layout_version;
397 __u32 ff_range; /* range of layout version that
398 * write are allowed */
399 } __attribute__((packed));
401 /* Userspace should treat lu_fid as opaque, and only use the following methods
402 * to print or parse them. Other functions (e.g. compare, swab) could be moved
403 * here from lustre_idl.h if needed. */
407 LMAC_HSM = 0x00000001,
408 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
409 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
410 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
411 * under /O/<seq>/d<x>. */
412 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
413 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
414 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
418 * Masks for all features that should be supported by a Lustre version to
419 * access a specific file.
420 * This information is stored in lustre_mdt_attrs::lma_incompat.
423 LMAI_RELEASED = 0x00000001, /* file is released */
424 LMAI_AGENT = 0x00000002, /* agent inode */
425 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
426 is on the remote MDT */
427 LMAI_STRIPED = 0x00000008, /* striped directory inode */
428 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
429 LMAI_ENCRYPT = 0x00000020, /* inode is encrypted */
430 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
431 LMAI_STRIPED | LMAI_ORPHAN | LMAI_ENCRYPT)
436 * Following struct for object attributes, that will be kept inode's EA.
437 * Introduced in 2.0 release (please see b15993, for details)
438 * Added to all objects since Lustre 2.4 as contains self FID
440 struct lustre_mdt_attrs {
442 * Bitfield for supported data in this structure. From enum lma_compat.
443 * lma_self_fid and lma_flags are always available.
447 * Per-file incompat feature list. Lustre version should support all
448 * flags set in this field. The supported feature mask is available in
452 /** FID of this inode */
453 struct lu_fid lma_self_fid;
456 struct lustre_ost_attrs {
457 /* Use lustre_mdt_attrs directly for now, need a common header
458 * structure if want to change lustre_mdt_attrs in future. */
459 struct lustre_mdt_attrs loa_lma;
461 /* Below five elements are for OST-object's PFID EA, the
462 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
463 * and the stripe_index (low 16 bits), the size should not exceed
464 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
465 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
466 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
467 * loa_comp_* elements are valid. */
468 struct lu_fid loa_parent_fid;
469 __u32 loa_stripe_size;
471 __u64 loa_comp_start;
476 * Prior to 2.4, the LMA structure also included SOM attributes which has since
477 * been moved to a dedicated xattr
478 * lma_flags was also removed because of lma_compat/incompat fields.
480 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
482 enum lustre_som_flags {
483 /* Unknow or no SoM data, must get size from OSTs. */
484 SOM_FL_UNKNOWN = 0x0000,
485 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
486 SOM_FL_STRICT = 0x0001,
487 /* Known stale - was right at some point in the past, but it is
488 * known (or likely) to be incorrect now (e.g. opened for write). */
489 SOM_FL_STALE = 0x0002,
490 /* Approximate, may never have been strictly correct,
491 * need to sync SOM data to achieve eventual consistency. */
492 SOM_FL_LAZY = 0x0004,
495 struct lustre_som_attrs {
497 __u16 lsa_reserved[3];
503 * OST object IDentifier.
511 struct lu_fid oi_fid;
513 } __attribute__((packed));
515 #define DOSTID "%#llx:%llu"
516 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
517 ((unsigned long long)ostid_id(oi))
519 struct ll_futimes_3 {
521 __u64 lfu_atime_nsec;
523 __u64 lfu_mtime_nsec;
525 __u64 lfu_ctime_nsec;
529 * Maximum number of mirrors currently implemented.
531 #define LUSTRE_MIRROR_COUNT_MAX 16
533 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
535 LL_LEASE_RDLCK = 0x01,
536 LL_LEASE_WRLCK = 0x02,
537 LL_LEASE_UNLCK = 0x04,
540 enum ll_lease_flags {
541 LL_LEASE_RESYNC = 0x1,
542 LL_LEASE_RESYNC_DONE = 0x2,
543 LL_LEASE_LAYOUT_MERGE = 0x4,
544 LL_LEASE_LAYOUT_SPLIT = 0x8,
545 LL_LEASE_PCC_ATTACH = 0x10,
548 #define IOC_IDS_MAX 4096
549 struct ll_ioc_lease {
556 struct ll_ioc_lease_id {
567 * The ioctl naming rules:
568 * LL_* - works on the currently opened filehandle instead of parent dir
569 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
570 * *_MDC_* - gets/sets data related to MDC
571 * *_LOV_* - gets/sets data related to OSC/LOV
572 * *FILE* - called on parent dir and passes in a filename
573 * *STRIPE* - set/get lov_user_md
574 * *INFO - set/get lov_user_mds_data
576 /* lustre_ioctl.h 101-150 */
577 /* ioctl codes 128-143 are reserved for fsverity */
578 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
579 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
580 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
581 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
582 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
583 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
584 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
585 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
586 /* LL_IOC_RECREATE_OBJ 157 obsolete */
587 /* LL_IOC_RECREATE_FID 157 obsolete */
588 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
589 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
590 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
591 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
592 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
593 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
594 /* IOC_LOV_GETINFO 165 obsolete */
595 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
596 /* LL_IOC_RMTACL 167 obsolete */
597 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
598 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
599 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
600 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
601 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
602 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
603 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
604 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
605 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
606 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
607 /* lustre_ioctl.h 177-210 */
608 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
609 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
610 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
611 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
612 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
613 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
614 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
615 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
616 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
617 struct lustre_swap_layouts)
618 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
619 struct hsm_current_action)
620 /* lustre_ioctl.h 221-232 */
621 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
622 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
623 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
624 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
625 #define LL_IOC_UNLOCK_FOREIGN _IO('f', 242)
626 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
627 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
628 #define LL_IOC_GET_LEASE _IO('f', 244)
629 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
630 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
631 #define LL_IOC_MIGRATE _IOR('f', 247, int)
632 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
633 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
634 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
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_DETACH _IOW('f', 252, struct lu_pcc_detach)
638 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
639 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
640 #define LL_IOC_PROJECT _IOW('f', 253, struct lu_project)
642 #ifndef FS_IOC_FSGETXATTR
644 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
647 __u32 fsx_xflags; /* xflags field value (get/set) */
648 __u32 fsx_extsize; /* extsize field value (get/set)*/
649 __u32 fsx_nextents; /* nextents field value (get) */
650 __u32 fsx_projid; /* project identifier (get/set) */
651 unsigned char fsx_pad[12];
653 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
654 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
656 #ifndef FS_XFLAG_PROJINHERIT
657 #define FS_XFLAG_PROJINHERIT 0x00000200
661 #define LL_STATFS_LMV 1
662 #define LL_STATFS_LOV 2
663 #define LL_STATFS_NODELAY 4
665 #define IOC_MDC_TYPE 'i'
666 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
667 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
668 #define IOC_MDC_GETFILEINFO_V1 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
669 #define IOC_MDC_GETFILEINFO_V2 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
670 #define LL_IOC_MDC_GETINFO_V1 _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
671 #define LL_IOC_MDC_GETINFO_V2 _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
672 #define IOC_MDC_GETFILEINFO IOC_MDC_GETFILEINFO_V1
673 #define LL_IOC_MDC_GETINFO LL_IOC_MDC_GETINFO_V1
675 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
677 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
678 * files, but are unlikely to be used in practice and are not harmful if
679 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
680 * devices and are safe for use on new files. See LU-4209. */
681 /* To be compatible with old statically linked binary we keep the check for
682 * the older 0100000000 flag. This is already removed upstream. LU-812. */
683 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
685 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
687 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
688 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
689 O_LOV_DELAY_CREATE_MASK)
690 /* O_FILE_ENC principle is similar to O_LOV_DELAY_CREATE above,
691 * for access to encrypted files without the encryption key.
693 #define O_FILE_ENC (O_NOCTTY | O_NDELAY)
695 #define LL_FILE_IGNORE_LOCK 0x00000001
696 #define LL_FILE_GROUP_LOCKED 0x00000002
697 #define LL_FILE_READAHEA 0x00000004
698 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
699 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
701 #define LOV_USER_MAGIC_V1 0x0BD10BD0
702 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
703 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
704 #define LOV_USER_MAGIC_V3 0x0BD30BD0
705 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
706 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
707 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
708 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
709 #define LOV_USER_MAGIC_SEL 0x0BD80BD0
711 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
712 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
713 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
715 #define LOV_PATTERN_NONE 0x000
716 #define LOV_PATTERN_RAID0 0x001
717 #define LOV_PATTERN_RAID1 0x002
718 #define LOV_PATTERN_MDT 0x100
719 #define LOV_PATTERN_OVERSTRIPING 0x200
720 #define LOV_PATTERN_FOREIGN 0x400
722 #define LOV_PATTERN_F_MASK 0xffff0000
723 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
724 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
725 #define LOV_PATTERN_DEFAULT 0xffffffff
727 #define LOV_OFFSET_DEFAULT ((__u16)-1)
728 #define LMV_OFFSET_DEFAULT ((__u32)-1)
730 static inline bool lov_pattern_supported(__u32 pattern)
732 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
733 (pattern & ~LOV_PATTERN_F_RELEASED) ==
734 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
735 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
738 /* RELEASED and MDT patterns are not valid in many places, so rather than
739 * having many extra checks on lov_pattern_supported, we have this separate
740 * check for non-released, non-DOM components
742 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
744 return pattern == LOV_PATTERN_RAID0 ||
745 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
749 #define LOV_MAXPOOLNAME 15
750 #define LOV_POOLNAMEF "%.15s"
752 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
753 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
754 #define LOV_MAX_STRIPE_COUNT_OLD 160
755 /* This calculation is crafted so that input of 4096 will result in 160
756 * which in turn is equal to old maximal stripe count.
757 * XXX: In fact this is too simpified for now, what it also need is to get
758 * ea_type argument to clearly know how much space each stripe consumes.
760 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
761 * allocation that is sufficient for the current generation of systems.
763 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
764 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
765 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
766 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
768 #define XATTR_LUSTRE_PREFIX "lustre."
769 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
771 /* Please update if XATTR_LUSTRE_LOV".set" groks more flags in the future */
772 #define allowed_lustre_lov(att) (strcmp((att), XATTR_LUSTRE_LOV".add") == 0 || \
773 strcmp((att), XATTR_LUSTRE_LOV".set") == 0 || \
774 strcmp((att), XATTR_LUSTRE_LOV".set.flags") == 0 || \
775 strcmp((att), XATTR_LUSTRE_LOV".del") == 0)
777 #define lov_user_ost_data lov_user_ost_data_v1
778 struct lov_user_ost_data_v1 { /* per-stripe data structure */
779 struct ost_id l_ost_oi; /* OST object ID */
780 __u32 l_ost_gen; /* generation of this OST index */
781 __u32 l_ost_idx; /* OST index in LOV */
782 } __attribute__((packed));
784 #define lov_user_md lov_user_md_v1
785 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
786 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
787 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
788 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
789 __u32 lmm_stripe_size; /* size of stripe in bytes */
790 __u16 lmm_stripe_count; /* num stripes in use for this object */
792 __u16 lmm_stripe_offset; /* starting stripe offset in
793 * lmm_objects, use when writing */
794 __u16 lmm_layout_gen; /* layout generation number
795 * used when reading */
797 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
798 } __attribute__((packed, __may_alias__));
800 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
801 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
802 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
803 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
804 __u32 lmm_stripe_size; /* size of stripe in bytes */
805 __u16 lmm_stripe_count; /* num stripes in use for this object */
807 __u16 lmm_stripe_offset; /* starting stripe offset in
808 * lmm_objects, use when writing */
809 __u16 lmm_layout_gen; /* layout generation number
810 * used when reading */
812 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
813 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
814 } __attribute__((packed, __may_alias__));
816 struct lov_foreign_md {
817 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
818 __u32 lfm_length; /* length of lfm_value */
819 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
820 __u32 lfm_flags; /* flags, type specific */
822 } __attribute__((packed));
824 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
825 offsetof(struct lov_foreign_md, lfm_value))
827 #define foreign_size_le(lfm) \
828 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
829 offsetof(struct lov_foreign_md, lfm_value))
832 * The stripe size fields are shared for the extension size storage, however
833 * the extension size is stored in KB, not bytes.
835 #define SEL_UNIT_SIZE 1024llu
840 } __attribute__((packed));
842 #define DEXT "[%#llx, %#llx)"
843 #define PEXT(ext) (unsigned long long)(ext)->e_start, (unsigned long long)(ext)->e_end
845 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
846 struct lu_extent *e2)
848 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
851 static inline bool lu_extent_is_whole(struct lu_extent *e)
853 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
856 enum lov_comp_md_entry_flags {
857 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
858 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
859 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
860 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
861 LCME_FL_OFFLINE = 0x00000008, /* Not used */
862 LCME_FL_INIT = 0x00000010, /* instantiated */
863 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
864 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
865 LCME_FL_PARITY = 0x00000080, /* EC: a parity code component */
866 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
867 * won't be stored on disk
871 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
872 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
875 /* The component flags can be set by users at creation/modification time. */
876 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
879 /* The mirror flags can be set by users at creation time. */
880 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
882 /* The allowed flags obtained from the client at component creation time. */
883 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
885 /* The mirror flags sent by client */
886 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
888 /* These flags have meaning when set in a default layout and will be inherited
889 * from the default/template layout set on a directory.
891 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
894 /* lcme_id can be specified as certain flags, and the the first
895 * bit of lcme_id is used to indicate that the ID is representing
896 * certain LCME_FL_* but not a real ID. Which implies we can have
897 * at most 31 flags (see LCME_FL_XXX). */
900 LCME_ID_MAX = 0x7FFFFFFF,
901 LCME_ID_ALL = 0xFFFFFFFF,
902 LCME_ID_NOT_ID = LCME_FL_NEG
905 /* layout version equals to lcme_id, except some bits have special meanings */
906 enum layout_version_flags {
907 /* layout version reaches the high water mark to be increased to
908 * circularly reuse the smallest value
910 LU_LAYOUT_HIGEN = 0x40000000,
911 /* the highest bit is used to mark if the file is being resynced */
912 LU_LAYOUT_RESYNC = 0x80000000,
915 #define LCME_ID_MASK LCME_ID_MAX
917 struct lov_comp_md_entry_v1 {
918 __u32 lcme_id; /* unique id of component */
919 __u32 lcme_flags; /* LCME_FL_XXX */
920 /* file extent for component. If it's an EC code component, its flags
921 * contains LCME_FL_PARITY, and its extent covers the same extent of
922 * its corresponding data component.
924 struct lu_extent lcme_extent;
925 __u32 lcme_offset; /* offset of component blob,
926 start from lov_comp_md_v1 */
927 __u32 lcme_size; /* size of component blob */
928 __u32 lcme_layout_gen;
929 __u64 lcme_timestamp; /* snapshot time if applicable*/
930 __u8 lcme_dstripe_count; /* data stripe count,
933 __u8 lcme_cstripe_count; /* code stripe count,
936 __u16 lcme_padding_1;
937 } __attribute__((packed));
939 #define SEQ_ID_MAX 0x0000FFFF
940 #define SEQ_ID_MASK SEQ_ID_MAX
941 /* bit 30:16 of lcme_id is used to store mirror id */
942 #define MIRROR_ID_MASK 0x7FFF0000
943 #define MIRROR_ID_NEG 0x8000
944 #define MIRROR_ID_SHIFT 16
946 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
948 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
951 static inline __u16 mirror_id_of(__u32 id)
953 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
957 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
959 enum lov_comp_md_flags {
960 /* the least 4 bits are used by FLR to record file state */
963 LCM_FL_WRITE_PENDING = 0x2,
964 LCM_FL_SYNC_PENDING = 0x3,
965 LCM_FL_PCC_RDONLY = 0x8,
966 LCM_FL_FLR_MASK = 0xB,
969 struct lov_comp_md_v1 {
970 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
971 __u32 lcm_size; /* overall size including this struct */
972 __u32 lcm_layout_gen;
974 __u16 lcm_entry_count;
975 /* lcm_mirror_count stores the number of actual mirrors minus 1,
976 * so that non-flr files will have value 0 meaning 1 mirror. */
977 __u16 lcm_mirror_count;
978 /* code components count, non-EC file contains 0 ec_count */
980 __u8 lcm_padding3[1];
981 __u16 lcm_padding1[2];
983 struct lov_comp_md_entry_v1 lcm_entries[0];
984 } __attribute__((packed));
986 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
988 if (stripes == (__u16)-1)
991 if (lmm_magic == LOV_USER_MAGIC_V1)
992 return sizeof(struct lov_user_md_v1) +
993 stripes * sizeof(struct lov_user_ost_data_v1);
994 return sizeof(struct lov_user_md_v3) +
995 stripes * sizeof(struct lov_user_ost_data_v1);
998 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
999 * use this. It is unsafe to #define those values in this header as it
1000 * is possible the application has already #included <sys/stat.h>. */
1001 #define lov_user_mds_data lov_user_mds_data_v2
1002 struct lov_user_mds_data_v1 {
1003 lstat_t lmd_st; /* MDS stat struct */
1004 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
1005 } __attribute__((packed));
1007 struct lov_user_mds_data_v2 {
1008 struct lu_fid lmd_fid; /* Lustre FID */
1009 lstatx_t lmd_stx; /* MDS statx struct */
1010 __u64 lmd_flags; /* MDS stat flags */
1011 __u32 lmd_lmmsize; /* LOV EA size */
1012 __u32 lmd_padding; /* unused */
1013 struct lov_user_md_v1 lmd_lmm; /* LOV EA user data */
1014 } __attribute__((packed));
1016 struct lmv_user_mds_data {
1017 struct lu_fid lum_fid;
1020 } __attribute__((packed, __may_alias__));
1022 enum lmv_hash_type {
1023 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
1024 LMV_HASH_TYPE_ALL_CHARS = 1, /* simple sum of characters */
1025 LMV_HASH_TYPE_FNV_1A_64 = 2, /* reasonable non-cryptographic hash */
1026 LMV_HASH_TYPE_CRUSH = 3, /* double-hash to optimize migration */
1027 LMV_HASH_TYPE_CRUSH2 = 4, /* CRUSH with small fixes, LU-15692 */
1029 LMV_HASH_TYPE_DEFAULT = LMV_HASH_TYPE_FNV_1A_64
1032 static __attribute__((unused)) const char *mdt_hash_name[] = {
1041 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1042 * and the higher part will be the flag to indicate the status of object,
1043 * for example the object is being migrated. And the hash function
1044 * might be interpreted differently with different flags. */
1045 #define LMV_HASH_TYPE_MASK 0x0000ffff
1047 static inline bool lmv_is_known_hash_type(__u32 type)
1049 return (type & LMV_HASH_TYPE_MASK) > LMV_HASH_TYPE_UNKNOWN &&
1050 (type & LMV_HASH_TYPE_MASK) < LMV_HASH_TYPE_MAX;
1053 /* fixed layout, such directories won't split automatically */
1054 /* NB, update LMV_HASH_FLAG_KNOWN when adding new flag */
1055 #define LMV_HASH_FLAG_FIXED 0x02000000
1056 #define LMV_HASH_FLAG_MERGE 0x04000000
1057 #define LMV_HASH_FLAG_SPLIT 0x08000000
1059 /* The striped directory has ever lost its master LMV EA, then LFSCK
1060 * re-generated it. This flag is used to indicate such case. It is an
1062 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
1064 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1065 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1067 #define LMV_HASH_FLAG_LAYOUT_CHANGE \
1068 (LMV_HASH_FLAG_MIGRATION | LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MERGE)
1070 #define LMV_HASH_FLAG_KNOWN 0xbe000000
1072 /* both SPLIT and MIGRATION are set for directory split */
1073 static inline bool lmv_hash_is_splitting(__u32 hash)
1075 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1076 (LMV_HASH_FLAG_SPLIT | LMV_HASH_FLAG_MIGRATION);
1079 /* both MERGE and MIGRATION are set for directory merge */
1080 static inline bool lmv_hash_is_merging(__u32 hash)
1082 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) ==
1083 (LMV_HASH_FLAG_MERGE | LMV_HASH_FLAG_MIGRATION);
1086 /* only MIGRATION is set for directory migration */
1087 static inline bool lmv_hash_is_migrating(__u32 hash)
1089 return (hash & LMV_HASH_FLAG_LAYOUT_CHANGE) == LMV_HASH_FLAG_MIGRATION;
1092 static inline bool lmv_hash_is_restriping(__u32 hash)
1094 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash);
1097 static inline bool lmv_hash_is_layout_changing(__u32 hash)
1099 return lmv_hash_is_splitting(hash) || lmv_hash_is_merging(hash) ||
1100 lmv_hash_is_migrating(hash);
1103 struct lustre_foreign_type {
1105 const char *lft_name;
1109 * LOV/LMV foreign types
1111 enum lustre_foreign_types {
1112 LU_FOREIGN_TYPE_NONE = 0,
1113 LU_FOREIGN_TYPE_SYMLINK = 0xda05,
1114 /* must be the max/last one */
1115 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
1118 extern struct lustre_foreign_type lu_foreign_types[];
1120 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
1121 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
1122 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
1123 #define lmv_user_md lmv_user_md_v1
1124 struct lmv_user_md_v1 {
1125 __u32 lum_magic; /* must be the first field */
1126 __u32 lum_stripe_count; /* dirstripe count */
1127 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
1128 __u32 lum_hash_type; /* Dir stripe policy */
1129 __u32 lum_type; /* LMV type: default */
1130 __u8 lum_max_inherit; /* inherit depth of default LMV */
1131 __u8 lum_max_inherit_rr; /* inherit depth of default LMV to round-robin mkdir */
1135 char lum_pool_name[LOV_MAXPOOLNAME + 1];
1136 struct lmv_user_mds_data lum_objects[0];
1137 } __attribute__((packed));
1139 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
1141 if (size <= sizeof(struct lmv_user_md))
1144 size -= sizeof(struct lmv_user_md);
1145 return (size + sizeof(struct lmv_user_mds_data) - 1) /
1146 sizeof(struct lmv_user_mds_data);
1150 * NB, historically default layout didn't set type, but use XATTR name to differ
1151 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
1152 * and still use the same method.
1155 LMV_TYPE_DEFAULT = 0x0000,
1158 /* lum_max_inherit will be decreased by 1 after each inheritance if it's not
1159 * LMV_INHERIT_UNLIMITED or > LMV_INHERIT_MAX.
1162 /* for historical reason, 0 means unlimited inheritance */
1163 LMV_INHERIT_UNLIMITED = 0,
1164 /* unlimited lum_max_inherit by default for plain stripe (0 or 1) */
1165 LMV_INHERIT_DEFAULT_PLAIN = LMV_INHERIT_UNLIMITED,
1166 /* not inherit any more */
1167 LMV_INHERIT_END = 1,
1168 /* for multiple stripes, the default lum_max_inherit is 3 */
1169 LMV_INHERIT_DEFAULT_STRIPED = 3,
1170 /* max inherit depth */
1171 LMV_INHERIT_MAX = 250,
1172 /* [251, 254] are reserved */
1173 /* not set, or when inherit depth goes beyond end, */
1174 LMV_INHERIT_NONE = 255,
1178 /* not set, or when inherit_rr depth goes beyond end, */
1179 LMV_INHERIT_RR_NONE = 0,
1180 /* disable lum_max_inherit_rr by default */
1181 LMV_INHERIT_RR_DEFAULT = LMV_INHERIT_RR_NONE,
1182 /* not inherit any more */
1183 LMV_INHERIT_RR_END = 1,
1184 /* default inherit_rr of ROOT */
1185 LMV_INHERIT_RR_ROOT = 3,
1186 /* max inherit depth */
1187 LMV_INHERIT_RR_MAX = 250,
1188 /* [251, 254] are reserved */
1189 /* unlimited inheritance */
1190 LMV_INHERIT_RR_UNLIMITED = 255,
1193 static inline int lmv_user_md_size(int stripes, int lmm_magic)
1195 int size = sizeof(struct lmv_user_md);
1197 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
1198 size += stripes * sizeof(struct lmv_user_mds_data);
1203 struct ll_recreate_obj {
1209 __u64 id; /* holds object id */
1210 __u32 generation; /* holds object generation */
1211 __u32 f_type; /* holds object type or stripe idx when passing it to
1212 * OST for saving into EA. */
1217 char uuid[UUID_MAX];
1220 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
1221 const struct obd_uuid *u2)
1223 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
1226 static inline int obd_uuid_empty(struct obd_uuid *uuid)
1228 return uuid->uuid[0] == '\0';
1231 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
1233 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
1234 uuid->uuid[sizeof(*uuid) - 1] = '\0';
1237 /* For printf's only, make sure uuid is terminated */
1238 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
1243 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
1244 /* Obviously not safe, but for printfs, no real harm done...
1245 we're always null-terminated, even in a race. */
1246 static char temp[sizeof(*uuid->uuid)];
1248 memcpy(temp, uuid->uuid, sizeof(*uuid->uuid) - 1);
1249 temp[sizeof(*uuid->uuid) - 1] = '\0';
1253 return (char *)(uuid->uuid);
1256 #define LUSTRE_MAXFSNAME 8
1257 #define LUSTRE_MAXINSTANCE 16
1259 /* Extract fsname from uuid (or target name) of a target
1260 e.g. (myfs-OST0007_UUID -> myfs)
1261 see also deuuidify. */
1262 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
1266 strncpy(buf, uuid, buflen - 1);
1267 buf[buflen - 1] = '\0';
1268 p = strrchr(buf, '-');
1273 /* printf display format for Lustre FIDs
1274 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1275 #define FID_NOBRACE_LEN 40
1276 #define FID_LEN (FID_NOBRACE_LEN + 2)
1277 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1278 #define DFID "[" DFID_NOBRACE "]"
1279 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1281 /* scanf input parse format for fids in DFID_NOBRACE format
1282 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1283 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1284 #define SFID "0x%llx:0x%x:0x%x"
1285 #define RFID(fid) (unsigned long long *)&((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1287 /********* Quotas **********/
1289 /* From linux/fs/quota/quota.c */
1290 static inline __u64 toqb(__kernel_size_t space)
1292 return (space + QIF_DQBLKSIZE - 1) >> QIF_DQBLKSIZE_BITS;
1295 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1296 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1297 #define Q_GETOINFO 0x800102 /* get obd quota info */
1298 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1299 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1301 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1302 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1303 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1304 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1305 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1306 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1307 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1308 /* lustre-specific control commands */
1309 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1310 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1311 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1312 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1313 #define LUSTRE_Q_GETQUOTAPOOL 0x80000f /* get user pool quota */
1314 #define LUSTRE_Q_SETQUOTAPOOL 0x800010 /* set user pool quota */
1315 #define LUSTRE_Q_GETINFOPOOL 0x800011 /* get pool quota info */
1316 #define LUSTRE_Q_SETINFOPOOL 0x800012 /* set pool quota info */
1317 #define LUSTRE_Q_GETDEFAULT_POOL 0x800013 /* get default pool quota*/
1318 #define LUSTRE_Q_SETDEFAULT_POOL 0x800014 /* set default pool quota */
1319 #define LUSTRE_Q_DELETEQID 0x800015 /* delete quota ID */
1320 /* In the current Lustre implementation, the grace time is either the time
1321 * or the timestamp to be used after some quota ID exceeds the soft limt,
1322 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1324 #define LQUOTA_GRACE_BITS 48
1325 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1326 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1327 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1328 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1329 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1331 /* special grace time, only notify the user when its quota is over soft limit
1332 * but doesn't block new writes until the hard limit is reached. */
1333 #define NOTIFY_GRACE "notify"
1334 #define NOTIFY_GRACE_TIME LQUOTA_GRACE_MASK
1336 /* different quota flags */
1338 /* the default quota flag, the corresponding quota ID will use the default
1339 * quota setting, the hardlimit and softlimit of its quota record in the global
1340 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1341 * and high 16 bits will contain this flag (see above comment).
1343 #define LQUOTA_FLAG_DEFAULT 0x0001
1344 #define LQUOTA_FLAG_DELETED 0x0002
1346 #define LUSTRE_Q_CMD_IS_POOL(cmd) \
1347 (cmd == LUSTRE_Q_GETQUOTAPOOL || \
1348 cmd == LUSTRE_Q_SETQUOTAPOOL || \
1349 cmd == LUSTRE_Q_SETINFOPOOL || \
1350 cmd == LUSTRE_Q_GETINFOPOOL || \
1351 cmd == LUSTRE_Q_SETDEFAULT_POOL || \
1352 cmd == LUSTRE_Q_GETDEFAULT_POOL)
1354 #define ALLQUOTA 255 /* set all quota */
1355 static inline const char *qtype_name(int qtype)
1368 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1371 #define N_PERMS_MAX 64
1373 struct perm_downcall_data {
1379 struct identity_downcall_data {
1386 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1387 __u32 idd_groups[0];
1390 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
1391 /* old interface struct is deprecated in 2.14 */
1392 #define SEPOL_DOWNCALL_MAGIC_OLD 0x8b8bb842
1393 struct sepol_downcall_data_old {
1395 __s64 sdd_sepol_mtime;
1396 __u16 sdd_sepol_len;
1401 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb843
1402 struct sepol_downcall_data {
1404 __u16 sdd_sepol_len;
1406 __s64 sdd_sepol_mtime;
1410 #ifdef NEED_QUOTA_DEFS
1412 #define QIF_BLIMITS 1
1414 #define QIF_ILIMITS 4
1415 #define QIF_INODES 8
1416 #define QIF_BTIME 16
1417 #define QIF_ITIME 32
1418 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1419 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1420 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1421 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1424 #endif /* !__KERNEL__ */
1426 /* lustre volatile file support
1427 * file name header: ".^L^S^T^R:volatile"
1429 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1430 #define LUSTRE_VOLATILE_HDR_LEN 14
1432 enum lustre_quota_version {
1436 /* XXX: same as if_dqinfo struct in kernel */
1444 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1446 __u64 dqb_bhardlimit; /* kbytes unit */
1447 __u64 dqb_bsoftlimit; /* kbytes unit */
1448 __u64 dqb_curspace; /* bytes unit */
1449 __u64 dqb_ihardlimit;
1450 __u64 dqb_isoftlimit;
1451 __u64 dqb_curinodes;
1465 struct if_quotactl {
1472 struct obd_dqinfo qc_dqinfo;
1473 struct obd_dqblk qc_dqblk;
1475 struct obd_uuid obd_uuid;
1476 char qc_poolname[0];
1479 /* swap layout flags */
1480 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1481 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1482 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1483 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1484 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1486 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1487 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1488 struct lustre_swap_layouts {
1496 /** Bit-mask of valid attributes */
1497 /* The LA_* flags are written to disk as part of the ChangeLog records
1498 * so they are part of the on-disk and network protocol, and cannot be changed.
1499 * Only the first 12 bits are currently saved.
1502 LA_ATIME = 1 << 0, /* 0x00001 */
1503 LA_MTIME = 1 << 1, /* 0x00002 */
1504 LA_CTIME = 1 << 2, /* 0x00004 */
1505 LA_SIZE = 1 << 3, /* 0x00008 */
1506 LA_MODE = 1 << 4, /* 0x00010 */
1507 LA_UID = 1 << 5, /* 0x00020 */
1508 LA_GID = 1 << 6, /* 0x00040 */
1509 LA_BLOCKS = 1 << 7, /* 0x00080 */
1510 LA_TYPE = 1 << 8, /* 0x00100 */
1511 LA_FLAGS = 1 << 9, /* 0x00200 */
1512 LA_NLINK = 1 << 10, /* 0x00400 */
1513 LA_RDEV = 1 << 11, /* 0x00800 */
1514 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1515 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1516 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1517 LA_PROJID = 1 << 15, /* 0x08000 */
1518 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1519 LA_LSIZE = 1 << 17, /* 0x20000 */
1520 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1521 LA_BTIME = 1 << 19, /* 0x80000 */
1523 * Attributes must be transmitted to OST objects
1525 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1528 #define MDS_FMODE_READ 00000001
1529 #define MDS_FMODE_WRITE 00000002
1531 #define MDS_FMODE_CLOSED 00000000
1532 #define MDS_FMODE_EXEC 00000004
1533 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1534 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1535 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1537 #define MDS_OPEN_CREATED 00000010
1538 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1540 #define MDS_OPEN_CREAT 00000100
1541 #define MDS_OPEN_EXCL 00000200
1542 #define MDS_OPEN_TRUNC 00001000
1543 #define MDS_OPEN_APPEND 00002000
1544 #define MDS_OPEN_SYNC 00010000
1545 #define MDS_OPEN_DIRECTORY 00200000
1547 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1548 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1549 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1550 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1551 * We do not support JOIN FILE
1552 * anymore, reserve this flags
1553 * just for preventing such bit
1556 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1557 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1558 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1559 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1560 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1562 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1564 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1565 * delegation, succeed if it's not
1566 * being opened with conflict mode.
1568 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1570 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1571 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1572 * for newly created file */
1573 #define MDS_OP_WITH_FID 020000000000000ULL /* operation carried out by FID */
1574 #define MDS_OPEN_DEFAULT_LMV 040000000000000ULL /* open fetches default LMV */
1576 /* lustre internal open flags, which should not be set from user space */
1577 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1578 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1579 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1580 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1581 MDS_OPEN_PCC | MDS_OP_WITH_FID | \
1582 MDS_OPEN_DEFAULT_LMV)
1585 /********* Changelogs **********/
1586 /** Changelog record types */
1587 enum changelog_rec_type {
1590 CL_CREATE = 1, /* namespace */
1591 CL_MKDIR = 2, /* namespace */
1592 CL_HARDLINK = 3, /* namespace */
1593 CL_SOFTLINK = 4, /* namespace */
1594 CL_MKNOD = 5, /* namespace */
1595 CL_UNLINK = 6, /* namespace */
1596 CL_RMDIR = 7, /* namespace */
1597 CL_RENAME = 8, /* namespace */
1598 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1599 CL_OPEN = 10, /* not currently used */
1600 CL_CLOSE = 11, /* may be written to log only with mtime change */
1601 CL_LAYOUT = 12, /* file layout/striping modified */
1605 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1606 CL_HSM = 16, /* HSM specific events, see flags */
1607 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1611 CL_FLRW = 21, /* FLR: file was firstly written */
1612 CL_RESYNC = 22, /* FLR: file was resync-ed */
1614 CL_DN_OPEN = 24, /* denied open */
1618 static inline const char *changelog_type2str(int type) {
1619 static const char *const changelog_str[] = {
1620 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1621 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1622 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1623 "FLRW", "RESYNC","GXATR", "NOPEN",
1626 if (type >= 0 && type < CL_LAST)
1627 return changelog_str[type];
1631 /* 12 bits of per-record data can be stored in the bottom of the flags */
1632 #define CLF_FLAGSHIFT 12
1633 enum changelog_rec_flags {
1634 CLF_VERSION = 0x1000,
1635 CLF_RENAME = 0x2000,
1637 CLF_EXTRA_FLAGS = 0x8000,
1638 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1640 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1641 CLF_VERMASK = ~CLF_FLAGMASK,
1645 /* Anything under the flagmask may be per-type (if desired) */
1646 /* Flags for unlink */
1647 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1648 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1649 /* HSM cleaning needed */
1650 /* Flags for rename */
1651 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1653 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1654 * has an archive in backend */
1657 /* 12b used (from high weight to low weight):
1662 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1663 #define CLF_HSM_ERR_H 6
1664 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1665 #define CLF_HSM_EVENT_H 9
1666 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1667 #define CLF_HSM_FLAG_H 11
1668 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1669 #define CLF_HSM_SPARE_H 15
1670 #define CLF_HSM_LAST 15
1672 /* Remove bits higher than _h, then extract the value
1673 * between _h and _l by shifting lower weigth to bit 0. */
1674 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1675 >> (CLF_HSM_LAST - _h + _l))
1677 #define CLF_HSM_SUCCESS 0x00
1678 #define CLF_HSM_MAXERROR 0x7E
1679 #define CLF_HSM_ERROVERFLOW 0x7F
1681 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1683 /* 3 bits field => 8 values allowed */
1695 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1697 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1701 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1704 *clf_flags = (enum changelog_rec_flags)
1705 (*clf_flags | (he << CLF_HSM_EVENT_L));
1708 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1710 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1713 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1716 *clf_flags = (enum changelog_rec_flags)
1717 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1720 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1722 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1725 static inline int hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1728 /* In case a negative error is given */
1731 if (error > CLF_HSM_MAXERROR)
1732 error = CLF_HSM_ERROVERFLOW;
1734 *clf_flags = (enum changelog_rec_flags)
1735 (*clf_flags | (error << CLF_HSM_ERR_L));
1737 return error == CLF_HSM_ERROVERFLOW ? -EOVERFLOW : 0;
1740 enum changelog_rec_extra_flags {
1742 CLFE_UIDGID = 0x0001,
1745 CLFE_XATTR = 0x0008,
1746 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1749 enum changelog_send_flag {
1750 /* Not yet implemented */
1751 CHANGELOG_FLAG_FOLLOW = 0x01,
1752 /* Blocking IO makes sense in case of slow user parsing of the records,
1753 * but it also prevents us from cleaning up if the records are not
1755 CHANGELOG_FLAG_BLOCK = 0x02,
1756 /* Pack jobid into the changelog records if available. */
1757 CHANGELOG_FLAG_JOBID = 0x04,
1758 /* Pack additional flag bits into the changelog record */
1759 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1762 enum changelog_send_extra_flag {
1763 /* Pack uid/gid into the changelog record */
1764 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1765 /* Pack nid into the changelog record */
1766 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1767 /* Pack open mode into the changelog record */
1768 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1769 /* Pack xattr name into the changelog record */
1770 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1773 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1774 changelog_rec_offset(CLF_SUPPORTED, \
1777 /* 31 usable bytes string + null terminator. */
1778 #define LUSTRE_JOBID_SIZE 32
1780 /* This is the minimal changelog record. It can contain extensions
1781 * such as rename fields or process jobid. Its exact content is described
1782 * by the cr_flags and cr_extra_flags.
1784 * Extensions are packed in the same order as their corresponding flags,
1785 * then in the same order as their corresponding extra flags.
1787 struct changelog_rec {
1789 __u16 cr_flags; /**< \a changelog_rec_flags */
1790 __u32 cr_type; /**< \a changelog_rec_type */
1791 __u64 cr_index; /**< changelog record number */
1792 __u64 cr_prev; /**< last index for this target fid */
1795 struct lu_fid cr_tfid; /**< target fid */
1796 __u32 cr_markerflags; /**< CL_MARK flags */
1798 struct lu_fid cr_pfid; /**< parent fid */
1799 } __attribute__ ((packed));
1801 /* Changelog extension for RENAME. */
1802 struct changelog_ext_rename {
1803 struct lu_fid cr_sfid; /**< source fid, or zero */
1804 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1807 /* Changelog extension to include JOBID. */
1808 struct changelog_ext_jobid {
1809 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1812 /* Changelog extension to include additional flags. */
1813 struct changelog_ext_extra_flags {
1814 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1817 /* Changelog extra extension to include UID/GID. */
1818 struct changelog_ext_uidgid {
1823 /* Changelog extra extension to include NID. */
1824 struct changelog_ext_nid {
1825 /* have __u64 instead of lnet_nid_t type for use by client api */
1827 /* for use when IPv6 support is added */
1832 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1833 struct changelog_ext_openmode {
1837 /* Changelog extra extension to include xattr */
1838 struct changelog_ext_xattr {
1839 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1842 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1843 const struct changelog_rec *rec);
1845 static inline __kernel_size_t changelog_rec_offset(enum changelog_rec_flags crf,
1846 enum changelog_rec_extra_flags cref)
1848 __kernel_size_t size = sizeof(struct changelog_rec);
1850 if (crf & CLF_RENAME)
1851 size += sizeof(struct changelog_ext_rename);
1853 if (crf & CLF_JOBID)
1854 size += sizeof(struct changelog_ext_jobid);
1856 if (crf & CLF_EXTRA_FLAGS) {
1857 size += sizeof(struct changelog_ext_extra_flags);
1858 if (cref & CLFE_UIDGID)
1859 size += sizeof(struct changelog_ext_uidgid);
1860 if (cref & CLFE_NID)
1861 size += sizeof(struct changelog_ext_nid);
1862 if (cref & CLFE_OPEN)
1863 size += sizeof(struct changelog_ext_openmode);
1864 if (cref & CLFE_XATTR)
1865 size += sizeof(struct changelog_ext_xattr);
1871 static inline __kernel_size_t changelog_rec_size(const struct changelog_rec *rec)
1873 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1875 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1876 cref = (enum changelog_rec_extra_flags)
1877 changelog_rec_extra_flags(rec)->cr_extra_flags;
1879 return changelog_rec_offset(
1880 (enum changelog_rec_flags)rec->cr_flags, cref);
1883 static inline __kernel_size_t changelog_rec_varsize(const struct changelog_rec *rec)
1885 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1889 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1891 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1892 (rec->cr_flags & CLF_VERSION);
1894 return (struct changelog_ext_rename *)((char *)rec +
1895 changelog_rec_offset(crf,
1899 /* The jobid follows the rename extension, if present */
1901 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1903 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1904 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1906 return (struct changelog_ext_jobid *)((char *)rec +
1907 changelog_rec_offset(crf,
1911 /* The additional flags follow the rename and jobid extensions, if present */
1913 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1914 const struct changelog_rec *rec)
1916 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1917 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1919 return (struct changelog_ext_extra_flags *)((char *)rec +
1920 changelog_rec_offset(crf,
1924 /* The uid/gid is the first extra extension */
1926 struct changelog_ext_uidgid *changelog_rec_uidgid(
1927 const struct changelog_rec *rec)
1929 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1931 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1933 return (struct changelog_ext_uidgid *)((char *)rec +
1934 changelog_rec_offset(crf,
1938 /* The nid is the second extra extension */
1940 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1942 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1944 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1945 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1947 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1948 cref = (enum changelog_rec_extra_flags)
1949 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1952 return (struct changelog_ext_nid *)((char *)rec +
1953 changelog_rec_offset(crf, cref));
1956 /* The OPEN mode is the third extra extension */
1958 struct changelog_ext_openmode *changelog_rec_openmode(
1959 const struct changelog_rec *rec)
1961 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1963 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1964 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1966 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1967 cref = (enum changelog_rec_extra_flags)
1968 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1969 (CLFE_UIDGID | CLFE_NID));
1972 return (struct changelog_ext_openmode *)((char *)rec +
1973 changelog_rec_offset(crf, cref));
1976 /* The xattr name is the fourth extra extension */
1978 struct changelog_ext_xattr *changelog_rec_xattr(
1979 const struct changelog_rec *rec)
1981 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1983 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1984 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1986 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1987 cref = (enum changelog_rec_extra_flags)
1988 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1989 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
1991 return (struct changelog_ext_xattr *)((char *)rec +
1992 changelog_rec_offset(crf, cref));
1995 /* The name follows the rename, jobid and extra flags extns, if present */
1996 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1998 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2000 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2001 cref = (enum changelog_rec_extra_flags)
2002 changelog_rec_extra_flags(rec)->cr_extra_flags;
2004 return (char *)rec + changelog_rec_offset(
2005 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
2006 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
2009 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
2011 return strchrnul(changelog_rec_name(rec), '\0') + 1;
2014 static inline __kernel_size_t changelog_rec_snamelen(const struct changelog_rec *rec)
2016 return strlen(changelog_rec_sname(rec));
2020 * Remap a record to the desired format as specified by the crf flags.
2021 * The record must be big enough to contain the final remapped version.
2022 * Superfluous extension fields are removed and missing ones are added
2023 * and zeroed. The flags of the record are updated accordingly.
2025 * The jobid and rename extensions can be added to a record, to match the
2026 * format an application expects, typically. In this case, the newly added
2027 * fields will be zeroed.
2028 * The Jobid field can be removed, to guarantee compatibility with older
2029 * clients that don't expect this field in the records they process.
2031 * The following assumptions are being made:
2032 * - CLF_RENAME will not be removed
2033 * - CLF_JOBID will not be added without CLF_RENAME being added too
2034 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
2036 * @param[in,out] rec The record to remap.
2037 * @param[in] crf_wanted Flags describing the desired extensions.
2038 * @param[in] cref_want Flags describing the desired extra extensions.
2040 static inline void changelog_remap_rec(struct changelog_rec *rec,
2041 enum changelog_rec_flags crf_wanted,
2042 enum changelog_rec_extra_flags cref_want)
2044 char *xattr_mov = NULL;
2045 char *omd_mov = NULL;
2046 char *nid_mov = NULL;
2047 char *uidgid_mov = NULL;
2051 enum changelog_rec_extra_flags cref = CLFE_INVALID;
2053 crf_wanted = (enum changelog_rec_flags)
2054 (crf_wanted & CLF_SUPPORTED);
2055 cref_want = (enum changelog_rec_extra_flags)
2056 (cref_want & CLFE_SUPPORTED);
2058 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
2059 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
2060 (rec->cr_flags & CLF_EXTRA_FLAGS &&
2061 (changelog_rec_extra_flags(rec)->cr_extra_flags &
2067 /* First move the variable-length name field */
2068 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
2069 changelog_rec_name(rec), rec->cr_namelen);
2071 /* Locations of extensions in the remapped record */
2072 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
2073 xattr_mov = (char *)rec +
2074 changelog_rec_offset(
2075 (enum changelog_rec_flags)
2076 (crf_wanted & CLF_SUPPORTED),
2077 (enum changelog_rec_extra_flags)
2078 (cref_want & ~CLFE_XATTR));
2079 omd_mov = (char *)rec +
2080 changelog_rec_offset(
2081 (enum changelog_rec_flags)
2082 (crf_wanted & CLF_SUPPORTED),
2083 (enum changelog_rec_extra_flags)
2084 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
2085 nid_mov = (char *)rec +
2086 changelog_rec_offset(
2087 (enum changelog_rec_flags)
2088 (crf_wanted & CLF_SUPPORTED),
2089 (enum changelog_rec_extra_flags)
2091 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
2092 uidgid_mov = (char *)rec +
2093 changelog_rec_offset(
2094 (enum changelog_rec_flags)
2095 (crf_wanted & CLF_SUPPORTED),
2096 (enum changelog_rec_extra_flags)
2097 (cref_want & ~(CLFE_UIDGID |
2101 cref = (enum changelog_rec_extra_flags)
2102 changelog_rec_extra_flags(rec)->cr_extra_flags;
2105 ef_mov = (char *)rec +
2106 changelog_rec_offset(
2107 (enum changelog_rec_flags)
2108 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
2109 jid_mov = (char *)rec +
2110 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2111 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
2113 rnm_mov = (char *)rec +
2114 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
2120 /* Move the extension fields to the desired positions */
2121 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2122 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
2123 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
2124 memmove(xattr_mov, changelog_rec_xattr(rec),
2125 sizeof(struct changelog_ext_xattr));
2127 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
2128 memmove(omd_mov, changelog_rec_openmode(rec),
2129 sizeof(struct changelog_ext_openmode));
2131 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
2132 memmove(nid_mov, changelog_rec_nid(rec),
2133 sizeof(struct changelog_ext_nid));
2135 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
2136 memmove(uidgid_mov, changelog_rec_uidgid(rec),
2137 sizeof(struct changelog_ext_uidgid));
2139 memmove(ef_mov, changelog_rec_extra_flags(rec),
2140 sizeof(struct changelog_ext_extra_flags));
2143 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
2144 memmove(jid_mov, changelog_rec_jobid(rec),
2145 sizeof(struct changelog_ext_jobid));
2147 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
2148 memmove(rnm_mov, changelog_rec_rename(rec),
2149 sizeof(struct changelog_ext_rename));
2151 /* Clear newly added fields */
2152 if (xattr_mov && (cref_want & CLFE_XATTR) &&
2153 !(cref & CLFE_XATTR))
2154 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
2156 if (omd_mov && (cref_want & CLFE_OPEN) &&
2157 !(cref & CLFE_OPEN))
2158 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
2160 if (nid_mov && (cref_want & CLFE_NID) &&
2162 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
2164 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
2165 !(cref & CLFE_UIDGID))
2166 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
2168 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
2169 !(rec->cr_flags & CLF_EXTRA_FLAGS))
2170 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
2172 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
2173 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
2175 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
2176 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
2178 /* Update the record's flags accordingly */
2179 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
2180 if (rec->cr_flags & CLF_EXTRA_FLAGS)
2181 changelog_rec_extra_flags(rec)->cr_extra_flags =
2182 changelog_rec_extra_flags(rec)->cr_extra_flags |
2186 enum changelog_message_type {
2187 CL_RECORD = 10, /* message is a changelog_rec */
2188 CL_EOF = 11, /* at end of current changelog */
2191 /********* Misc **********/
2193 struct ioc_data_version {
2195 __u32 idv_layout_version; /* FLR: layout version for OST objects */
2196 __u32 idv_flags; /* enum ioc_data_version_flags */
2199 enum ioc_data_version_flags {
2200 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
2201 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
2202 LL_DV_SZ_UPDATE = (1 << 2), /* Update the file size on the client */
2206 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
2209 #define dot_lustre_name ".lustre"
2210 #define dot_fscrypt_name ".fscrypt"
2213 /********* HSM **********/
2215 /** HSM per-file state
2216 * See HSM_FLAGS below.
2219 HS_NONE = 0x00000000,
2220 HS_EXISTS = 0x00000001,
2221 HS_DIRTY = 0x00000002,
2222 HS_RELEASED = 0x00000004,
2223 HS_ARCHIVED = 0x00000008,
2224 HS_NORELEASE = 0x00000010,
2225 HS_NOARCHIVE = 0x00000020,
2226 HS_LOST = 0x00000040,
2229 /* HSM user-setable flags. */
2230 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
2232 /* Other HSM flags. */
2233 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
2236 * All HSM-related possible flags that could be applied to a file.
2237 * This should be kept in sync with hsm_states.
2239 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
2242 * HSM request progress state
2244 enum hsm_progress_states {
2251 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
2254 case HPS_WAITING: return "waiting";
2255 case HPS_RUNNING: return "running";
2256 case HPS_DONE: return "done";
2257 default: return "unknown";
2264 } __attribute__((packed));
2267 * Current HSM states of a Lustre file.
2269 * This structure purpose is to be sent to user-space mainly. It describes the
2270 * current HSM flags and in-progress action.
2272 struct hsm_user_state {
2273 /** Current HSM states, from enum hsm_states. */
2275 __u32 hus_archive_id;
2276 /** The current undergoing action, if there is one */
2277 __u32 hus_in_progress_state;
2278 __u32 hus_in_progress_action;
2279 struct hsm_extent hus_in_progress_location;
2280 char hus_extended_info[];
2283 struct hsm_state_set_ioc {
2284 struct lu_fid hssi_fid;
2286 __u64 hssi_clearmask;
2290 * This structure describes the current in-progress action for a file.
2291 * it is retuned to user space and send over the wire
2293 struct hsm_current_action {
2294 /** The current undergoing action, if there is one */
2295 /* state is one of hsm_progress_states */
2297 /* action is one of hsm_user_action */
2299 struct hsm_extent hca_location;
2302 /***** HSM user requests ******/
2303 /* User-generated (lfs/ioctl) request types */
2304 enum hsm_user_action {
2305 HUA_NONE = 1, /* no action (noop) */
2306 HUA_ARCHIVE = 10, /* copy to hsm */
2307 HUA_RESTORE = 11, /* prestage */
2308 HUA_RELEASE = 12, /* drop ost objects */
2309 HUA_REMOVE = 13, /* remove from archive */
2310 HUA_CANCEL = 14 /* cancel a request */
2313 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2316 case HUA_NONE: return "NOOP";
2317 case HUA_ARCHIVE: return "ARCHIVE";
2318 case HUA_RESTORE: return "RESTORE";
2319 case HUA_RELEASE: return "RELEASE";
2320 case HUA_REMOVE: return "REMOVE";
2321 case HUA_CANCEL: return "CANCEL";
2322 default: return "UNKNOWN";
2327 * List of hr_flags (bit field)
2329 #define HSM_FORCE_ACTION 0x0001
2330 /* used by CT, cannot be set by user */
2331 #define HSM_GHOST_COPY 0x0002
2334 * Contains all the fixed part of struct hsm_user_request.
2337 struct hsm_request {
2338 __u32 hr_action; /* enum hsm_user_action */
2339 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2340 __u64 hr_flags; /* request flags */
2341 __u32 hr_itemcount; /* item count in hur_user_item vector */
2345 struct hsm_user_item {
2346 struct lu_fid hui_fid;
2347 struct hsm_extent hui_extent;
2348 } __attribute__((packed));
2350 struct hsm_user_request {
2351 struct hsm_request hur_request;
2352 struct hsm_user_item hur_user_item[0];
2353 /* extra data blob at end of struct (after all
2354 * hur_user_items), only use helpers to access it
2356 } __attribute__((packed));
2358 /** Return pointer to data field in a hsm user request */
2359 static inline void *hur_data(struct hsm_user_request *hur)
2361 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2365 * Compute the current length of the provided hsm_user_request. This returns -1
2366 * instead of an errno because __kernel_ssize_t is defined to be only
2369 * return -1 on bounds check error.
2371 static inline __kernel_size_t hur_len(struct hsm_user_request *hur)
2375 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2376 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2377 (__u64)hur->hur_request.hr_itemcount *
2378 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2380 if ((__kernel_ssize_t)size < 0)
2386 /****** HSM RPCs to copytool *****/
2387 /* Message types the copytool may receive */
2388 enum hsm_message_type {
2389 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2392 /* Actions the copytool may be instructed to take for a given action_item */
2393 enum hsm_copytool_action {
2394 HSMA_NONE = 10, /* no action */
2395 HSMA_ARCHIVE = 20, /* arbitrary offset */
2401 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2404 case HSMA_NONE: return "NOOP";
2405 case HSMA_ARCHIVE: return "ARCHIVE";
2406 case HSMA_RESTORE: return "RESTORE";
2407 case HSMA_REMOVE: return "REMOVE";
2408 case HSMA_CANCEL: return "CANCEL";
2409 default: return "UNKNOWN";
2413 /* Copytool item action description */
2414 struct hsm_action_item {
2415 __u32 hai_len; /* valid size of this struct */
2416 __u32 hai_action; /* hsm_copytool_action, but use known size */
2417 struct lu_fid hai_fid; /* Lustre FID to operate on */
2418 struct lu_fid hai_dfid; /* fid used for data access */
2419 struct hsm_extent hai_extent; /* byte range to operate on */
2420 __u64 hai_cookie; /* action cookie from coordinator */
2421 __u64 hai_gid; /* grouplock id */
2422 char hai_data[0]; /* variable length */
2423 } __attribute__((packed));
2426 * helper function which print in hexa the first bytes of
2429 * \param hai [IN] record to print
2430 * \param buffer [IN,OUT] buffer to write the hex string to
2431 * \param len [IN] max buffer length
2435 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2436 char *buffer, __kernel_size_t len)
2443 data_len = hai->hai_len - sizeof(*hai);
2444 for (i = 0; (i < data_len) && (len > 2); i++) {
2445 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2455 /* Copytool action list */
2456 #define HAL_VERSION 1
2457 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2458 struct hsm_action_list {
2460 __u32 hal_count; /* number of hai's to follow */
2461 __u64 hal_compound_id; /* returned by coordinator, ignored */
2463 __u32 hal_archive_id; /* which archive backend */
2465 char hal_fsname[0]; /* null-terminated */
2466 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2467 boundaries. See hai_zero */
2468 } __attribute__((packed));
2470 /* Return pointer to first hai in action list */
2471 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2473 __kernel_size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2475 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2478 /* Return pointer to next hai */
2479 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2481 __kernel_size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2483 return (struct hsm_action_item *)((char *)hai + offset);
2486 /* Return size of an hsm_action_list */
2487 static inline __kernel_size_t hal_size(struct hsm_action_list *hal)
2491 struct hsm_action_item *hai;
2493 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2494 hai = hai_first(hal);
2495 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2496 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2502 * describe the attributes to be set on imported file
2504 struct hsm_user_import {
2513 __u32 hui_archive_id;
2516 /* Copytool progress reporting */
2517 #define HP_FLAG_COMPLETED 0x01
2518 #define HP_FLAG_RETRY 0x02
2520 struct hsm_progress {
2521 struct lu_fid hp_fid;
2523 struct hsm_extent hp_extent;
2525 __u16 hp_errval; /* positive val */
2530 __u64 hc_data_version;
2532 __u16 hc_errval; /* positive val */
2534 struct hsm_action_item hc_hai;
2537 enum lu_ladvise_type {
2538 LU_LADVISE_INVALID = 0,
2539 LU_LADVISE_WILLREAD = 1,
2540 LU_LADVISE_DONTNEED = 2,
2541 LU_LADVISE_LOCKNOEXPAND = 3,
2542 LU_LADVISE_LOCKAHEAD = 4,
2546 #define LU_LADVISE_NAMES { \
2547 [LU_LADVISE_WILLREAD] = "willread", \
2548 [LU_LADVISE_DONTNEED] = "dontneed", \
2549 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2550 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2553 /* This is the userspace argument for ladvise. It is currently the same as
2554 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2555 * may need info which is only used locally. */
2556 struct llapi_lu_ladvise {
2557 __u16 lla_advice; /* advice type */
2558 __u16 lla_value1; /* values for different advice types */
2560 __u64 lla_start; /* first byte of extent for advice */
2561 __u64 lla_end; /* last byte of extent for advice */
2567 LF_ASYNC = 0x00000001,
2568 LF_UNSET = 0x00000002,
2571 #define LADVISE_MAGIC 0x1ADF1CE0
2572 /* Masks of valid flags for each advice */
2573 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2574 /* Flags valid for all advices not explicitly specified */
2575 #define LF_DEFAULT_MASK LF_ASYNC
2577 #define LF_MASK (LF_ASYNC | LF_UNSET)
2579 #define lla_lockahead_mode lla_value1
2580 #define lla_peradvice_flags lla_value2
2581 #define lla_lockahead_result lla_value3
2583 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2584 * is used on the wire. It is defined separately as we may need info which is
2585 * only used locally. */
2586 struct llapi_ladvise_hdr {
2587 __u32 lah_magic; /* LADVISE_MAGIC */
2588 __u32 lah_count; /* number of advices */
2589 __u64 lah_flags; /* from enum ladvise_flag */
2590 __u32 lah_value1; /* unused */
2591 __u32 lah_value2; /* unused */
2592 __u64 lah_value3; /* unused */
2593 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2596 #define LAH_COUNT_MAX (1024)
2600 SK_CRYPT_INVALID = -1,
2602 SK_CRYPT_AES256_CTR = 1,
2606 SK_HMAC_INVALID = -1,
2612 struct sk_crypt_type {
2613 const char *sct_name;
2617 struct sk_hmac_type {
2618 const char *sht_name;
2622 enum lock_mode_user {
2628 #define LOCK_MODE_NAMES { \
2629 [MODE_READ_USER] = "READ",\
2630 [MODE_WRITE_USER] = "WRITE"\
2633 enum lockahead_results {
2634 LLA_RESULT_SENT = 0,
2635 LLA_RESULT_DIFFERENT,
2639 enum lu_heat_flag_bit {
2640 LU_HEAT_FLAG_BIT_INVALID = 0,
2641 LU_HEAT_FLAG_BIT_OFF,
2642 LU_HEAT_FLAG_BIT_CLEAR,
2646 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2647 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2650 enum obd_heat_type {
2651 OBD_HEAT_READSAMPLE = 0,
2652 OBD_HEAT_WRITESAMPLE = 1,
2653 OBD_HEAT_READBYTE = 2,
2654 OBD_HEAT_WRITEBYTE = 3,
2658 #define LU_HEAT_NAMES { \
2659 [OBD_HEAT_READSAMPLE] = "readsample", \
2660 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2661 [OBD_HEAT_READBYTE] = "readbyte", \
2662 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2677 static inline const char *pcc_type2string(enum lu_pcc_type type)
2682 case LU_PCC_READWRITE:
2689 struct lu_pcc_attach {
2690 __u32 pcca_type; /* PCC type */
2691 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2694 enum lu_pcc_detach_opts {
2695 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2696 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2699 struct lu_pcc_detach_fid {
2700 /* fid of the file to detach */
2701 struct lu_fid pccd_fid;
2705 struct lu_pcc_detach {
2709 enum lu_pcc_state_flags {
2710 PCC_STATE_FL_NONE = 0x0,
2711 /* The inode attr is cached locally */
2712 PCC_STATE_FL_ATTR_VALID = 0x01,
2713 /* The file is being attached into PCC */
2714 PCC_STATE_FL_ATTACHING = 0x02,
2717 struct lu_pcc_state {
2718 __u32 pccs_type; /* enum lu_pcc_type */
2719 __u32 pccs_open_count;
2720 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2722 char pccs_path[PATH_MAX];
2725 enum lu_project_type {
2726 LU_PROJECT_NONE = 0,
2733 __u32 project_type; /* enum lu_project_type */
2735 __u32 project_xflags;
2736 __u32 project_reserved;
2737 char project_name[NAME_MAX + 1];
2742 /* make header's size equal lu_fid */
2745 struct lu_fid fa_fids[0];
2747 #define OBD_MAX_FIDS_IN_ARRAY 4096
2749 /* more types could be defined upon need for more complex
2750 * format to be used in foreign symlink LOV/LMV EAs, like
2751 * one to describe a delimiter string and occurence number
2752 * of delimited sub-string, ...
2754 enum ll_foreign_symlink_upcall_item_type {
2760 /* may need to be modified to allow for more format items to be defined, and
2761 * like for ll_foreign_symlink_upcall_item_type enum
2763 struct ll_foreign_symlink_upcall_item {
2773 /* internal storage of constant string */
2775 /* upcall stores constant string in a raw */
2782 #define POSLEN_ITEM_SZ (offsetof(struct ll_foreign_symlink_upcall_item, len) + \
2783 sizeof(((struct ll_foreign_symlink_upcall_item *)0)->len))
2784 #define STRING_ITEM_SZ(sz) ( \
2785 offsetof(struct ll_foreign_symlink_upcall_item, bytestring) + \
2786 (sz + sizeof(__u32) - 1) / sizeof(__u32) * sizeof(__u32))
2788 /* presently limited to not cause max stack frame size to be reached
2789 * because of temporary automatic array of
2790 * "struct ll_foreign_symlink_upcall_item" presently used in
2791 * foreign_symlink_upcall_info_store()
2793 #define MAX_NB_UPCALL_ITEMS 32
2795 #if defined(__cplusplus)
2799 /** @} lustreuser */
2801 #endif /* _LUSTRE_USER_H */