4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_user.h
34 * Lustre public user-space interface definitions.
37 #ifndef _LUSTRE_USER_H
38 #define _LUSTRE_USER_H
40 /** \defgroup lustreuser lustreuser
45 #include <linux/kernel.h>
46 #include <linux/types.h>
49 # include <linux/fs.h>
50 # include <linux/quota.h>
51 # include <linux/string.h> /* snprintf() */
52 # include <linux/version.h>
53 # include <uapi/linux/lustre/lustre_fiemap.h>
54 #else /* !__KERNEL__ */
57 # include <stdio.h> /* snprintf() */
59 # define NEED_QUOTA_DEFS
60 /* # include <sys/quota.h> - this causes complaints about caddr_t */
61 # include <sys/stat.h>
62 # include <linux/lustre/lustre_fiemap.h>
63 #endif /* __KERNEL__ */
65 /* Handle older distros */
66 #ifndef __ALIGN_KERNEL
67 # define __ALIGN_KERNEL(x, a) __ALIGN_KERNEL_MASK(x, (typeof(x))(a) - 1)
68 # define __ALIGN_KERNEL_MASK(x, mask) (((x) + (mask)) & ~(mask))
71 #if defined(__cplusplus)
76 * This is a temporary solution of adding quota type.
77 * Should be removed as soon as system header is updated.
80 #define LL_MAXQUOTAS 3
82 #define INITQFNAMES { \
83 "user", /* USRQUOTA */ \
84 "group", /* GRPQUOTA */ \
85 "project", /* PRJQUOTA */ \
99 * We need to always use 64bit version because the structure
100 * is shared across entire cluster where 32bit and 64bit machines
103 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
104 typedef struct stat64 lstat_t;
105 #define lstat_f lstat64
106 #define fstat_f fstat64
107 #define fstatat_f fstatat64
109 typedef struct stat lstat_t;
110 #define lstat_f lstat
111 #define fstat_f fstat
112 #define fstatat_f fstatat
115 #define HAVE_LOV_USER_MDS_DATA
117 #define LUSTRE_EOF 0xffffffffffffffffULL
120 #define LL_SUPER_MAGIC 0x0BD00BD0
122 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
124 /* FIEMAP flags supported by Lustre */
125 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
127 enum obd_statfs_state {
128 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
129 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
130 OS_STATE_NOPRECREATE = 0x00000004, /**< no object precreation */
131 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
132 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
133 OS_STATE_SUM = 0x00000100, /**< aggregated for all tagrets */
136 /** filesystem statistics/attributes for target device */
138 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
139 __u64 os_blocks; /* total size in #os_bsize blocks */
140 __u64 os_bfree; /* number of unused blocks */
141 __u64 os_bavail; /* blocks available for allocation */
142 __u64 os_files; /* total number of objects */
143 __u64 os_ffree; /* # objects that could be created */
144 __u8 os_fsid[40]; /* identifier for filesystem */
145 __u32 os_bsize; /* block size in bytes for os_blocks */
146 __u32 os_namelen; /* maximum length of filename in bytes*/
147 __u64 os_maxbytes; /* maximum object size in bytes */
148 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
149 __u32 os_fprecreated; /* objs available now to the caller */
150 /* used in QoS code to find preferred
152 __u32 os_granted; /* space granted for MDS */
153 __u32 os_spare3; /* Unused padding fields. Remember */
154 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
165 * FID is a cluster-wide unique identifier of a file or an object (stripe).
166 * FIDs are never reused.
170 * FID sequence. Sequence is a unit of migration: all files (objects)
171 * with FIDs from a given sequence are stored on the same server.
172 * Lustre should support 2^64 objects, so even if each sequence
173 * has only a single object we can still enumerate 2^64 objects.
176 /* FID number within sequence. */
179 * FID version, used to distinguish different versions (in the sense
180 * of snapshots, etc.) of the same file system object. Not currently
186 static inline bool fid_is_zero(const struct lu_fid *fid)
188 return fid->f_seq == 0 && fid->f_oid == 0;
191 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
192 * MDT-object's FID::f_ver, instead it is the OST-object index in its
193 * parent MDT-object's layout EA. */
194 #define f_stripe_idx f_ver
197 __u32 ol_stripe_size;
198 __u32 ol_stripe_count;
202 } __attribute__((packed));
204 /* The filter_fid structure has changed several times over its lifetime.
205 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
206 * stripe_index and the "self FID" (objid/seq) to be able to recover the
207 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
208 * the OST, the "trusted.lma" xattr was added to the OST objects to store
209 * the "self FID" to be consistent with the MDT on-disk format, and the
210 * filter_fid only stored the MDT inode parent FID and stripe index.
212 * In 2.10, the addition of PFL composite layouts required more information
213 * to be stored into the filter_fid in order to be able to identify which
214 * component the OST object belonged. As well, the stripe size may vary
215 * between components, so it was no longer safe to assume the stripe size
216 * or stripe_count of a file. This is also more robust for plain layouts.
218 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
219 * enough space to store both the filter_fid and LMA in the inode, so they
220 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
221 * an extra seek for every OST object access.
223 * In 2.11, FLR mirror layouts also need to store the layout version and
224 * range so that writes to old versions of the layout are not allowed.
225 * That ensures that mirrored objects are not modified by evicted clients,
226 * and ensures that the components are correctly marked stale on the MDT.
228 struct filter_fid_18_23 {
229 struct lu_fid ff_parent; /* stripe_idx in f_ver */
234 struct filter_fid_24_29 {
235 struct lu_fid ff_parent; /* stripe_idx in f_ver */
238 struct filter_fid_210 {
239 struct lu_fid ff_parent; /* stripe_idx in f_ver */
240 struct ost_layout ff_layout;
244 struct lu_fid ff_parent; /* stripe_idx in f_ver */
245 struct ost_layout ff_layout;
246 __u32 ff_layout_version;
247 __u32 ff_range; /* range of layout version that
248 * write are allowed */
249 } __attribute__((packed));
251 /* Userspace should treat lu_fid as opaque, and only use the following methods
252 * to print or parse them. Other functions (e.g. compare, swab) could be moved
253 * here from lustre_idl.h if needed. */
257 LMAC_HSM = 0x00000001,
258 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
259 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
260 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
261 * under /O/<seq>/d<x>. */
262 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
263 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
264 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
268 * Masks for all features that should be supported by a Lustre version to
269 * access a specific file.
270 * This information is stored in lustre_mdt_attrs::lma_incompat.
273 LMAI_RELEASED = 0x00000001, /* file is released */
274 LMAI_AGENT = 0x00000002, /* agent inode */
275 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
276 is on the remote MDT */
277 LMAI_STRIPED = 0x00000008, /* striped directory inode */
278 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
279 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
280 LMAI_STRIPED | LMAI_ORPHAN)
285 * Following struct for object attributes, that will be kept inode's EA.
286 * Introduced in 2.0 release (please see b15993, for details)
287 * Added to all objects since Lustre 2.4 as contains self FID
289 struct lustre_mdt_attrs {
291 * Bitfield for supported data in this structure. From enum lma_compat.
292 * lma_self_fid and lma_flags are always available.
296 * Per-file incompat feature list. Lustre version should support all
297 * flags set in this field. The supported feature mask is available in
301 /** FID of this inode */
302 struct lu_fid lma_self_fid;
305 struct lustre_ost_attrs {
306 /* Use lustre_mdt_attrs directly for now, need a common header
307 * structure if want to change lustre_mdt_attrs in future. */
308 struct lustre_mdt_attrs loa_lma;
310 /* Below five elements are for OST-object's PFID EA, the
311 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
312 * and the stripe_index (low 16 bits), the size should not exceed
313 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
314 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
315 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
316 * loa_comp_* elements are valid. */
317 struct lu_fid loa_parent_fid;
318 __u32 loa_stripe_size;
320 __u64 loa_comp_start;
325 * Prior to 2.4, the LMA structure also included SOM attributes which has since
326 * been moved to a dedicated xattr
327 * lma_flags was also removed because of lma_compat/incompat fields.
329 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
331 enum lustre_som_flags {
332 /* Unknow or no SoM data, must get size from OSTs. */
333 SOM_FL_UNKNOWN = 0x0000,
334 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
335 SOM_FL_STRICT = 0x0001,
336 /* Known stale - was right at some point in the past, but it is
337 * known (or likely) to be incorrect now (e.g. opened for write). */
338 SOM_FL_STALE = 0x0002,
339 /* Approximate, may never have been strictly correct,
340 * need to sync SOM data to achieve eventual consistency. */
341 SOM_FL_LAZY = 0x0004,
344 struct lustre_som_attrs {
346 __u16 lsa_reserved[3];
352 * OST object IDentifier.
360 struct lu_fid oi_fid;
364 #define DOSTID "%#llx:%llu"
365 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
366 ((unsigned long long)ostid_id(oi))
368 struct ll_futimes_3 {
370 __u64 lfu_atime_nsec;
372 __u64 lfu_mtime_nsec;
374 __u64 lfu_ctime_nsec;
378 * Maximum number of mirrors currently implemented.
380 #define LUSTRE_MIRROR_COUNT_MAX 16
382 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
384 LL_LEASE_RDLCK = 0x01,
385 LL_LEASE_WRLCK = 0x02,
386 LL_LEASE_UNLCK = 0x04,
389 enum ll_lease_flags {
390 LL_LEASE_RESYNC = 0x1,
391 LL_LEASE_RESYNC_DONE = 0x2,
392 LL_LEASE_LAYOUT_MERGE = 0x4,
393 LL_LEASE_LAYOUT_SPLIT = 0x8,
396 #define IOC_IDS_MAX 4096
397 struct ll_ioc_lease {
404 struct ll_ioc_lease_id {
415 * The ioctl naming rules:
416 * LL_* - works on the currently opened filehandle instead of parent dir
417 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
418 * *_MDC_* - gets/sets data related to MDC
419 * *_LOV_* - gets/sets data related to OSC/LOV
420 * *FILE* - called on parent dir and passes in a filename
421 * *STRIPE* - set/get lov_user_md
422 * *INFO - set/get lov_user_mds_data
424 /* lustre_ioctl.h 101-150 */
425 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
426 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
427 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
428 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
429 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
430 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
431 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
432 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
433 /* LL_IOC_RECREATE_OBJ 157 obsolete */
434 /* LL_IOC_RECREATE_FID 157 obsolete */
435 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
436 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
437 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
438 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
439 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
440 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
441 /* IOC_LOV_GETINFO 165 obsolete */
442 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
443 /* LL_IOC_RMTACL 167 obsolete */
444 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
445 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
446 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
447 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
448 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
449 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
450 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
451 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
452 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
453 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
454 /* lustre_ioctl.h 177-210 */
455 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
456 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
457 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
458 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
459 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
460 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
461 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
462 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
463 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
464 struct lustre_swap_layouts)
465 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
466 struct hsm_current_action)
467 /* lustre_ioctl.h 221-232 */
468 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
469 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
470 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
471 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
472 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
473 #define LL_IOC_GET_LEASE _IO('f', 244)
474 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
475 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
476 #define LL_IOC_MIGRATE _IOR('f', 247, int)
477 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
478 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
479 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
480 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
481 #define LL_IOC_HEAT_SET _IOW('f', 252, long)
483 #ifndef FS_IOC_FSGETXATTR
485 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
488 __u32 fsx_xflags; /* xflags field value (get/set) */
489 __u32 fsx_extsize; /* extsize field value (get/set)*/
490 __u32 fsx_nextents; /* nextents field value (get) */
491 __u32 fsx_projid; /* project identifier (get/set) */
492 unsigned char fsx_pad[12];
494 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
495 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
497 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
498 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
499 #ifndef FS_XFLAG_PROJINHERIT
500 #define FS_XFLAG_PROJINHERIT 0x00000200
504 #define LL_STATFS_LMV 1
505 #define LL_STATFS_LOV 2
506 #define LL_STATFS_NODELAY 4
508 #define IOC_MDC_TYPE 'i'
509 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
510 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
511 #define IOC_MDC_GETFILEINFO_OLD _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data_v1 *)
512 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data)
513 #define LL_IOC_MDC_GETINFO_OLD _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data_v1 *)
514 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data)
516 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
518 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
519 * files, but are unlikely to be used in practice and are not harmful if
520 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
521 * devices and are safe for use on new files. See LU-4209. */
522 /* To be compatible with old statically linked binary we keep the check for
523 * the older 0100000000 flag. This is already removed upstream. LU-812. */
524 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
526 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
528 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
529 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
530 O_LOV_DELAY_CREATE_MASK)
532 #define LL_FILE_IGNORE_LOCK 0x00000001
533 #define LL_FILE_GROUP_LOCKED 0x00000002
534 #define LL_FILE_READAHEA 0x00000004
535 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
536 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
538 #define LOV_USER_MAGIC_V1 0x0BD10BD0
539 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
540 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
541 #define LOV_USER_MAGIC_V3 0x0BD30BD0
542 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
543 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
544 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
546 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
547 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
548 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
550 #define LOV_PATTERN_NONE 0x000
551 #define LOV_PATTERN_RAID0 0x001
552 #define LOV_PATTERN_RAID1 0x002
553 #define LOV_PATTERN_MDT 0x100
554 #define LOV_PATTERN_CMOBD 0x200
556 #define LOV_PATTERN_F_MASK 0xffff0000
557 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
558 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
559 #define LOV_PATTERN_DEFAULT 0xffffffff
561 #define LOV_OFFSET_DEFAULT ((__u16)-1)
563 static inline bool lov_pattern_supported(__u32 pattern)
565 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
566 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
569 #define LOV_MAXPOOLNAME 15
570 #define LOV_POOLNAMEF "%.15s"
572 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
573 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
574 #define LOV_MAX_STRIPE_COUNT_OLD 160
575 /* This calculation is crafted so that input of 4096 will result in 160
576 * which in turn is equal to old maximal stripe count.
577 * XXX: In fact this is too simpified for now, what it also need is to get
578 * ea_type argument to clearly know how much space each stripe consumes.
580 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
581 * allocation that is sufficient for the current generation of systems.
583 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
584 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
585 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
586 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
588 #define XATTR_LUSTRE_PREFIX "lustre."
589 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
591 #define lov_user_ost_data lov_user_ost_data_v1
592 struct lov_user_ost_data_v1 { /* per-stripe data structure */
593 struct ost_id l_ost_oi; /* OST object ID */
594 __u32 l_ost_gen; /* generation of this OST index */
595 __u32 l_ost_idx; /* OST index in LOV */
596 } __attribute__((packed));
598 #define lov_user_md lov_user_md_v1
599 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
600 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
601 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
602 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
603 __u32 lmm_stripe_size; /* size of stripe in bytes */
604 __u16 lmm_stripe_count; /* num stripes in use for this object */
606 __u16 lmm_stripe_offset; /* starting stripe offset in
607 * lmm_objects, use when writing */
608 __u16 lmm_layout_gen; /* layout generation number
609 * used when reading */
611 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
612 } __attribute__((packed, __may_alias__));
614 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
615 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
616 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
617 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
618 __u32 lmm_stripe_size; /* size of stripe in bytes */
619 __u16 lmm_stripe_count; /* num stripes in use for this object */
621 __u16 lmm_stripe_offset; /* starting stripe offset in
622 * lmm_objects, use when writing */
623 __u16 lmm_layout_gen; /* layout generation number
624 * used when reading */
626 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
627 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
628 } __attribute__((packed));
635 #define DEXT "[%#llx, %#llx)"
636 #define PEXT(ext) (ext)->e_start, (ext)->e_end
638 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
639 struct lu_extent *e2)
641 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
644 static inline bool lu_extent_is_whole(struct lu_extent *e)
646 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
649 enum lov_comp_md_entry_flags {
650 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
651 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
652 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
653 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
654 LCME_FL_OFFLINE = 0x00000008, /* Not used */
655 LCME_FL_INIT = 0x00000010, /* instantiated */
656 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
657 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
658 won't be stored on disk */
661 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
662 LCME_FL_PREF_RW | LCME_FL_NOSYNC)
663 /* The flags can be set by users at mirror creation time. */
664 #define LCME_USER_FLAGS (LCME_FL_PREF_RW)
666 /* The flags are for mirrors */
667 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
669 /* These flags have meaning when set in a default layout and will be inherited
670 * from the default/template layout set on a directory.
672 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC)
674 /* the highest bit in obdo::o_layout_version is used to mark if the file is
676 #define LU_LAYOUT_RESYNC LCME_FL_NEG
678 /* lcme_id can be specified as certain flags, and the the first
679 * bit of lcme_id is used to indicate that the ID is representing
680 * certain LCME_FL_* but not a real ID. Which implies we can have
681 * at most 31 flags (see LCME_FL_XXX). */
684 LCME_ID_MAX = 0x7FFFFFFF,
685 LCME_ID_ALL = 0xFFFFFFFF,
686 LCME_ID_NOT_ID = LCME_FL_NEG
689 #define LCME_ID_MASK LCME_ID_MAX
691 struct lov_comp_md_entry_v1 {
692 __u32 lcme_id; /* unique id of component */
693 __u32 lcme_flags; /* LCME_FL_XXX */
694 struct lu_extent lcme_extent; /* file extent for component */
695 __u32 lcme_offset; /* offset of component blob,
696 start from lov_comp_md_v1 */
697 __u32 lcme_size; /* size of component blob */
698 __u32 lcme_layout_gen;
699 __u64 lcme_timestamp; /* snapshot time if applicable*/
700 __u32 lcme_padding_1;
701 } __attribute__((packed));
703 #define SEQ_ID_MAX 0x0000FFFF
704 #define SEQ_ID_MASK SEQ_ID_MAX
705 /* bit 30:16 of lcme_id is used to store mirror id */
706 #define MIRROR_ID_MASK 0x7FFF0000
707 #define MIRROR_ID_NEG 0x8000
708 #define MIRROR_ID_SHIFT 16
710 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
712 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
715 static inline __u16 mirror_id_of(__u32 id)
717 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
721 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
723 enum lov_comp_md_flags {
724 /* the least 2 bits are used by FLR to record file state */
727 LCM_FL_WRITE_PENDING = 2,
728 LCM_FL_SYNC_PENDING = 3,
729 LCM_FL_FLR_MASK = 0x3,
732 struct lov_comp_md_v1 {
733 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
734 __u32 lcm_size; /* overall size including this struct */
735 __u32 lcm_layout_gen;
737 __u16 lcm_entry_count;
738 /* lcm_mirror_count stores the number of actual mirrors minus 1,
739 * so that non-flr files will have value 0 meaning 1 mirror. */
740 __u16 lcm_mirror_count;
741 __u16 lcm_padding1[3];
743 struct lov_comp_md_entry_v1 lcm_entries[0];
744 } __attribute__((packed));
746 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
748 if (stripes == (__u16)-1)
751 if (lmm_magic == LOV_USER_MAGIC_V1)
752 return sizeof(struct lov_user_md_v1) +
753 stripes * sizeof(struct lov_user_ost_data_v1);
754 return sizeof(struct lov_user_md_v3) +
755 stripes * sizeof(struct lov_user_ost_data_v1);
758 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
759 * use this. It is unsafe to #define those values in this header as it
760 * is possible the application has already #included <sys/stat.h>. */
761 #ifdef HAVE_LOV_USER_MDS_DATA
762 #define lov_user_mds_data lov_user_mds_data_v2
763 struct lov_user_mds_data_v1 {
764 lstat_t lmd_st; /* MDS stat struct */
765 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
766 } __attribute__((packed));
768 struct lov_user_mds_data_v2 {
769 lstat_t lmd_st; /* MDS stat struct */
770 __u64 lmd_flags; /* MDS stat flags */
771 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
772 } __attribute__((packed));
775 struct lmv_user_mds_data {
776 struct lu_fid lum_fid;
782 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
783 LMV_HASH_TYPE_ALL_CHARS = 1,
784 LMV_HASH_TYPE_FNV_1A_64 = 2,
788 #define LMV_HASH_NAME_ALL_CHARS "all_char"
789 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
791 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
793 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
794 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
795 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
796 #define lmv_user_md lmv_user_md_v1
797 struct lmv_user_md_v1 {
798 __u32 lum_magic; /* must be the first field */
799 __u32 lum_stripe_count; /* dirstripe count */
800 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
801 __u32 lum_hash_type; /* Dir stripe policy */
802 __u32 lum_type; /* LMV type: default or normal */
806 char lum_pool_name[LOV_MAXPOOLNAME + 1];
807 struct lmv_user_mds_data lum_objects[0];
808 } __attribute__((packed));
810 static inline int lmv_user_md_size(int stripes, int lmm_magic)
812 int size = sizeof(struct lmv_user_md);
814 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
815 size += stripes * sizeof(struct lmv_user_mds_data);
820 struct ll_recreate_obj {
826 __u64 id; /* holds object id */
827 __u32 generation; /* holds object generation */
828 __u32 f_type; /* holds object type or stripe idx when passing it to
829 * OST for saving into EA. */
837 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
838 const struct obd_uuid *u2)
840 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
843 static inline int obd_uuid_empty(struct obd_uuid *uuid)
845 return uuid->uuid[0] == '\0';
848 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
850 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
851 uuid->uuid[sizeof(*uuid) - 1] = '\0';
854 /* For printf's only, make sure uuid is terminated */
855 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
860 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
861 /* Obviously not safe, but for printfs, no real harm done...
862 we're always null-terminated, even in a race. */
863 static char temp[sizeof(*uuid)];
864 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
865 temp[sizeof(*uuid) - 1] = '\0';
868 return (char *)(uuid->uuid);
871 #define LUSTRE_MAXFSNAME 8
873 /* Extract fsname from uuid (or target name) of a target
874 e.g. (myfs-OST0007_UUID -> myfs)
875 see also deuuidify. */
876 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
880 strncpy(buf, uuid, buflen - 1);
881 buf[buflen - 1] = '\0';
882 p = strrchr(buf, '-');
887 /* printf display format for Lustre FIDs
888 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
889 #define FID_NOBRACE_LEN 40
890 #define FID_LEN (FID_NOBRACE_LEN + 2)
891 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
892 #define DFID "["DFID_NOBRACE"]"
893 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
895 /* scanf input parse format for fids in DFID_NOBRACE format
896 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
897 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
898 #define SFID "0x%llx:0x%x:0x%x"
899 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
901 /********* Quotas **********/
903 #define LUSTRE_QUOTABLOCK_BITS 10
904 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
906 static inline __u64 lustre_stoqb(size_t space)
908 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
911 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
912 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
913 #define Q_GETOINFO 0x800102 /* get obd quota info */
914 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
915 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
917 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
918 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
919 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
920 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
921 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
922 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
923 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
924 /* lustre-specific control commands */
925 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
926 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
927 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
928 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
930 /* In the current Lustre implementation, the grace time is either the time
931 * or the timestamp to be used after some quota ID exceeds the soft limt,
932 * 48 bits should be enough, its high 16 bits can be used as quota flags.
934 #define LQUOTA_GRACE_BITS 48
935 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
936 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
937 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
938 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
939 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
941 /* different quota flags */
943 /* the default quota flag, the corresponding quota ID will use the default
944 * quota setting, the hardlimit and softlimit of its quota record in the global
945 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
946 * and high 16 bits will contain this flag (see above comment).
948 #define LQUOTA_FLAG_DEFAULT 0x0001
950 #define ALLQUOTA 255 /* set all quota */
951 static inline char *qtype_name(int qtype)
964 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
965 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
968 #define N_PERMS_MAX 64
970 struct perm_downcall_data {
976 struct identity_downcall_data {
983 struct perm_downcall_data idd_perms[N_PERMS_MAX];
987 struct sepol_downcall_data {
989 time_t sdd_sepol_mtime;
994 #ifdef NEED_QUOTA_DEFS
996 #define QIF_BLIMITS 1
998 #define QIF_ILIMITS 4
1000 #define QIF_BTIME 16
1001 #define QIF_ITIME 32
1002 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1003 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1004 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1005 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1008 #endif /* !__KERNEL__ */
1010 /* lustre volatile file support
1011 * file name header: ".^L^S^T^R:volatile"
1013 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1014 #define LUSTRE_VOLATILE_HDR_LEN 14
1016 enum lustre_quota_version {
1020 /* XXX: same as if_dqinfo struct in kernel */
1028 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1030 __u64 dqb_bhardlimit;
1031 __u64 dqb_bsoftlimit;
1033 __u64 dqb_ihardlimit;
1034 __u64 dqb_isoftlimit;
1035 __u64 dqb_curinodes;
1049 struct if_quotactl {
1056 struct obd_dqinfo qc_dqinfo;
1057 struct obd_dqblk qc_dqblk;
1059 struct obd_uuid obd_uuid;
1062 /* swap layout flags */
1063 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1064 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1065 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1066 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1067 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1069 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1070 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1071 struct lustre_swap_layouts {
1079 /** Bit-mask of valid attributes */
1080 /* The LA_* flags are written to disk as part of the ChangeLog records
1081 * so they are part of the on-disk and network protocol, and cannot be changed.
1082 * Only the first 12 bits are currently saved.
1085 LA_ATIME = 1 << 0, /* 0x00001 */
1086 LA_MTIME = 1 << 1, /* 0x00002 */
1087 LA_CTIME = 1 << 2, /* 0x00004 */
1088 LA_SIZE = 1 << 3, /* 0x00008 */
1089 LA_MODE = 1 << 4, /* 0x00010 */
1090 LA_UID = 1 << 5, /* 0x00020 */
1091 LA_GID = 1 << 6, /* 0x00040 */
1092 LA_BLOCKS = 1 << 7, /* 0x00080 */
1093 LA_TYPE = 1 << 8, /* 0x00100 */
1094 LA_FLAGS = 1 << 9, /* 0x00200 */
1095 LA_NLINK = 1 << 10, /* 0x00400 */
1096 LA_RDEV = 1 << 11, /* 0x00800 */
1097 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1098 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1099 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1100 LA_PROJID = 1 << 15, /* 0x08000 */
1101 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1102 LA_LSIZE = 1 << 17, /* 0x20000 */
1103 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1105 * Attributes must be transmitted to OST objects
1107 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1110 #define MDS_FMODE_READ 00000001
1111 #define MDS_FMODE_WRITE 00000002
1113 #define MDS_FMODE_CLOSED 00000000
1114 #define MDS_FMODE_EXEC 00000004
1115 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1116 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1117 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1119 #define MDS_OPEN_CREATED 00000010
1120 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1122 #define MDS_OPEN_CREAT 00000100
1123 #define MDS_OPEN_EXCL 00000200
1124 #define MDS_OPEN_TRUNC 00001000
1125 #define MDS_OPEN_APPEND 00002000
1126 #define MDS_OPEN_SYNC 00010000
1127 #define MDS_OPEN_DIRECTORY 00200000
1129 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1130 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1131 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1132 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1133 * We do not support JOIN FILE
1134 * anymore, reserve this flags
1135 * just for preventing such bit
1138 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1139 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1140 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1141 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1142 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1144 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1146 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1147 * delegation, succeed if it's not
1148 * being opened with conflict mode.
1150 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1152 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1154 /* lustre internal open flags, which should not be set from user space */
1155 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1156 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1157 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1158 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC)
1161 /********* Changelogs **********/
1162 /** Changelog record types */
1163 enum changelog_rec_type {
1166 CL_CREATE = 1, /* namespace */
1167 CL_MKDIR = 2, /* namespace */
1168 CL_HARDLINK = 3, /* namespace */
1169 CL_SOFTLINK = 4, /* namespace */
1170 CL_MKNOD = 5, /* namespace */
1171 CL_UNLINK = 6, /* namespace */
1172 CL_RMDIR = 7, /* namespace */
1173 CL_RENAME = 8, /* namespace */
1174 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1175 CL_OPEN = 10, /* not currently used */
1176 CL_CLOSE = 11, /* may be written to log only with mtime change */
1177 CL_LAYOUT = 12, /* file layout/striping modified */
1181 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1182 CL_HSM = 16, /* HSM specific events, see flags */
1183 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1187 CL_FLRW = 21, /* FLR: file was firstly written */
1188 CL_RESYNC = 22, /* FLR: file was resync-ed */
1190 CL_DN_OPEN = 24, /* denied open */
1194 static inline const char *changelog_type2str(int type) {
1195 static const char *changelog_str[] = {
1196 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1197 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1198 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1199 "FLRW", "RESYNC","GXATR", "NOPEN",
1202 if (type >= 0 && type < CL_LAST)
1203 return changelog_str[type];
1207 /* 12 bits of per-record data can be stored in the bottom of the flags */
1208 #define CLF_FLAGSHIFT 12
1209 enum changelog_rec_flags {
1210 CLF_VERSION = 0x1000,
1211 CLF_RENAME = 0x2000,
1213 CLF_EXTRA_FLAGS = 0x8000,
1214 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1216 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1217 CLF_VERMASK = ~CLF_FLAGMASK,
1221 /* Anything under the flagmask may be per-type (if desired) */
1222 /* Flags for unlink */
1223 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1224 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1225 /* HSM cleaning needed */
1226 /* Flags for rename */
1227 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1229 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1230 * has an archive in backend */
1233 /* 12b used (from high weight to low weight):
1238 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1239 #define CLF_HSM_ERR_H 6
1240 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1241 #define CLF_HSM_EVENT_H 9
1242 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1243 #define CLF_HSM_FLAG_H 11
1244 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1245 #define CLF_HSM_SPARE_H 15
1246 #define CLF_HSM_LAST 15
1248 /* Remove bits higher than _h, then extract the value
1249 * between _h and _l by shifting lower weigth to bit 0. */
1250 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1251 >> (CLF_HSM_LAST - _h + _l))
1253 #define CLF_HSM_SUCCESS 0x00
1254 #define CLF_HSM_MAXERROR 0x7E
1255 #define CLF_HSM_ERROVERFLOW 0x7F
1257 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1259 /* 3 bits field => 8 values allowed */
1271 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1273 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1277 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1280 *clf_flags |= (he << CLF_HSM_EVENT_L);
1283 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1285 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1288 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1291 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1294 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1296 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1299 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1302 *clf_flags |= (error << CLF_HSM_ERR_L);
1305 enum changelog_rec_extra_flags {
1307 CLFE_UIDGID = 0x0001,
1310 CLFE_XATTR = 0x0008,
1311 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1314 enum changelog_send_flag {
1315 /* Not yet implemented */
1316 CHANGELOG_FLAG_FOLLOW = 0x01,
1317 /* Blocking IO makes sense in case of slow user parsing of the records,
1318 * but it also prevents us from cleaning up if the records are not
1320 CHANGELOG_FLAG_BLOCK = 0x02,
1321 /* Pack jobid into the changelog records if available. */
1322 CHANGELOG_FLAG_JOBID = 0x04,
1323 /* Pack additional flag bits into the changelog record */
1324 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1327 enum changelog_send_extra_flag {
1328 /* Pack uid/gid into the changelog record */
1329 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1330 /* Pack nid into the changelog record */
1331 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1332 /* Pack open mode into the changelog record */
1333 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1334 /* Pack xattr name into the changelog record */
1335 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1338 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1339 changelog_rec_offset(CLF_SUPPORTED, \
1342 /* 31 usable bytes string + null terminator. */
1343 #define LUSTRE_JOBID_SIZE 32
1345 /* This is the minimal changelog record. It can contain extensions
1346 * such as rename fields or process jobid. Its exact content is described
1347 * by the cr_flags and cr_extra_flags.
1349 * Extensions are packed in the same order as their corresponding flags,
1350 * then in the same order as their corresponding extra flags.
1352 struct changelog_rec {
1354 __u16 cr_flags; /**< \a changelog_rec_flags */
1355 __u32 cr_type; /**< \a changelog_rec_type */
1356 __u64 cr_index; /**< changelog record number */
1357 __u64 cr_prev; /**< last index for this target fid */
1360 struct lu_fid cr_tfid; /**< target fid */
1361 __u32 cr_markerflags; /**< CL_MARK flags */
1363 struct lu_fid cr_pfid; /**< parent fid */
1366 /* Changelog extension for RENAME. */
1367 struct changelog_ext_rename {
1368 struct lu_fid cr_sfid; /**< source fid, or zero */
1369 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1372 /* Changelog extension to include JOBID. */
1373 struct changelog_ext_jobid {
1374 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1377 /* Changelog extension to include additional flags. */
1378 struct changelog_ext_extra_flags {
1379 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1382 /* Changelog extra extension to include UID/GID. */
1383 struct changelog_ext_uidgid {
1388 /* Changelog extra extension to include NID. */
1389 struct changelog_ext_nid {
1390 /* have __u64 instead of lnet_nid_t type for use by client api */
1392 /* for use when IPv6 support is added */
1397 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1398 struct changelog_ext_openmode {
1402 /* Changelog extra extension to include xattr */
1403 struct changelog_ext_xattr {
1404 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1407 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1408 const struct changelog_rec *rec);
1410 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1411 enum changelog_rec_extra_flags cref)
1413 size_t size = sizeof(struct changelog_rec);
1415 if (crf & CLF_RENAME)
1416 size += sizeof(struct changelog_ext_rename);
1418 if (crf & CLF_JOBID)
1419 size += sizeof(struct changelog_ext_jobid);
1421 if (crf & CLF_EXTRA_FLAGS) {
1422 size += sizeof(struct changelog_ext_extra_flags);
1423 if (cref & CLFE_UIDGID)
1424 size += sizeof(struct changelog_ext_uidgid);
1425 if (cref & CLFE_NID)
1426 size += sizeof(struct changelog_ext_nid);
1427 if (cref & CLFE_OPEN)
1428 size += sizeof(struct changelog_ext_openmode);
1429 if (cref & CLFE_XATTR)
1430 size += sizeof(struct changelog_ext_xattr);
1436 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1438 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1440 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1441 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1443 return changelog_rec_offset(rec->cr_flags, cref);
1446 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1448 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1452 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1454 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1456 return (struct changelog_ext_rename *)((char *)rec +
1457 changelog_rec_offset(crf,
1461 /* The jobid follows the rename extension, if present */
1463 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1465 enum changelog_rec_flags crf = rec->cr_flags &
1466 (CLF_VERSION | CLF_RENAME);
1468 return (struct changelog_ext_jobid *)((char *)rec +
1469 changelog_rec_offset(crf,
1473 /* The additional flags follow the rename and jobid extensions, if present */
1475 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1476 const struct changelog_rec *rec)
1478 enum changelog_rec_flags crf = rec->cr_flags &
1479 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1481 return (struct changelog_ext_extra_flags *)((char *)rec +
1482 changelog_rec_offset(crf,
1486 /* The uid/gid is the first extra extension */
1488 struct changelog_ext_uidgid *changelog_rec_uidgid(
1489 const struct changelog_rec *rec)
1491 enum changelog_rec_flags crf = rec->cr_flags &
1492 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1494 return (struct changelog_ext_uidgid *)((char *)rec +
1495 changelog_rec_offset(crf,
1499 /* The nid is the second extra extension */
1501 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1503 enum changelog_rec_flags crf = rec->cr_flags &
1504 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1505 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1507 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1508 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1511 return (struct changelog_ext_nid *)((char *)rec +
1512 changelog_rec_offset(crf, cref));
1515 /* The OPEN mode is the third extra extension */
1517 struct changelog_ext_openmode *changelog_rec_openmode(
1518 const struct changelog_rec *rec)
1520 enum changelog_rec_flags crf = rec->cr_flags &
1521 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1522 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1524 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1525 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1526 (CLFE_UIDGID | CLFE_NID);
1528 return (struct changelog_ext_openmode *)((char *)rec +
1529 changelog_rec_offset(crf, cref));
1532 /* The xattr name is the fourth extra extension */
1534 struct changelog_ext_xattr *changelog_rec_xattr(
1535 const struct changelog_rec *rec)
1537 enum changelog_rec_flags crf = rec->cr_flags &
1538 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1539 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1541 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1542 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1543 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1545 return (struct changelog_ext_xattr *)((char *)rec +
1546 changelog_rec_offset(crf, cref));
1549 /* The name follows the rename, jobid and extra flags extns, if present */
1550 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1552 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1554 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1555 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1557 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1558 cref & CLFE_SUPPORTED);
1561 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1563 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1566 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1568 char *cr_name = changelog_rec_name(rec);
1570 return cr_name + strlen(cr_name) + 1;
1574 * Remap a record to the desired format as specified by the crf flags.
1575 * The record must be big enough to contain the final remapped version.
1576 * Superfluous extension fields are removed and missing ones are added
1577 * and zeroed. The flags of the record are updated accordingly.
1579 * The jobid and rename extensions can be added to a record, to match the
1580 * format an application expects, typically. In this case, the newly added
1581 * fields will be zeroed.
1582 * The Jobid field can be removed, to guarantee compatibility with older
1583 * clients that don't expect this field in the records they process.
1585 * The following assumptions are being made:
1586 * - CLF_RENAME will not be removed
1587 * - CLF_JOBID will not be added without CLF_RENAME being added too
1588 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1590 * @param[in,out] rec The record to remap.
1591 * @param[in] crf_wanted Flags describing the desired extensions.
1592 * @param[in] cref_want Flags describing the desired extra extensions.
1594 static inline void changelog_remap_rec(struct changelog_rec *rec,
1595 enum changelog_rec_flags crf_wanted,
1596 enum changelog_rec_extra_flags cref_want)
1598 char *xattr_mov = NULL;
1599 char *omd_mov = NULL;
1600 char *nid_mov = NULL;
1601 char *uidgid_mov = NULL;
1605 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1607 crf_wanted &= CLF_SUPPORTED;
1608 cref_want &= CLFE_SUPPORTED;
1610 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1611 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1612 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1613 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1619 /* First move the variable-length name field */
1620 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1621 changelog_rec_name(rec), rec->cr_namelen);
1623 /* Locations of extensions in the remapped record */
1624 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1625 xattr_mov = (char *)rec +
1626 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1627 cref_want & ~CLFE_XATTR);
1628 omd_mov = (char *)rec +
1629 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1630 cref_want & ~(CLFE_OPEN |
1632 nid_mov = (char *)rec +
1633 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1634 cref_want & ~(CLFE_NID |
1637 uidgid_mov = (char *)rec +
1638 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1639 cref_want & ~(CLFE_UIDGID |
1643 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1646 ef_mov = (char *)rec +
1647 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1649 jid_mov = (char *)rec +
1650 changelog_rec_offset(crf_wanted &
1651 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1653 rnm_mov = (char *)rec +
1654 changelog_rec_offset(crf_wanted &
1660 /* Move the extension fields to the desired positions */
1661 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1662 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1663 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1664 memmove(xattr_mov, changelog_rec_xattr(rec),
1665 sizeof(struct changelog_ext_xattr));
1667 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1668 memmove(omd_mov, changelog_rec_openmode(rec),
1669 sizeof(struct changelog_ext_openmode));
1671 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1672 memmove(nid_mov, changelog_rec_nid(rec),
1673 sizeof(struct changelog_ext_nid));
1675 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1676 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1677 sizeof(struct changelog_ext_uidgid));
1679 memmove(ef_mov, changelog_rec_extra_flags(rec),
1680 sizeof(struct changelog_ext_extra_flags));
1683 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1684 memmove(jid_mov, changelog_rec_jobid(rec),
1685 sizeof(struct changelog_ext_jobid));
1687 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1688 memmove(rnm_mov, changelog_rec_rename(rec),
1689 sizeof(struct changelog_ext_rename));
1691 /* Clear newly added fields */
1692 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1693 !(cref & CLFE_XATTR))
1694 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1696 if (omd_mov && (cref_want & CLFE_OPEN) &&
1697 !(cref & CLFE_OPEN))
1698 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1700 if (nid_mov && (cref_want & CLFE_NID) &&
1702 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1704 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1705 !(cref & CLFE_UIDGID))
1706 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1708 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1709 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1710 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1712 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1713 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1715 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1716 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1718 /* Update the record's flags accordingly */
1719 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1720 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1721 changelog_rec_extra_flags(rec)->cr_extra_flags =
1722 changelog_rec_extra_flags(rec)->cr_extra_flags |
1726 enum changelog_message_type {
1727 CL_RECORD = 10, /* message is a changelog_rec */
1728 CL_EOF = 11, /* at end of current changelog */
1731 /********* Misc **********/
1733 struct ioc_data_version {
1735 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1736 __u32 idv_flags; /* enum ioc_data_version_flags */
1739 enum ioc_data_version_flags {
1740 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1741 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1745 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1748 #define dot_lustre_name ".lustre"
1751 /********* HSM **********/
1753 /** HSM per-file state
1754 * See HSM_FLAGS below.
1757 HS_NONE = 0x00000000,
1758 HS_EXISTS = 0x00000001,
1759 HS_DIRTY = 0x00000002,
1760 HS_RELEASED = 0x00000004,
1761 HS_ARCHIVED = 0x00000008,
1762 HS_NORELEASE = 0x00000010,
1763 HS_NOARCHIVE = 0x00000020,
1764 HS_LOST = 0x00000040,
1767 /* HSM user-setable flags. */
1768 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1770 /* Other HSM flags. */
1771 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1774 * All HSM-related possible flags that could be applied to a file.
1775 * This should be kept in sync with hsm_states.
1777 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1780 * HSM request progress state
1782 enum hsm_progress_states {
1789 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1792 case HPS_WAITING: return "waiting";
1793 case HPS_RUNNING: return "running";
1794 case HPS_DONE: return "done";
1795 default: return "unknown";
1802 } __attribute__((packed));
1805 * Current HSM states of a Lustre file.
1807 * This structure purpose is to be sent to user-space mainly. It describes the
1808 * current HSM flags and in-progress action.
1810 struct hsm_user_state {
1811 /** Current HSM states, from enum hsm_states. */
1813 __u32 hus_archive_id;
1814 /** The current undergoing action, if there is one */
1815 __u32 hus_in_progress_state;
1816 __u32 hus_in_progress_action;
1817 struct hsm_extent hus_in_progress_location;
1818 char hus_extended_info[];
1821 struct hsm_state_set_ioc {
1822 struct lu_fid hssi_fid;
1824 __u64 hssi_clearmask;
1828 * This structure describes the current in-progress action for a file.
1829 * it is retuned to user space and send over the wire
1831 struct hsm_current_action {
1832 /** The current undergoing action, if there is one */
1833 /* state is one of hsm_progress_states */
1835 /* action is one of hsm_user_action */
1837 struct hsm_extent hca_location;
1840 /***** HSM user requests ******/
1841 /* User-generated (lfs/ioctl) request types */
1842 enum hsm_user_action {
1843 HUA_NONE = 1, /* no action (noop) */
1844 HUA_ARCHIVE = 10, /* copy to hsm */
1845 HUA_RESTORE = 11, /* prestage */
1846 HUA_RELEASE = 12, /* drop ost objects */
1847 HUA_REMOVE = 13, /* remove from archive */
1848 HUA_CANCEL = 14 /* cancel a request */
1851 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1854 case HUA_NONE: return "NOOP";
1855 case HUA_ARCHIVE: return "ARCHIVE";
1856 case HUA_RESTORE: return "RESTORE";
1857 case HUA_RELEASE: return "RELEASE";
1858 case HUA_REMOVE: return "REMOVE";
1859 case HUA_CANCEL: return "CANCEL";
1860 default: return "UNKNOWN";
1865 * List of hr_flags (bit field)
1867 #define HSM_FORCE_ACTION 0x0001
1868 /* used by CT, cannot be set by user */
1869 #define HSM_GHOST_COPY 0x0002
1872 * Contains all the fixed part of struct hsm_user_request.
1875 struct hsm_request {
1876 __u32 hr_action; /* enum hsm_user_action */
1877 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1878 __u64 hr_flags; /* request flags */
1879 __u32 hr_itemcount; /* item count in hur_user_item vector */
1883 struct hsm_user_item {
1884 struct lu_fid hui_fid;
1885 struct hsm_extent hui_extent;
1886 } __attribute__((packed));
1888 struct hsm_user_request {
1889 struct hsm_request hur_request;
1890 struct hsm_user_item hur_user_item[0];
1891 /* extra data blob at end of struct (after all
1892 * hur_user_items), only use helpers to access it
1894 } __attribute__((packed));
1896 /** Return pointer to data field in a hsm user request */
1897 static inline void *hur_data(struct hsm_user_request *hur)
1899 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1903 * Compute the current length of the provided hsm_user_request. This returns -1
1904 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1906 * return -1 on bounds check error.
1908 static inline ssize_t hur_len(struct hsm_user_request *hur)
1912 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1913 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1914 (__u64)hur->hur_request.hr_itemcount *
1915 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1917 if (size != (ssize_t)size)
1923 /****** HSM RPCs to copytool *****/
1924 /* Message types the copytool may receive */
1925 enum hsm_message_type {
1926 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1929 /* Actions the copytool may be instructed to take for a given action_item */
1930 enum hsm_copytool_action {
1931 HSMA_NONE = 10, /* no action */
1932 HSMA_ARCHIVE = 20, /* arbitrary offset */
1938 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1941 case HSMA_NONE: return "NOOP";
1942 case HSMA_ARCHIVE: return "ARCHIVE";
1943 case HSMA_RESTORE: return "RESTORE";
1944 case HSMA_REMOVE: return "REMOVE";
1945 case HSMA_CANCEL: return "CANCEL";
1946 default: return "UNKNOWN";
1950 /* Copytool item action description */
1951 struct hsm_action_item {
1952 __u32 hai_len; /* valid size of this struct */
1953 __u32 hai_action; /* hsm_copytool_action, but use known size */
1954 struct lu_fid hai_fid; /* Lustre FID to operate on */
1955 struct lu_fid hai_dfid; /* fid used for data access */
1956 struct hsm_extent hai_extent; /* byte range to operate on */
1957 __u64 hai_cookie; /* action cookie from coordinator */
1958 __u64 hai_gid; /* grouplock id */
1959 char hai_data[0]; /* variable length */
1960 } __attribute__((packed));
1963 * helper function which print in hexa the first bytes of
1966 * \param hai [IN] record to print
1967 * \param buffer [IN,OUT] buffer to write the hex string to
1968 * \param len [IN] max buffer length
1972 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1973 char *buffer, size_t len)
1980 data_len = hai->hai_len - sizeof(*hai);
1981 for (i = 0; (i < data_len) && (len > 2); i++) {
1982 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1992 /* Copytool action list */
1993 #define HAL_VERSION 1
1994 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1995 struct hsm_action_list {
1997 __u32 hal_count; /* number of hai's to follow */
1998 __u64 hal_compound_id; /* returned by coordinator, ignored */
2000 __u32 hal_archive_id; /* which archive backend */
2002 char hal_fsname[0]; /* null-terminated */
2003 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2004 boundaries. See hai_zero */
2005 } __attribute__((packed));
2007 /* Return pointer to first hai in action list */
2008 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2010 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2012 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2015 /* Return pointer to next hai */
2016 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2018 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2020 return (struct hsm_action_item *)((char *)hai + offset);
2023 /* Return size of an hsm_action_list */
2024 static inline size_t hal_size(struct hsm_action_list *hal)
2028 struct hsm_action_item *hai;
2030 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2031 hai = hai_first(hal);
2032 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2033 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2039 * describe the attributes to be set on imported file
2041 struct hsm_user_import {
2050 __u32 hui_archive_id;
2053 /* Copytool progress reporting */
2054 #define HP_FLAG_COMPLETED 0x01
2055 #define HP_FLAG_RETRY 0x02
2057 struct hsm_progress {
2058 struct lu_fid hp_fid;
2060 struct hsm_extent hp_extent;
2062 __u16 hp_errval; /* positive val */
2067 __u64 hc_data_version;
2069 __u16 hc_errval; /* positive val */
2071 struct hsm_action_item hc_hai;
2075 enum llapi_json_types {
2076 LLAPI_JSON_INTEGER = 1,
2082 struct llapi_json_item {
2091 struct llapi_json_item *lji_next;
2094 struct llapi_json_item_list {
2095 int ljil_item_count;
2096 struct llapi_json_item *ljil_items;
2099 enum lu_ladvise_type {
2100 LU_LADVISE_INVALID = 0,
2101 LU_LADVISE_WILLREAD = 1,
2102 LU_LADVISE_DONTNEED = 2,
2103 LU_LADVISE_LOCKNOEXPAND = 3,
2104 LU_LADVISE_LOCKAHEAD = 4,
2108 #define LU_LADVISE_NAMES { \
2109 [LU_LADVISE_WILLREAD] = "willread", \
2110 [LU_LADVISE_DONTNEED] = "dontneed", \
2111 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2112 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2115 /* This is the userspace argument for ladvise. It is currently the same as
2116 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2117 * may need info which is only used locally. */
2118 struct llapi_lu_ladvise {
2119 __u16 lla_advice; /* advice type */
2120 __u16 lla_value1; /* values for different advice types */
2122 __u64 lla_start; /* first byte of extent for advice */
2123 __u64 lla_end; /* last byte of extent for advice */
2129 LF_ASYNC = 0x00000001,
2130 LF_UNSET = 0x00000002,
2133 #define LADVISE_MAGIC 0x1ADF1CE0
2134 /* Masks of valid flags for each advice */
2135 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2136 /* Flags valid for all advices not explicitly specified */
2137 #define LF_DEFAULT_MASK LF_ASYNC
2139 #define LF_MASK (LF_ASYNC | LF_UNSET)
2141 #define lla_lockahead_mode lla_value1
2142 #define lla_peradvice_flags lla_value2
2143 #define lla_lockahead_result lla_value3
2145 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2146 * is used on the wire. It is defined separately as we may need info which is
2147 * only used locally. */
2148 struct llapi_ladvise_hdr {
2149 __u32 lah_magic; /* LADVISE_MAGIC */
2150 __u32 lah_count; /* number of advices */
2151 __u64 lah_flags; /* from enum ladvise_flag */
2152 __u32 lah_value1; /* unused */
2153 __u32 lah_value2; /* unused */
2154 __u64 lah_value3; /* unused */
2155 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2158 #define LAH_COUNT_MAX (1024)
2162 SK_CRYPT_INVALID = -1,
2164 SK_CRYPT_AES256_CTR = 1,
2168 SK_HMAC_INVALID = -1,
2174 struct sk_crypt_type {
2175 const char *sct_name;
2179 struct sk_hmac_type {
2180 const char *sht_name;
2184 enum lock_mode_user {
2190 #define LOCK_MODE_NAMES { \
2191 [MODE_READ_USER] = "READ",\
2192 [MODE_WRITE_USER] = "WRITE"\
2195 enum lockahead_results {
2196 LLA_RESULT_SENT = 0,
2197 LLA_RESULT_DIFFERENT,
2201 enum lu_heat_flag_bit {
2202 LU_HEAT_FLAG_BIT_INVALID = 0,
2203 LU_HEAT_FLAG_BIT_OFF,
2204 LU_HEAT_FLAG_BIT_CLEAR,
2207 #define LU_HEAT_FLAG_CLEAR (1 << LU_HEAT_FLAG_BIT_CLEAR)
2208 #define LU_HEAT_FLAG_OFF (1 << LU_HEAT_FLAG_BIT_OFF)
2210 enum obd_heat_type {
2211 OBD_HEAT_READSAMPLE = 0,
2212 OBD_HEAT_WRITESAMPLE = 1,
2213 OBD_HEAT_READBYTE = 2,
2214 OBD_HEAT_WRITEBYTE = 3,
2218 #define LU_HEAT_NAMES { \
2219 [OBD_HEAT_READSAMPLE] = "readsample", \
2220 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2221 [OBD_HEAT_READBYTE] = "readbyte", \
2222 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2231 #if defined(__cplusplus)
2235 /** @} lustreuser */
2237 #endif /* _LUSTRE_USER_H */