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 _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
512 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
514 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
516 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
517 * files, but are unlikely to be used in practice and are not harmful if
518 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
519 * devices and are safe for use on new files. See LU-4209. */
520 /* To be compatible with old statically linked binary we keep the check for
521 * the older 0100000000 flag. This is already removed upstream. LU-812. */
522 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
524 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
526 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
527 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
528 O_LOV_DELAY_CREATE_MASK)
530 #define LL_FILE_IGNORE_LOCK 0x00000001
531 #define LL_FILE_GROUP_LOCKED 0x00000002
532 #define LL_FILE_READAHEA 0x00000004
533 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
534 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
536 #define LOV_USER_MAGIC_V1 0x0BD10BD0
537 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
538 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
539 #define LOV_USER_MAGIC_V3 0x0BD30BD0
540 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
541 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
542 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
544 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
545 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
546 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
548 #define LOV_PATTERN_NONE 0x000
549 #define LOV_PATTERN_RAID0 0x001
550 #define LOV_PATTERN_RAID1 0x002
551 #define LOV_PATTERN_MDT 0x100
552 #define LOV_PATTERN_CMOBD 0x200
554 #define LOV_PATTERN_F_MASK 0xffff0000
555 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
556 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
557 #define LOV_PATTERN_DEFAULT 0xffffffff
559 static inline bool lov_pattern_supported(__u32 pattern)
561 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
562 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
565 #define LOV_MAXPOOLNAME 15
566 #define LOV_POOLNAMEF "%.15s"
568 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
569 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
570 #define LOV_MAX_STRIPE_COUNT_OLD 160
571 /* This calculation is crafted so that input of 4096 will result in 160
572 * which in turn is equal to old maximal stripe count.
573 * XXX: In fact this is too simpified for now, what it also need is to get
574 * ea_type argument to clearly know how much space each stripe consumes.
576 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
577 * allocation that is sufficient for the current generation of systems.
579 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
580 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
581 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
582 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
584 #define XATTR_LUSTRE_PREFIX "lustre."
585 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
587 #define lov_user_ost_data lov_user_ost_data_v1
588 struct lov_user_ost_data_v1 { /* per-stripe data structure */
589 struct ost_id l_ost_oi; /* OST object ID */
590 __u32 l_ost_gen; /* generation of this OST index */
591 __u32 l_ost_idx; /* OST index in LOV */
592 } __attribute__((packed));
594 #define lov_user_md lov_user_md_v1
595 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
596 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
597 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
598 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
599 __u32 lmm_stripe_size; /* size of stripe in bytes */
600 __u16 lmm_stripe_count; /* num stripes in use for this object */
602 __u16 lmm_stripe_offset; /* starting stripe offset in
603 * lmm_objects, use when writing */
604 __u16 lmm_layout_gen; /* layout generation number
605 * used when reading */
607 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
608 } __attribute__((packed, __may_alias__));
610 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
611 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
612 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
613 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
614 __u32 lmm_stripe_size; /* size of stripe in bytes */
615 __u16 lmm_stripe_count; /* num stripes in use for this object */
617 __u16 lmm_stripe_offset; /* starting stripe offset in
618 * lmm_objects, use when writing */
619 __u16 lmm_layout_gen; /* layout generation number
620 * used when reading */
622 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
623 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
624 } __attribute__((packed));
631 #define DEXT "[%#llx, %#llx)"
632 #define PEXT(ext) (ext)->e_start, (ext)->e_end
634 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
635 struct lu_extent *e2)
637 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
640 static inline bool lu_extent_is_whole(struct lu_extent *e)
642 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
645 enum lov_comp_md_entry_flags {
646 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
647 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
648 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
649 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
650 LCME_FL_OFFLINE = 0x00000008, /* Not used */
651 LCME_FL_INIT = 0x00000010, /* instantiated */
652 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
653 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
654 won't be stored on disk */
657 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
658 LCME_FL_PREF_RW | LCME_FL_NOSYNC)
659 /* The flags can be set by users at mirror creation time. */
660 #define LCME_USER_FLAGS (LCME_FL_PREF_RW)
662 /* The flags are for mirrors */
663 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
665 /* These flags have meaning when set in a default layout and will be inherited
666 * from the default/template layout set on a directory.
668 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC)
670 /* the highest bit in obdo::o_layout_version is used to mark if the file is
672 #define LU_LAYOUT_RESYNC LCME_FL_NEG
674 /* lcme_id can be specified as certain flags, and the the first
675 * bit of lcme_id is used to indicate that the ID is representing
676 * certain LCME_FL_* but not a real ID. Which implies we can have
677 * at most 31 flags (see LCME_FL_XXX). */
680 LCME_ID_MAX = 0x7FFFFFFF,
681 LCME_ID_ALL = 0xFFFFFFFF,
682 LCME_ID_NOT_ID = LCME_FL_NEG
685 #define LCME_ID_MASK LCME_ID_MAX
687 struct lov_comp_md_entry_v1 {
688 __u32 lcme_id; /* unique id of component */
689 __u32 lcme_flags; /* LCME_FL_XXX */
690 struct lu_extent lcme_extent; /* file extent for component */
691 __u32 lcme_offset; /* offset of component blob,
692 start from lov_comp_md_v1 */
693 __u32 lcme_size; /* size of component blob */
694 __u32 lcme_layout_gen;
695 __u64 lcme_timestamp; /* snapshot time if applicable*/
696 __u32 lcme_padding_1;
697 } __attribute__((packed));
699 #define SEQ_ID_MAX 0x0000FFFF
700 #define SEQ_ID_MASK SEQ_ID_MAX
701 /* bit 30:16 of lcme_id is used to store mirror id */
702 #define MIRROR_ID_MASK 0x7FFF0000
703 #define MIRROR_ID_NEG 0x8000
704 #define MIRROR_ID_SHIFT 16
706 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
708 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
711 static inline __u16 mirror_id_of(__u32 id)
713 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
717 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
719 enum lov_comp_md_flags {
720 /* the least 2 bits are used by FLR to record file state */
723 LCM_FL_WRITE_PENDING = 2,
724 LCM_FL_SYNC_PENDING = 3,
725 LCM_FL_FLR_MASK = 0x3,
728 struct lov_comp_md_v1 {
729 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
730 __u32 lcm_size; /* overall size including this struct */
731 __u32 lcm_layout_gen;
733 __u16 lcm_entry_count;
734 /* lcm_mirror_count stores the number of actual mirrors minus 1,
735 * so that non-flr files will have value 0 meaning 1 mirror. */
736 __u16 lcm_mirror_count;
737 __u16 lcm_padding1[3];
739 struct lov_comp_md_entry_v1 lcm_entries[0];
740 } __attribute__((packed));
742 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
744 if (stripes == (__u16)-1)
747 if (lmm_magic == LOV_USER_MAGIC_V1)
748 return sizeof(struct lov_user_md_v1) +
749 stripes * sizeof(struct lov_user_ost_data_v1);
750 return sizeof(struct lov_user_md_v3) +
751 stripes * sizeof(struct lov_user_ost_data_v1);
754 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
755 * use this. It is unsafe to #define those values in this header as it
756 * is possible the application has already #included <sys/stat.h>. */
757 #ifdef HAVE_LOV_USER_MDS_DATA
758 #define lov_user_mds_data lov_user_mds_data_v1
759 struct lov_user_mds_data_v1 {
760 lstat_t lmd_st; /* MDS stat struct */
761 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
762 } __attribute__((packed));
765 struct lmv_user_mds_data {
766 struct lu_fid lum_fid;
772 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
773 LMV_HASH_TYPE_ALL_CHARS = 1,
774 LMV_HASH_TYPE_FNV_1A_64 = 2,
778 #define LMV_HASH_NAME_ALL_CHARS "all_char"
779 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
781 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
783 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
784 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
785 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
786 #define lmv_user_md lmv_user_md_v1
787 struct lmv_user_md_v1 {
788 __u32 lum_magic; /* must be the first field */
789 __u32 lum_stripe_count; /* dirstripe count */
790 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
791 __u32 lum_hash_type; /* Dir stripe policy */
792 __u32 lum_type; /* LMV type: default or normal */
796 char lum_pool_name[LOV_MAXPOOLNAME + 1];
797 struct lmv_user_mds_data lum_objects[0];
798 } __attribute__((packed));
800 static inline int lmv_user_md_size(int stripes, int lmm_magic)
802 int size = sizeof(struct lmv_user_md);
804 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
805 size += stripes * sizeof(struct lmv_user_mds_data);
810 struct ll_recreate_obj {
816 __u64 id; /* holds object id */
817 __u32 generation; /* holds object generation */
818 __u32 f_type; /* holds object type or stripe idx when passing it to
819 * OST for saving into EA. */
827 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
828 const struct obd_uuid *u2)
830 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
833 static inline int obd_uuid_empty(struct obd_uuid *uuid)
835 return uuid->uuid[0] == '\0';
838 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
840 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
841 uuid->uuid[sizeof(*uuid) - 1] = '\0';
844 /* For printf's only, make sure uuid is terminated */
845 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
850 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
851 /* Obviously not safe, but for printfs, no real harm done...
852 we're always null-terminated, even in a race. */
853 static char temp[sizeof(*uuid)];
854 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
855 temp[sizeof(*uuid) - 1] = '\0';
858 return (char *)(uuid->uuid);
861 #define LUSTRE_MAXFSNAME 8
863 /* Extract fsname from uuid (or target name) of a target
864 e.g. (myfs-OST0007_UUID -> myfs)
865 see also deuuidify. */
866 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
870 strncpy(buf, uuid, buflen - 1);
871 buf[buflen - 1] = '\0';
872 p = strrchr(buf, '-');
877 /* printf display format for Lustre FIDs
878 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
879 #define FID_NOBRACE_LEN 40
880 #define FID_LEN (FID_NOBRACE_LEN + 2)
881 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
882 #define DFID "["DFID_NOBRACE"]"
883 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
885 /* scanf input parse format for fids in DFID_NOBRACE format
886 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
887 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
888 #define SFID "0x%llx:0x%x:0x%x"
889 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
891 /********* Quotas **********/
893 #define LUSTRE_QUOTABLOCK_BITS 10
894 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
896 static inline __u64 lustre_stoqb(size_t space)
898 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
901 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
902 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
903 #define Q_GETOINFO 0x800102 /* get obd quota info */
904 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
905 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
907 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
908 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
909 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
910 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
911 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
912 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
913 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
914 /* lustre-specific control commands */
915 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
916 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
917 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
918 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
920 /* In the current Lustre implementation, the grace time is either the time
921 * or the timestamp to be used after some quota ID exceeds the soft limt,
922 * 48 bits should be enough, its high 16 bits can be used as quota flags.
924 #define LQUOTA_GRACE_BITS 48
925 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
926 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
927 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
928 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
929 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
931 /* different quota flags */
933 /* the default quota flag, the corresponding quota ID will use the default
934 * quota setting, the hardlimit and softlimit of its quota record in the global
935 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
936 * and high 16 bits will contain this flag (see above comment).
938 #define LQUOTA_FLAG_DEFAULT 0x0001
940 #define ALLQUOTA 255 /* set all quota */
941 static inline char *qtype_name(int qtype)
954 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
955 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
958 #define N_PERMS_MAX 64
960 struct perm_downcall_data {
966 struct identity_downcall_data {
973 struct perm_downcall_data idd_perms[N_PERMS_MAX];
977 struct sepol_downcall_data {
979 time_t sdd_sepol_mtime;
984 #ifdef NEED_QUOTA_DEFS
986 #define QIF_BLIMITS 1
988 #define QIF_ILIMITS 4
992 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
993 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
994 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
995 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
998 #endif /* !__KERNEL__ */
1000 /* lustre volatile file support
1001 * file name header: ".^L^S^T^R:volatile"
1003 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1004 #define LUSTRE_VOLATILE_HDR_LEN 14
1006 enum lustre_quota_version {
1010 /* XXX: same as if_dqinfo struct in kernel */
1018 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1020 __u64 dqb_bhardlimit;
1021 __u64 dqb_bsoftlimit;
1023 __u64 dqb_ihardlimit;
1024 __u64 dqb_isoftlimit;
1025 __u64 dqb_curinodes;
1039 struct if_quotactl {
1046 struct obd_dqinfo qc_dqinfo;
1047 struct obd_dqblk qc_dqblk;
1049 struct obd_uuid obd_uuid;
1052 /* swap layout flags */
1053 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1054 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1055 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1056 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1057 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1059 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1060 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1061 struct lustre_swap_layouts {
1069 /** Bit-mask of valid attributes */
1070 /* The LA_* flags are written to disk as part of the ChangeLog records
1071 * so they are part of the on-disk and network protocol, and cannot be changed.
1072 * Only the first 12 bits are currently saved.
1075 LA_ATIME = 1 << 0, /* 0x00001 */
1076 LA_MTIME = 1 << 1, /* 0x00002 */
1077 LA_CTIME = 1 << 2, /* 0x00004 */
1078 LA_SIZE = 1 << 3, /* 0x00008 */
1079 LA_MODE = 1 << 4, /* 0x00010 */
1080 LA_UID = 1 << 5, /* 0x00020 */
1081 LA_GID = 1 << 6, /* 0x00040 */
1082 LA_BLOCKS = 1 << 7, /* 0x00080 */
1083 LA_TYPE = 1 << 8, /* 0x00100 */
1084 LA_FLAGS = 1 << 9, /* 0x00200 */
1085 LA_NLINK = 1 << 10, /* 0x00400 */
1086 LA_RDEV = 1 << 11, /* 0x00800 */
1087 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1088 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1089 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1090 LA_PROJID = 1 << 15, /* 0x08000 */
1091 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1092 LA_LSIZE = 1 << 17, /* 0x20000 */
1093 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1095 * Attributes must be transmitted to OST objects
1097 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1100 #define MDS_FMODE_READ 00000001
1101 #define MDS_FMODE_WRITE 00000002
1103 #define MDS_FMODE_CLOSED 00000000
1104 #define MDS_FMODE_EXEC 00000004
1105 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1106 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1107 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1109 #define MDS_OPEN_CREATED 00000010
1110 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1112 #define MDS_OPEN_CREAT 00000100
1113 #define MDS_OPEN_EXCL 00000200
1114 #define MDS_OPEN_TRUNC 00001000
1115 #define MDS_OPEN_APPEND 00002000
1116 #define MDS_OPEN_SYNC 00010000
1117 #define MDS_OPEN_DIRECTORY 00200000
1119 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1120 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1121 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1122 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1123 * We do not support JOIN FILE
1124 * anymore, reserve this flags
1125 * just for preventing such bit
1128 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1129 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1130 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1131 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1132 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1134 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1136 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1137 * delegation, succeed if it's not
1138 * being opened with conflict mode.
1140 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1142 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1144 /* lustre internal open flags, which should not be set from user space */
1145 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1146 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1147 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1148 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC)
1151 /********* Changelogs **********/
1152 /** Changelog record types */
1153 enum changelog_rec_type {
1156 CL_CREATE = 1, /* namespace */
1157 CL_MKDIR = 2, /* namespace */
1158 CL_HARDLINK = 3, /* namespace */
1159 CL_SOFTLINK = 4, /* namespace */
1160 CL_MKNOD = 5, /* namespace */
1161 CL_UNLINK = 6, /* namespace */
1162 CL_RMDIR = 7, /* namespace */
1163 CL_RENAME = 8, /* namespace */
1164 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1165 CL_OPEN = 10, /* not currently used */
1166 CL_CLOSE = 11, /* may be written to log only with mtime change */
1167 CL_LAYOUT = 12, /* file layout/striping modified */
1171 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1172 CL_HSM = 16, /* HSM specific events, see flags */
1173 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1177 CL_FLRW = 21, /* FLR: file was firstly written */
1178 CL_RESYNC = 22, /* FLR: file was resync-ed */
1180 CL_DN_OPEN = 24, /* denied open */
1184 static inline const char *changelog_type2str(int type) {
1185 static const char *changelog_str[] = {
1186 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1187 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1188 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1189 "FLRW", "RESYNC","GXATR", "NOPEN",
1192 if (type >= 0 && type < CL_LAST)
1193 return changelog_str[type];
1197 /* 12 bits of per-record data can be stored in the bottom of the flags */
1198 #define CLF_FLAGSHIFT 12
1199 enum changelog_rec_flags {
1200 CLF_VERSION = 0x1000,
1201 CLF_RENAME = 0x2000,
1203 CLF_EXTRA_FLAGS = 0x8000,
1204 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1206 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1207 CLF_VERMASK = ~CLF_FLAGMASK,
1211 /* Anything under the flagmask may be per-type (if desired) */
1212 /* Flags for unlink */
1213 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1214 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1215 /* HSM cleaning needed */
1216 /* Flags for rename */
1217 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1219 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1220 * has an archive in backend */
1223 /* 12b used (from high weight to low weight):
1228 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1229 #define CLF_HSM_ERR_H 6
1230 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1231 #define CLF_HSM_EVENT_H 9
1232 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1233 #define CLF_HSM_FLAG_H 11
1234 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1235 #define CLF_HSM_SPARE_H 15
1236 #define CLF_HSM_LAST 15
1238 /* Remove bits higher than _h, then extract the value
1239 * between _h and _l by shifting lower weigth to bit 0. */
1240 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1241 >> (CLF_HSM_LAST - _h + _l))
1243 #define CLF_HSM_SUCCESS 0x00
1244 #define CLF_HSM_MAXERROR 0x7E
1245 #define CLF_HSM_ERROVERFLOW 0x7F
1247 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1249 /* 3 bits field => 8 values allowed */
1261 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1263 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1267 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1270 *clf_flags |= (he << CLF_HSM_EVENT_L);
1273 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1275 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1278 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1281 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1284 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1286 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1289 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1292 *clf_flags |= (error << CLF_HSM_ERR_L);
1295 enum changelog_rec_extra_flags {
1297 CLFE_UIDGID = 0x0001,
1300 CLFE_XATTR = 0x0008,
1301 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1304 enum changelog_send_flag {
1305 /* Not yet implemented */
1306 CHANGELOG_FLAG_FOLLOW = 0x01,
1307 /* Blocking IO makes sense in case of slow user parsing of the records,
1308 * but it also prevents us from cleaning up if the records are not
1310 CHANGELOG_FLAG_BLOCK = 0x02,
1311 /* Pack jobid into the changelog records if available. */
1312 CHANGELOG_FLAG_JOBID = 0x04,
1313 /* Pack additional flag bits into the changelog record */
1314 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1317 enum changelog_send_extra_flag {
1318 /* Pack uid/gid into the changelog record */
1319 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1320 /* Pack nid into the changelog record */
1321 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1322 /* Pack open mode into the changelog record */
1323 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1324 /* Pack xattr name into the changelog record */
1325 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1328 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1329 changelog_rec_offset(CLF_SUPPORTED, \
1332 /* 31 usable bytes string + null terminator. */
1333 #define LUSTRE_JOBID_SIZE 32
1335 /* This is the minimal changelog record. It can contain extensions
1336 * such as rename fields or process jobid. Its exact content is described
1337 * by the cr_flags and cr_extra_flags.
1339 * Extensions are packed in the same order as their corresponding flags,
1340 * then in the same order as their corresponding extra flags.
1342 struct changelog_rec {
1344 __u16 cr_flags; /**< \a changelog_rec_flags */
1345 __u32 cr_type; /**< \a changelog_rec_type */
1346 __u64 cr_index; /**< changelog record number */
1347 __u64 cr_prev; /**< last index for this target fid */
1350 struct lu_fid cr_tfid; /**< target fid */
1351 __u32 cr_markerflags; /**< CL_MARK flags */
1353 struct lu_fid cr_pfid; /**< parent fid */
1356 /* Changelog extension for RENAME. */
1357 struct changelog_ext_rename {
1358 struct lu_fid cr_sfid; /**< source fid, or zero */
1359 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1362 /* Changelog extension to include JOBID. */
1363 struct changelog_ext_jobid {
1364 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1367 /* Changelog extension to include additional flags. */
1368 struct changelog_ext_extra_flags {
1369 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1372 /* Changelog extra extension to include UID/GID. */
1373 struct changelog_ext_uidgid {
1378 /* Changelog extra extension to include NID. */
1379 struct changelog_ext_nid {
1380 /* have __u64 instead of lnet_nid_t type for use by client api */
1382 /* for use when IPv6 support is added */
1387 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1388 struct changelog_ext_openmode {
1392 /* Changelog extra extension to include xattr */
1393 struct changelog_ext_xattr {
1394 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1397 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1398 const struct changelog_rec *rec);
1400 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1401 enum changelog_rec_extra_flags cref)
1403 size_t size = sizeof(struct changelog_rec);
1405 if (crf & CLF_RENAME)
1406 size += sizeof(struct changelog_ext_rename);
1408 if (crf & CLF_JOBID)
1409 size += sizeof(struct changelog_ext_jobid);
1411 if (crf & CLF_EXTRA_FLAGS) {
1412 size += sizeof(struct changelog_ext_extra_flags);
1413 if (cref & CLFE_UIDGID)
1414 size += sizeof(struct changelog_ext_uidgid);
1415 if (cref & CLFE_NID)
1416 size += sizeof(struct changelog_ext_nid);
1417 if (cref & CLFE_OPEN)
1418 size += sizeof(struct changelog_ext_openmode);
1419 if (cref & CLFE_XATTR)
1420 size += sizeof(struct changelog_ext_xattr);
1426 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1428 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1430 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1431 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1433 return changelog_rec_offset(rec->cr_flags, cref);
1436 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1438 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1442 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1444 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1446 return (struct changelog_ext_rename *)((char *)rec +
1447 changelog_rec_offset(crf,
1451 /* The jobid follows the rename extension, if present */
1453 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1455 enum changelog_rec_flags crf = rec->cr_flags &
1456 (CLF_VERSION | CLF_RENAME);
1458 return (struct changelog_ext_jobid *)((char *)rec +
1459 changelog_rec_offset(crf,
1463 /* The additional flags follow the rename and jobid extensions, if present */
1465 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1466 const struct changelog_rec *rec)
1468 enum changelog_rec_flags crf = rec->cr_flags &
1469 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1471 return (struct changelog_ext_extra_flags *)((char *)rec +
1472 changelog_rec_offset(crf,
1476 /* The uid/gid is the first extra extension */
1478 struct changelog_ext_uidgid *changelog_rec_uidgid(
1479 const struct changelog_rec *rec)
1481 enum changelog_rec_flags crf = rec->cr_flags &
1482 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1484 return (struct changelog_ext_uidgid *)((char *)rec +
1485 changelog_rec_offset(crf,
1489 /* The nid is the second extra extension */
1491 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1493 enum changelog_rec_flags crf = rec->cr_flags &
1494 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1495 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1497 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1498 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1501 return (struct changelog_ext_nid *)((char *)rec +
1502 changelog_rec_offset(crf, cref));
1505 /* The OPEN mode is the third extra extension */
1507 struct changelog_ext_openmode *changelog_rec_openmode(
1508 const struct changelog_rec *rec)
1510 enum changelog_rec_flags crf = rec->cr_flags &
1511 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1512 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1514 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1515 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1516 (CLFE_UIDGID | CLFE_NID);
1518 return (struct changelog_ext_openmode *)((char *)rec +
1519 changelog_rec_offset(crf, cref));
1522 /* The xattr name is the fourth extra extension */
1524 struct changelog_ext_xattr *changelog_rec_xattr(
1525 const struct changelog_rec *rec)
1527 enum changelog_rec_flags crf = rec->cr_flags &
1528 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1529 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1531 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1532 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1533 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1535 return (struct changelog_ext_xattr *)((char *)rec +
1536 changelog_rec_offset(crf, cref));
1539 /* The name follows the rename, jobid and extra flags extns, if present */
1540 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1542 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1544 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1545 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1547 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1548 cref & CLFE_SUPPORTED);
1551 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1553 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1556 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1558 char *cr_name = changelog_rec_name(rec);
1560 return cr_name + strlen(cr_name) + 1;
1564 * Remap a record to the desired format as specified by the crf flags.
1565 * The record must be big enough to contain the final remapped version.
1566 * Superfluous extension fields are removed and missing ones are added
1567 * and zeroed. The flags of the record are updated accordingly.
1569 * The jobid and rename extensions can be added to a record, to match the
1570 * format an application expects, typically. In this case, the newly added
1571 * fields will be zeroed.
1572 * The Jobid field can be removed, to guarantee compatibility with older
1573 * clients that don't expect this field in the records they process.
1575 * The following assumptions are being made:
1576 * - CLF_RENAME will not be removed
1577 * - CLF_JOBID will not be added without CLF_RENAME being added too
1578 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1580 * @param[in,out] rec The record to remap.
1581 * @param[in] crf_wanted Flags describing the desired extensions.
1582 * @param[in] cref_want Flags describing the desired extra extensions.
1584 static inline void changelog_remap_rec(struct changelog_rec *rec,
1585 enum changelog_rec_flags crf_wanted,
1586 enum changelog_rec_extra_flags cref_want)
1588 char *xattr_mov = NULL;
1589 char *omd_mov = NULL;
1590 char *nid_mov = NULL;
1591 char *uidgid_mov = NULL;
1595 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1597 crf_wanted &= CLF_SUPPORTED;
1598 cref_want &= CLFE_SUPPORTED;
1600 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1601 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1602 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1603 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1609 /* First move the variable-length name field */
1610 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1611 changelog_rec_name(rec), rec->cr_namelen);
1613 /* Locations of extensions in the remapped record */
1614 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1615 xattr_mov = (char *)rec +
1616 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1617 cref_want & ~CLFE_XATTR);
1618 omd_mov = (char *)rec +
1619 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1620 cref_want & ~(CLFE_OPEN |
1622 nid_mov = (char *)rec +
1623 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1624 cref_want & ~(CLFE_NID |
1627 uidgid_mov = (char *)rec +
1628 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1629 cref_want & ~(CLFE_UIDGID |
1633 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1636 ef_mov = (char *)rec +
1637 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1639 jid_mov = (char *)rec +
1640 changelog_rec_offset(crf_wanted &
1641 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1643 rnm_mov = (char *)rec +
1644 changelog_rec_offset(crf_wanted &
1650 /* Move the extension fields to the desired positions */
1651 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1652 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1653 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1654 memmove(xattr_mov, changelog_rec_xattr(rec),
1655 sizeof(struct changelog_ext_xattr));
1657 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1658 memmove(omd_mov, changelog_rec_openmode(rec),
1659 sizeof(struct changelog_ext_openmode));
1661 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1662 memmove(nid_mov, changelog_rec_nid(rec),
1663 sizeof(struct changelog_ext_nid));
1665 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1666 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1667 sizeof(struct changelog_ext_uidgid));
1669 memmove(ef_mov, changelog_rec_extra_flags(rec),
1670 sizeof(struct changelog_ext_extra_flags));
1673 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1674 memmove(jid_mov, changelog_rec_jobid(rec),
1675 sizeof(struct changelog_ext_jobid));
1677 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1678 memmove(rnm_mov, changelog_rec_rename(rec),
1679 sizeof(struct changelog_ext_rename));
1681 /* Clear newly added fields */
1682 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1683 !(cref & CLFE_XATTR))
1684 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1686 if (omd_mov && (cref_want & CLFE_OPEN) &&
1687 !(cref & CLFE_OPEN))
1688 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1690 if (nid_mov && (cref_want & CLFE_NID) &&
1692 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1694 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1695 !(cref & CLFE_UIDGID))
1696 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1698 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1699 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1700 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1702 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1703 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1705 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1706 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1708 /* Update the record's flags accordingly */
1709 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1710 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1711 changelog_rec_extra_flags(rec)->cr_extra_flags =
1712 changelog_rec_extra_flags(rec)->cr_extra_flags |
1716 enum changelog_message_type {
1717 CL_RECORD = 10, /* message is a changelog_rec */
1718 CL_EOF = 11, /* at end of current changelog */
1721 /********* Misc **********/
1723 struct ioc_data_version {
1725 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1726 __u32 idv_flags; /* enum ioc_data_version_flags */
1729 enum ioc_data_version_flags {
1730 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1731 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1735 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1738 #define dot_lustre_name ".lustre"
1741 /********* HSM **********/
1743 /** HSM per-file state
1744 * See HSM_FLAGS below.
1747 HS_NONE = 0x00000000,
1748 HS_EXISTS = 0x00000001,
1749 HS_DIRTY = 0x00000002,
1750 HS_RELEASED = 0x00000004,
1751 HS_ARCHIVED = 0x00000008,
1752 HS_NORELEASE = 0x00000010,
1753 HS_NOARCHIVE = 0x00000020,
1754 HS_LOST = 0x00000040,
1757 /* HSM user-setable flags. */
1758 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1760 /* Other HSM flags. */
1761 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1764 * All HSM-related possible flags that could be applied to a file.
1765 * This should be kept in sync with hsm_states.
1767 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1770 * HSM request progress state
1772 enum hsm_progress_states {
1779 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1782 case HPS_WAITING: return "waiting";
1783 case HPS_RUNNING: return "running";
1784 case HPS_DONE: return "done";
1785 default: return "unknown";
1792 } __attribute__((packed));
1795 * Current HSM states of a Lustre file.
1797 * This structure purpose is to be sent to user-space mainly. It describes the
1798 * current HSM flags and in-progress action.
1800 struct hsm_user_state {
1801 /** Current HSM states, from enum hsm_states. */
1803 __u32 hus_archive_id;
1804 /** The current undergoing action, if there is one */
1805 __u32 hus_in_progress_state;
1806 __u32 hus_in_progress_action;
1807 struct hsm_extent hus_in_progress_location;
1808 char hus_extended_info[];
1811 struct hsm_state_set_ioc {
1812 struct lu_fid hssi_fid;
1814 __u64 hssi_clearmask;
1818 * This structure describes the current in-progress action for a file.
1819 * it is retuned to user space and send over the wire
1821 struct hsm_current_action {
1822 /** The current undergoing action, if there is one */
1823 /* state is one of hsm_progress_states */
1825 /* action is one of hsm_user_action */
1827 struct hsm_extent hca_location;
1830 /***** HSM user requests ******/
1831 /* User-generated (lfs/ioctl) request types */
1832 enum hsm_user_action {
1833 HUA_NONE = 1, /* no action (noop) */
1834 HUA_ARCHIVE = 10, /* copy to hsm */
1835 HUA_RESTORE = 11, /* prestage */
1836 HUA_RELEASE = 12, /* drop ost objects */
1837 HUA_REMOVE = 13, /* remove from archive */
1838 HUA_CANCEL = 14 /* cancel a request */
1841 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1844 case HUA_NONE: return "NOOP";
1845 case HUA_ARCHIVE: return "ARCHIVE";
1846 case HUA_RESTORE: return "RESTORE";
1847 case HUA_RELEASE: return "RELEASE";
1848 case HUA_REMOVE: return "REMOVE";
1849 case HUA_CANCEL: return "CANCEL";
1850 default: return "UNKNOWN";
1855 * List of hr_flags (bit field)
1857 #define HSM_FORCE_ACTION 0x0001
1858 /* used by CT, cannot be set by user */
1859 #define HSM_GHOST_COPY 0x0002
1862 * Contains all the fixed part of struct hsm_user_request.
1865 struct hsm_request {
1866 __u32 hr_action; /* enum hsm_user_action */
1867 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1868 __u64 hr_flags; /* request flags */
1869 __u32 hr_itemcount; /* item count in hur_user_item vector */
1873 struct hsm_user_item {
1874 struct lu_fid hui_fid;
1875 struct hsm_extent hui_extent;
1876 } __attribute__((packed));
1878 struct hsm_user_request {
1879 struct hsm_request hur_request;
1880 struct hsm_user_item hur_user_item[0];
1881 /* extra data blob at end of struct (after all
1882 * hur_user_items), only use helpers to access it
1884 } __attribute__((packed));
1886 /** Return pointer to data field in a hsm user request */
1887 static inline void *hur_data(struct hsm_user_request *hur)
1889 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1893 * Compute the current length of the provided hsm_user_request. This returns -1
1894 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1896 * return -1 on bounds check error.
1898 static inline ssize_t hur_len(struct hsm_user_request *hur)
1902 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1903 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1904 (__u64)hur->hur_request.hr_itemcount *
1905 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1907 if (size != (ssize_t)size)
1913 /****** HSM RPCs to copytool *****/
1914 /* Message types the copytool may receive */
1915 enum hsm_message_type {
1916 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1919 /* Actions the copytool may be instructed to take for a given action_item */
1920 enum hsm_copytool_action {
1921 HSMA_NONE = 10, /* no action */
1922 HSMA_ARCHIVE = 20, /* arbitrary offset */
1928 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1931 case HSMA_NONE: return "NOOP";
1932 case HSMA_ARCHIVE: return "ARCHIVE";
1933 case HSMA_RESTORE: return "RESTORE";
1934 case HSMA_REMOVE: return "REMOVE";
1935 case HSMA_CANCEL: return "CANCEL";
1936 default: return "UNKNOWN";
1940 /* Copytool item action description */
1941 struct hsm_action_item {
1942 __u32 hai_len; /* valid size of this struct */
1943 __u32 hai_action; /* hsm_copytool_action, but use known size */
1944 struct lu_fid hai_fid; /* Lustre FID to operate on */
1945 struct lu_fid hai_dfid; /* fid used for data access */
1946 struct hsm_extent hai_extent; /* byte range to operate on */
1947 __u64 hai_cookie; /* action cookie from coordinator */
1948 __u64 hai_gid; /* grouplock id */
1949 char hai_data[0]; /* variable length */
1950 } __attribute__((packed));
1953 * helper function which print in hexa the first bytes of
1956 * \param hai [IN] record to print
1957 * \param buffer [IN,OUT] buffer to write the hex string to
1958 * \param len [IN] max buffer length
1962 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1963 char *buffer, size_t len)
1970 data_len = hai->hai_len - sizeof(*hai);
1971 for (i = 0; (i < data_len) && (len > 2); i++) {
1972 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1982 /* Copytool action list */
1983 #define HAL_VERSION 1
1984 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1985 struct hsm_action_list {
1987 __u32 hal_count; /* number of hai's to follow */
1988 __u64 hal_compound_id; /* returned by coordinator, ignored */
1990 __u32 hal_archive_id; /* which archive backend */
1992 char hal_fsname[0]; /* null-terminated */
1993 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1994 boundaries. See hai_zero */
1995 } __attribute__((packed));
1997 /* Return pointer to first hai in action list */
1998 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2000 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2002 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2005 /* Return pointer to next hai */
2006 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2008 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2010 return (struct hsm_action_item *)((char *)hai + offset);
2013 /* Return size of an hsm_action_list */
2014 static inline size_t hal_size(struct hsm_action_list *hal)
2018 struct hsm_action_item *hai;
2020 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2021 hai = hai_first(hal);
2022 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2023 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2029 * describe the attributes to be set on imported file
2031 struct hsm_user_import {
2040 __u32 hui_archive_id;
2043 /* Copytool progress reporting */
2044 #define HP_FLAG_COMPLETED 0x01
2045 #define HP_FLAG_RETRY 0x02
2047 struct hsm_progress {
2048 struct lu_fid hp_fid;
2050 struct hsm_extent hp_extent;
2052 __u16 hp_errval; /* positive val */
2057 __u64 hc_data_version;
2059 __u16 hc_errval; /* positive val */
2061 struct hsm_action_item hc_hai;
2065 enum llapi_json_types {
2066 LLAPI_JSON_INTEGER = 1,
2072 struct llapi_json_item {
2081 struct llapi_json_item *lji_next;
2084 struct llapi_json_item_list {
2085 int ljil_item_count;
2086 struct llapi_json_item *ljil_items;
2089 enum lu_ladvise_type {
2090 LU_LADVISE_INVALID = 0,
2091 LU_LADVISE_WILLREAD = 1,
2092 LU_LADVISE_DONTNEED = 2,
2093 LU_LADVISE_LOCKNOEXPAND = 3,
2094 LU_LADVISE_LOCKAHEAD = 4,
2098 #define LU_LADVISE_NAMES { \
2099 [LU_LADVISE_WILLREAD] = "willread", \
2100 [LU_LADVISE_DONTNEED] = "dontneed", \
2101 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2102 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2105 /* This is the userspace argument for ladvise. It is currently the same as
2106 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2107 * may need info which is only used locally. */
2108 struct llapi_lu_ladvise {
2109 __u16 lla_advice; /* advice type */
2110 __u16 lla_value1; /* values for different advice types */
2112 __u64 lla_start; /* first byte of extent for advice */
2113 __u64 lla_end; /* last byte of extent for advice */
2119 LF_ASYNC = 0x00000001,
2120 LF_UNSET = 0x00000002,
2123 #define LADVISE_MAGIC 0x1ADF1CE0
2124 /* Masks of valid flags for each advice */
2125 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2126 /* Flags valid for all advices not explicitly specified */
2127 #define LF_DEFAULT_MASK LF_ASYNC
2129 #define LF_MASK (LF_ASYNC | LF_UNSET)
2131 #define lla_lockahead_mode lla_value1
2132 #define lla_peradvice_flags lla_value2
2133 #define lla_lockahead_result lla_value3
2135 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2136 * is used on the wire. It is defined separately as we may need info which is
2137 * only used locally. */
2138 struct llapi_ladvise_hdr {
2139 __u32 lah_magic; /* LADVISE_MAGIC */
2140 __u32 lah_count; /* number of advices */
2141 __u64 lah_flags; /* from enum ladvise_flag */
2142 __u32 lah_value1; /* unused */
2143 __u32 lah_value2; /* unused */
2144 __u64 lah_value3; /* unused */
2145 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2148 #define LAH_COUNT_MAX (1024)
2152 SK_CRYPT_INVALID = -1,
2154 SK_CRYPT_AES256_CTR = 1,
2158 SK_HMAC_INVALID = -1,
2164 struct sk_crypt_type {
2165 const char *sct_name;
2169 struct sk_hmac_type {
2170 const char *sht_name;
2174 enum lock_mode_user {
2180 #define LOCK_MODE_NAMES { \
2181 [MODE_READ_USER] = "READ",\
2182 [MODE_WRITE_USER] = "WRITE"\
2185 enum lockahead_results {
2186 LLA_RESULT_SENT = 0,
2187 LLA_RESULT_DIFFERENT,
2191 enum lu_heat_flag_bit {
2192 LU_HEAT_FLAG_BIT_INVALID = 0,
2193 LU_HEAT_FLAG_BIT_OFF,
2194 LU_HEAT_FLAG_BIT_CLEAR,
2197 #define LU_HEAT_FLAG_CLEAR (1 << LU_HEAT_FLAG_BIT_CLEAR)
2198 #define LU_HEAT_FLAG_OFF (1 << LU_HEAT_FLAG_BIT_OFF)
2200 enum obd_heat_type {
2201 OBD_HEAT_READSAMPLE = 0,
2202 OBD_HEAT_WRITESAMPLE = 1,
2203 OBD_HEAT_READBYTE = 2,
2204 OBD_HEAT_WRITEBYTE = 3,
2208 #define LU_HEAT_NAMES { \
2209 [OBD_HEAT_READSAMPLE] = "readsample", \
2210 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2211 [OBD_HEAT_READBYTE] = "readbyte", \
2212 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2221 #if defined(__cplusplus)
2225 /** @} lustreuser */
2227 #endif /* _LUSTRE_USER_H */