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, 2016, 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/types.h>
48 # include <linux/fs.h>
49 # include <linux/quota.h>
50 # include <linux/string.h> /* snprintf() */
51 # include <linux/version.h>
52 # include <uapi/linux/lustre/lustre_fiemap.h>
53 #else /* !__KERNEL__ */
54 # define NEED_QUOTA_DEFS
57 # include <stdio.h> /* snprintf() */
59 # include <sys/quota.h>
60 # include <sys/stat.h>
61 # include <linux/lustre/lustre_fiemap.h>
62 #endif /* __KERNEL__ */
65 * This is a temporary solution of adding quota type.
66 * Should be removed as soon as system header is updated.
69 #define LL_MAXQUOTAS 3
71 #define INITQFNAMES { \
72 "user", /* USRQUOTA */ \
73 "group", /* GRPQUOTA */ \
74 "project", /* PRJQUOTA */ \
79 #if defined(__x86_64__) || defined(__ia64__) || defined(__ppc64__) || \
80 defined(__craynv) || defined(__mips64__) || defined(__powerpc64__) || \
82 typedef struct stat lstat_t;
83 # define lstat_f lstat
84 # define fstat_f fstat
85 # define fstatat_f fstatat
86 # define HAVE_LOV_USER_MDS_DATA
87 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
88 typedef struct stat64 lstat_t;
89 # define lstat_f lstat64
90 # define fstat_f fstat64
91 # define fstatat_f fstatat64
92 # define HAVE_LOV_USER_MDS_DATA
95 #define LUSTRE_EOF 0xffffffffffffffffULL
98 #define LL_SUPER_MAGIC 0x0BD00BD0
100 #ifndef FSFILT_IOC_GETFLAGS
101 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
102 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
103 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
104 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
105 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
106 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
109 /* FIEMAP flags supported by Lustre */
110 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
112 enum obd_statfs_state {
113 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
114 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
115 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
116 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
130 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
131 __u32 os_fprecreated; /* objs available now to the caller */
132 /* used in QoS code to find preferred
147 * FID is a cluster-wide unique identifier of a file or an object (stripe).
148 * FIDs are never reused.
152 * FID sequence. Sequence is a unit of migration: all files (objects)
153 * with FIDs from a given sequence are stored on the same server.
154 * Lustre should support 2^64 objects, so even if each sequence
155 * has only a single object we can still enumerate 2^64 objects.
158 /* FID number within sequence. */
161 * FID version, used to distinguish different versions (in the sense
162 * of snapshots, etc.) of the same file system object. Not currently
168 static inline bool fid_is_zero(const struct lu_fid *fid)
170 return fid->f_seq == 0 && fid->f_oid == 0;
173 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
174 * MDT-object's FID::f_ver, instead it is the OST-object index in its
175 * parent MDT-object's layout EA. */
176 #define f_stripe_idx f_ver
179 __u32 ol_stripe_size;
180 __u32 ol_stripe_count;
184 } __attribute__((packed));
186 /* keep this one for compatibility */
187 struct filter_fid_old {
188 struct lu_fid ff_parent;
194 struct lu_fid ff_parent;
195 struct ost_layout ff_layout;
196 __u32 ff_layout_version;
197 __u32 ff_range; /* range of layout version that
198 * write are allowed */
199 } __attribute__((packed));
201 /* Userspace should treat lu_fid as opaque, and only use the following methods
202 * to print or parse them. Other functions (e.g. compare, swab) could be moved
203 * here from lustre_idl.h if needed. */
204 typedef struct lu_fid lustre_fid;
207 LMAC_HSM = 0x00000001,
208 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
209 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
210 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
211 * under /O/<seq>/d<x>. */
212 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
213 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
217 * Masks for all features that should be supported by a Lustre version to
218 * access a specific file.
219 * This information is stored in lustre_mdt_attrs::lma_incompat.
222 LMAI_RELEASED = 0x00000001, /* file is released */
223 LMAI_AGENT = 0x00000002, /* agent inode */
224 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
225 is on the remote MDT */
226 LMAI_STRIPED = 0x00000008, /* striped directory inode */
227 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
228 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
229 LMAI_STRIPED | LMAI_ORPHAN)
234 * Following struct for object attributes, that will be kept inode's EA.
235 * Introduced in 2.0 release (please see b15993, for details)
236 * Added to all objects since Lustre 2.4 as contains self FID
238 struct lustre_mdt_attrs {
240 * Bitfield for supported data in this structure. From enum lma_compat.
241 * lma_self_fid and lma_flags are always available.
245 * Per-file incompat feature list. Lustre version should support all
246 * flags set in this field. The supported feature mask is available in
250 /** FID of this inode */
251 struct lu_fid lma_self_fid;
254 struct lustre_ost_attrs {
255 /* Use lustre_mdt_attrs directly for now, need a common header
256 * structure if want to change lustre_mdt_attrs in future. */
257 struct lustre_mdt_attrs loa_lma;
259 /* Below five elements are for OST-object's PFID EA, the
260 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
261 * and the stripe_index (low 16 bits), the size should not exceed
262 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
263 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
264 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
265 * loa_comp_* elements are valid. */
266 struct lu_fid loa_parent_fid;
267 __u32 loa_stripe_size;
269 __u64 loa_comp_start;
274 * Prior to 2.4, the LMA structure also included SOM attributes which has since
275 * been moved to a dedicated xattr
276 * lma_flags was also removed because of lma_compat/incompat fields.
278 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
281 LSOM_FL_VALID = 1 << 0,
284 struct lustre_som_attrs {
286 __u16 lsa_reserved[3];
292 * OST object IDentifier.
300 struct lu_fid oi_fid;
304 #define DOSTID "%#llx:%llu"
305 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
306 ((unsigned long long)ostid_id(oi))
308 struct ll_futimes_3 {
310 __u64 lfu_atime_nsec;
312 __u64 lfu_mtime_nsec;
314 __u64 lfu_ctime_nsec;
318 * The ioctl naming rules:
319 * LL_* - works on the currently opened filehandle instead of parent dir
320 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
321 * *_MDC_* - gets/sets data related to MDC
322 * *_LOV_* - gets/sets data related to OSC/LOV
323 * *FILE* - called on parent dir and passes in a filename
324 * *STRIPE* - set/get lov_user_md
325 * *INFO - set/get lov_user_mds_data
327 /* lustre_ioctl.h 101-150 */
328 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
329 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
330 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
331 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
332 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
333 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
334 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
335 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
336 /* LL_IOC_RECREATE_OBJ 157 obsolete */
337 /* LL_IOC_RECREATE_FID 157 obsolete */
338 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
339 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
340 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
341 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
342 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
343 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
344 /* IOC_LOV_GETINFO 165 obsolete */
345 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
346 /* LL_IOC_RMTACL 167 obsolete */
347 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
348 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
349 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
350 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
351 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
352 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
353 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
354 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
355 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
356 /* lustre_ioctl.h 177-210 */
357 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
358 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
359 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
360 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
361 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
362 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
363 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
364 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
365 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
366 struct lustre_swap_layouts)
367 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
368 struct hsm_current_action)
369 /* lustre_ioctl.h 221-232 */
370 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
371 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
372 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
373 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
374 #define LL_IOC_GET_LEASE _IO('f', 244)
375 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
376 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
377 #define LL_IOC_MIGRATE _IOR('f', 247, int)
378 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
379 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
380 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
382 #ifndef FS_IOC_FSGETXATTR
384 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
387 __u32 fsx_xflags; /* xflags field value (get/set) */
388 __u32 fsx_extsize; /* extsize field value (get/set)*/
389 __u32 fsx_nextents; /* nextents field value (get) */
390 __u32 fsx_projid; /* project identifier (get/set) */
391 unsigned char fsx_pad[12];
393 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
394 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
396 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
397 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
400 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
402 LL_LEASE_RDLCK = 0x1,
403 LL_LEASE_WRLCK = 0x2,
404 LL_LEASE_UNLCK = 0x4,
407 #define LL_STATFS_LMV 1
408 #define LL_STATFS_LOV 2
409 #define LL_STATFS_NODELAY 4
411 #define IOC_MDC_TYPE 'i'
412 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
413 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
414 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
415 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
417 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
419 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
420 * files, but are unlikely to be used in practice and are not harmful if
421 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
422 * devices and are safe for use on new files. See LU-4209. */
423 /* To be compatible with old statically linked binary we keep the check for
424 * the older 0100000000 flag. This is already removed upstream. LU-812. */
425 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
426 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
427 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
428 O_LOV_DELAY_CREATE_MASK)
430 #define LL_FILE_IGNORE_LOCK 0x00000001
431 #define LL_FILE_GROUP_LOCKED 0x00000002
432 #define LL_FILE_READAHEA 0x00000004
433 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
434 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
436 #define LOV_USER_MAGIC_V1 0x0BD10BD0
437 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
438 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
439 #define LOV_USER_MAGIC_V3 0x0BD30BD0
440 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
441 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
442 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
444 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
446 #define LOV_PATTERN_NONE 0x000
447 #define LOV_PATTERN_RAID0 0x001
448 #define LOV_PATTERN_RAID1 0x002
449 #define LOV_PATTERN_MDT 0x100
450 #define LOV_PATTERN_CMOBD 0x200
452 #define LOV_PATTERN_F_MASK 0xffff0000
453 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
454 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
455 #define LOV_PATTERN_DEFAULT 0xffffffff
457 static inline bool lov_pattern_supported(__u32 pattern)
459 return pattern == LOV_PATTERN_RAID0 ||
460 pattern == LOV_PATTERN_MDT ||
461 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_F_RELEASED);
464 #define LOV_MAXPOOLNAME 15
465 #define LOV_POOLNAMEF "%.15s"
467 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
468 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
469 #define LOV_MAX_STRIPE_COUNT_OLD 160
470 /* This calculation is crafted so that input of 4096 will result in 160
471 * which in turn is equal to old maximal stripe count.
472 * XXX: In fact this is too simpified for now, what it also need is to get
473 * ea_type argument to clearly know how much space each stripe consumes.
475 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
476 * allocation that is sufficient for the current generation of systems.
478 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
479 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
480 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
481 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
483 #define XATTR_LUSTRE_PREFIX "lustre."
484 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
486 #define lov_user_ost_data lov_user_ost_data_v1
487 struct lov_user_ost_data_v1 { /* per-stripe data structure */
488 struct ost_id l_ost_oi; /* OST object ID */
489 __u32 l_ost_gen; /* generation of this OST index */
490 __u32 l_ost_idx; /* OST index in LOV */
491 } __attribute__((packed));
493 #define lov_user_md lov_user_md_v1
494 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
495 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
496 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
497 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
498 __u32 lmm_stripe_size; /* size of stripe in bytes */
499 __u16 lmm_stripe_count; /* num stripes in use for this object */
501 __u16 lmm_stripe_offset; /* starting stripe offset in
502 * lmm_objects, use when writing */
503 __u16 lmm_layout_gen; /* layout generation number
504 * used when reading */
506 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
507 } __attribute__((packed, __may_alias__));
509 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
510 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
511 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
512 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
513 __u32 lmm_stripe_size; /* size of stripe in bytes */
514 __u16 lmm_stripe_count; /* num stripes in use for this object */
516 __u16 lmm_stripe_offset; /* starting stripe offset in
517 * lmm_objects, use when writing */
518 __u16 lmm_layout_gen; /* layout generation number
519 * used when reading */
521 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
522 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
523 } __attribute__((packed));
530 #define DEXT "[%#llx, %#llx)"
531 #define PEXT(ext) (ext)->e_start, (ext)->e_end
533 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
534 struct lu_extent *e2)
536 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
539 static inline bool lu_extent_is_whole(struct lu_extent *e)
541 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
544 enum lov_comp_md_entry_flags {
545 LCME_FL_PRIMARY = 0x00000001, /* Not used */
546 LCME_FL_STALE = 0x00000002, /* Not used */
547 LCME_FL_OFFLINE = 0x00000004, /* Not used */
548 LCME_FL_PREFERRED = 0x00000008, /* Not used */
549 LCME_FL_INIT = 0x00000010, /* instantiated */
550 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
551 won't be stored on disk */
554 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT)
556 /* the highest bit in obdo::o_layout_version is used to mark if the file is
558 #define LU_LAYOUT_RESYNC LCME_FL_NEG
560 /* lcme_id can be specified as certain flags, and the the first
561 * bit of lcme_id is used to indicate that the ID is representing
562 * certain LCME_FL_* but not a real ID. Which implies we can have
563 * at most 31 flags (see LCME_FL_XXX). */
566 LCME_ID_MAX = 0x7FFFFFFF,
567 LCME_ID_ALL = 0xFFFFFFFF,
568 LCME_ID_NOT_ID = LCME_FL_NEG
571 #define LCME_ID_MASK LCME_ID_MAX
573 struct lov_comp_md_entry_v1 {
574 __u32 lcme_id; /* unique id of component */
575 __u32 lcme_flags; /* LCME_FL_XXX */
576 struct lu_extent lcme_extent; /* file extent for component */
577 __u32 lcme_offset; /* offset of component blob,
578 start from lov_comp_md_v1 */
579 __u32 lcme_size; /* size of component blob */
580 __u64 lcme_padding[2];
581 } __attribute__((packed));
583 #define SEQ_ID_MAX 0x0000FFFF
584 #define SEQ_ID_MASK SEQ_ID_MAX
585 /* bit 30:16 of lcme_id is used to store mirror id */
586 #define MIRROR_ID_MASK 0x7FFF0000
587 #define MIRROR_ID_SHIFT 16
589 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
591 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
594 static inline __u16 mirror_id_of(__u32 id)
596 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
600 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
602 enum lov_comp_md_flags {
603 /* the least 2 bits are used by FLR to record file state */
606 LCM_FL_WRITE_PENDING = 2,
607 LCM_FL_SYNC_PENDING = 3,
608 LCM_FL_FLR_MASK = 0x3,
611 struct lov_comp_md_v1 {
612 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
613 __u32 lcm_size; /* overall size including this struct */
614 __u32 lcm_layout_gen;
616 __u16 lcm_entry_count;
617 /* lcm_mirror_count stores the number of actual mirrors minus 1,
618 * so that non-flr files will have value 0 meaning 1 mirror. */
619 __u16 lcm_mirror_count;
620 __u16 lcm_padding1[3];
622 struct lov_comp_md_entry_v1 lcm_entries[0];
623 } __attribute__((packed));
626 * Maximum number of mirrors Lustre can support.
628 #define LUSTRE_MIRROR_COUNT_MAX 16
630 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
632 if (stripes == (__u16)-1)
635 if (lmm_magic == LOV_USER_MAGIC_V1)
636 return sizeof(struct lov_user_md_v1) +
637 stripes * sizeof(struct lov_user_ost_data_v1);
638 return sizeof(struct lov_user_md_v3) +
639 stripes * sizeof(struct lov_user_ost_data_v1);
642 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
643 * use this. It is unsafe to #define those values in this header as it
644 * is possible the application has already #included <sys/stat.h>. */
645 #ifdef HAVE_LOV_USER_MDS_DATA
646 #define lov_user_mds_data lov_user_mds_data_v1
647 struct lov_user_mds_data_v1 {
648 lstat_t lmd_st; /* MDS stat struct */
649 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
650 } __attribute__((packed));
652 struct lov_user_mds_data_v3 {
653 lstat_t lmd_st; /* MDS stat struct */
654 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
655 } __attribute__((packed));
658 struct lmv_user_mds_data {
659 struct lu_fid lum_fid;
665 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
666 LMV_HASH_TYPE_ALL_CHARS = 1,
667 LMV_HASH_TYPE_FNV_1A_64 = 2,
671 #define LMV_HASH_NAME_ALL_CHARS "all_char"
672 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
674 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
676 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
677 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
678 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
679 #define lmv_user_md lmv_user_md_v1
680 struct lmv_user_md_v1 {
681 __u32 lum_magic; /* must be the first field */
682 __u32 lum_stripe_count; /* dirstripe count */
683 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
684 __u32 lum_hash_type; /* Dir stripe policy */
685 __u32 lum_type; /* LMV type: default or normal */
689 char lum_pool_name[LOV_MAXPOOLNAME + 1];
690 struct lmv_user_mds_data lum_objects[0];
691 } __attribute__((packed));
693 static inline int lmv_user_md_size(int stripes, int lmm_magic)
695 return sizeof(struct lmv_user_md) +
696 stripes * sizeof(struct lmv_user_mds_data);
699 struct ll_recreate_obj {
705 __u64 id; /* holds object id */
706 __u32 generation; /* holds object generation */
707 __u32 f_type; /* holds object type or stripe idx when passing it to
708 * OST for saving into EA. */
716 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
717 const struct obd_uuid *u2)
719 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
722 static inline int obd_uuid_empty(struct obd_uuid *uuid)
724 return uuid->uuid[0] == '\0';
727 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
729 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
730 uuid->uuid[sizeof(*uuid) - 1] = '\0';
733 /* For printf's only, make sure uuid is terminated */
734 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
739 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
740 /* Obviously not safe, but for printfs, no real harm done...
741 we're always null-terminated, even in a race. */
742 static char temp[sizeof(*uuid)];
743 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
744 temp[sizeof(*uuid) - 1] = '\0';
747 return (char *)(uuid->uuid);
750 #define LUSTRE_MAXFSNAME 8
752 /* Extract fsname from uuid (or target name) of a target
753 e.g. (myfs-OST0007_UUID -> myfs)
754 see also deuuidify. */
755 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
759 strncpy(buf, uuid, buflen - 1);
760 buf[buflen - 1] = '\0';
761 p = strrchr(buf, '-');
766 /* printf display format for Lustre FIDs
767 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
768 #define FID_NOBRACE_LEN 40
769 #define FID_LEN (FID_NOBRACE_LEN + 2)
770 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
771 #define DFID "["DFID_NOBRACE"]"
772 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
774 /* scanf input parse format for fids in DFID_NOBRACE format
775 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
776 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
777 #define SFID "0x%llx:0x%x:0x%x"
778 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
780 /********* Quotas **********/
782 #define LUSTRE_QUOTABLOCK_BITS 10
783 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
785 static inline __u64 lustre_stoqb(size_t space)
787 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
790 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
791 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
792 #define Q_GETOINFO 0x800102 /* get obd quota info */
793 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
794 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
796 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
797 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
798 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
799 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
800 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
801 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
802 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
803 /* lustre-specific control commands */
804 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
805 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
807 #define ALLQUOTA 255 /* set all quota */
808 static inline char *qtype_name(int qtype)
821 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
824 #define N_PERMS_MAX 64
826 struct perm_downcall_data {
832 struct identity_downcall_data {
839 struct perm_downcall_data idd_perms[N_PERMS_MAX];
843 #ifdef NEED_QUOTA_DEFS
845 #define QIF_BLIMITS 1
847 #define QIF_ILIMITS 4
851 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
852 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
853 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
854 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
857 #endif /* !__KERNEL__ */
859 /* lustre volatile file support
860 * file name header: .^L^S^T^R:volatile"
862 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
863 #define LUSTRE_VOLATILE_HDR_LEN 14
865 typedef enum lustre_quota_version {
867 } lustre_quota_version_t;
869 /* XXX: same as if_dqinfo struct in kernel */
877 /* XXX: same as if_dqblk struct in kernel, plus one padding */
879 __u64 dqb_bhardlimit;
880 __u64 dqb_bsoftlimit;
882 __u64 dqb_ihardlimit;
883 __u64 dqb_isoftlimit;
905 struct obd_dqinfo qc_dqinfo;
906 struct obd_dqblk qc_dqblk;
908 struct obd_uuid obd_uuid;
911 /* swap layout flags */
912 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
913 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
914 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
915 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
916 #define SWAP_LAYOUTS_CLOSE (1 << 4)
917 #define MERGE_LAYOUTS_CLOSE (1 << 5)
918 #define INTENT_LAYOUTS_CLOSE (SWAP_LAYOUTS_CLOSE | MERGE_LAYOUTS_CLOSE)
920 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
921 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
922 struct lustre_swap_layouts {
931 /********* Changelogs **********/
932 /** Changelog record types */
933 enum changelog_rec_type {
935 CL_CREATE = 1, /* namespace */
936 CL_MKDIR = 2, /* namespace */
937 CL_HARDLINK = 3, /* namespace */
938 CL_SOFTLINK = 4, /* namespace */
939 CL_MKNOD = 5, /* namespace */
940 CL_UNLINK = 6, /* namespace */
941 CL_RMDIR = 7, /* namespace */
942 CL_RENAME = 8, /* namespace */
943 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
944 CL_OPEN = 10, /* not currently used */
945 CL_CLOSE = 11, /* may be written to log only with mtime change */
946 CL_LAYOUT = 12, /* file layout/striping modified */
950 CL_HSM = 16, /* HSM specific events, see flags */
951 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
955 CL_FLRW = 21, /* FLR: file was firstly written */
956 CL_RESYNC = 22, /* FLR: file was resync-ed */
960 static inline const char *changelog_type2str(int type) {
961 static const char *changelog_str[] = {
962 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
963 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
964 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
968 if (type >= 0 && type < CL_LAST)
969 return changelog_str[type];
973 /* per-record flags */
974 #define CLF_FLAGSHIFT 12
975 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
976 #define CLF_VERMASK (~CLF_FLAGMASK)
977 enum changelog_rec_flags {
978 CLF_VERSION = 0x1000,
981 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
985 /* Anything under the flagmask may be per-type (if desired) */
986 /* Flags for unlink */
987 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
988 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
989 /* HSM cleaning needed */
990 /* Flags for rename */
991 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
993 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
994 * has an archive in backend */
997 /* 12b used (from high weight to low weight):
1002 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1003 #define CLF_HSM_ERR_H 6
1004 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1005 #define CLF_HSM_EVENT_H 9
1006 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1007 #define CLF_HSM_FLAG_H 11
1008 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1009 #define CLF_HSM_SPARE_H 15
1010 #define CLF_HSM_LAST 15
1012 /* Remove bits higher than _h, then extract the value
1013 * between _h and _l by shifting lower weigth to bit 0. */
1014 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1015 >> (CLF_HSM_LAST - _h + _l))
1017 #define CLF_HSM_SUCCESS 0x00
1018 #define CLF_HSM_MAXERROR 0x7E
1019 #define CLF_HSM_ERROVERFLOW 0x7F
1021 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1023 /* 3 bits field => 8 values allowed */
1035 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1037 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1041 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
1043 *flags |= (he << CLF_HSM_EVENT_L);
1046 static inline __u16 hsm_get_cl_flags(int flags)
1048 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1051 static inline void hsm_set_cl_flags(int *flags, int bits)
1053 *flags |= (bits << CLF_HSM_FLAG_L);
1056 static inline int hsm_get_cl_error(int flags)
1058 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1061 static inline void hsm_set_cl_error(int *flags, int error)
1063 *flags |= (error << CLF_HSM_ERR_L);
1066 enum changelog_send_flag {
1067 /* Not yet implemented */
1068 CHANGELOG_FLAG_FOLLOW = 0x01,
1069 /* Blocking IO makes sense in case of slow user parsing of the records,
1070 * but it also prevents us from cleaning up if the records are not
1072 CHANGELOG_FLAG_BLOCK = 0x02,
1073 /* Pack jobid into the changelog records if available. */
1074 CHANGELOG_FLAG_JOBID = 0x04,
1077 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
1078 changelog_rec_offset(CLF_SUPPORTED))
1080 /* 31 usable bytes string + null terminator. */
1081 #define LUSTRE_JOBID_SIZE 32
1083 /* This is the minimal changelog record. It can contain extensions
1084 * such as rename fields or process jobid. Its exact content is described
1087 * Extensions are packed in the same order as their corresponding flags.
1089 struct changelog_rec {
1091 __u16 cr_flags; /**< \a changelog_rec_flags */
1092 __u32 cr_type; /**< \a changelog_rec_type */
1093 __u64 cr_index; /**< changelog record number */
1094 __u64 cr_prev; /**< last index for this target fid */
1097 lustre_fid cr_tfid; /**< target fid */
1098 __u32 cr_markerflags; /**< CL_MARK flags */
1100 lustre_fid cr_pfid; /**< parent fid */
1103 /* Changelog extension for RENAME. */
1104 struct changelog_ext_rename {
1105 lustre_fid cr_sfid; /**< source fid, or zero */
1106 lustre_fid cr_spfid; /**< source parent fid, or zero */
1109 /* Changelog extension to include JOBID. */
1110 struct changelog_ext_jobid {
1111 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1115 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
1117 size_t size = sizeof(struct changelog_rec);
1119 if (crf & CLF_RENAME)
1120 size += sizeof(struct changelog_ext_rename);
1122 if (crf & CLF_JOBID)
1123 size += sizeof(struct changelog_ext_jobid);
1128 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1130 return changelog_rec_offset(rec->cr_flags);
1133 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1135 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1139 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1141 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1143 return (struct changelog_ext_rename *)((char *)rec +
1144 changelog_rec_offset(crf));
1147 /* The jobid follows the rename extension, if present */
1149 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1151 enum changelog_rec_flags crf = rec->cr_flags &
1152 (CLF_VERSION | CLF_RENAME);
1154 return (struct changelog_ext_jobid *)((char *)rec +
1155 changelog_rec_offset(crf));
1158 /* The name follows the rename and jobid extensions, if present */
1159 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1161 return (char *)rec + changelog_rec_offset(rec->cr_flags &
1165 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1167 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1170 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1172 char *cr_name = changelog_rec_name(rec);
1174 return cr_name + strlen(cr_name) + 1;
1178 * Remap a record to the desired format as specified by the crf flags.
1179 * The record must be big enough to contain the final remapped version.
1180 * Superfluous extension fields are removed and missing ones are added
1181 * and zeroed. The flags of the record are updated accordingly.
1183 * The jobid and rename extensions can be added to a record, to match the
1184 * format an application expects, typically. In this case, the newly added
1185 * fields will be zeroed.
1186 * The Jobid field can be removed, to guarantee compatibility with older
1187 * clients that don't expect this field in the records they process.
1189 * The following assumptions are being made:
1190 * - CLF_RENAME will not be removed
1191 * - CLF_JOBID will not be added without CLF_RENAME being added too
1193 * @param[in,out] rec The record to remap.
1194 * @param[in] crf_wanted Flags describing the desired extensions.
1196 static inline void changelog_remap_rec(struct changelog_rec *rec,
1197 enum changelog_rec_flags crf_wanted)
1202 crf_wanted &= CLF_SUPPORTED;
1204 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
1207 /* First move the variable-length name field */
1208 memmove((char *)rec + changelog_rec_offset(crf_wanted),
1209 changelog_rec_name(rec), rec->cr_namelen);
1211 /* Locations of jobid and rename extensions in the remapped record */
1212 jid_mov = (char *)rec +
1213 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
1214 rnm_mov = (char *)rec +
1215 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
1217 /* Move the extension fields to the desired positions */
1218 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1219 memmove(jid_mov, changelog_rec_jobid(rec),
1220 sizeof(struct changelog_ext_jobid));
1222 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1223 memmove(rnm_mov, changelog_rec_rename(rec),
1224 sizeof(struct changelog_ext_rename));
1226 /* Clear newly added fields */
1227 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1228 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1230 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1231 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1233 /* Update the record's flags accordingly */
1234 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1237 enum changelog_message_type {
1238 CL_RECORD = 10, /* message is a changelog_rec */
1239 CL_EOF = 11, /* at end of current changelog */
1242 /********* Misc **********/
1244 struct ioc_data_version {
1246 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1247 __u32 idv_flags; /* enum ioc_data_version_flags */
1250 enum ioc_data_version_flags {
1251 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1252 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1256 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1259 #define dot_lustre_name ".lustre"
1262 /********* HSM **********/
1264 /** HSM per-file state
1265 * See HSM_FLAGS below.
1268 HS_NONE = 0x00000000,
1269 HS_EXISTS = 0x00000001,
1270 HS_DIRTY = 0x00000002,
1271 HS_RELEASED = 0x00000004,
1272 HS_ARCHIVED = 0x00000008,
1273 HS_NORELEASE = 0x00000010,
1274 HS_NOARCHIVE = 0x00000020,
1275 HS_LOST = 0x00000040,
1278 /* HSM user-setable flags. */
1279 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1281 /* Other HSM flags. */
1282 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1285 * All HSM-related possible flags that could be applied to a file.
1286 * This should be kept in sync with hsm_states.
1288 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1291 * HSMÂ request progress state
1293 enum hsm_progress_states {
1300 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1303 case HPS_WAITING: return "waiting";
1304 case HPS_RUNNING: return "running";
1305 case HPS_DONE: return "done";
1306 default: return "unknown";
1313 } __attribute__((packed));
1316 * Current HSM states of a Lustre file.
1318 * This structure purpose is to be sent to user-space mainly. It describes the
1319 * current HSM flags and in-progress action.
1321 struct hsm_user_state {
1322 /** Current HSM states, from enum hsm_states. */
1324 __u32 hus_archive_id;
1325 /** The current undergoing action, if there is one */
1326 __u32 hus_in_progress_state;
1327 __u32 hus_in_progress_action;
1328 struct hsm_extent hus_in_progress_location;
1329 char hus_extended_info[];
1332 struct hsm_state_set_ioc {
1333 struct lu_fid hssi_fid;
1335 __u64 hssi_clearmask;
1339 * This structure describes the current in-progress action for a file.
1340 * it is retuned to user space and send over the wire
1342 struct hsm_current_action {
1343 /** The current undergoing action, if there is one */
1344 /* state is one of hsm_progress_states */
1346 /* action is one of hsm_user_action */
1348 struct hsm_extent hca_location;
1351 /***** HSM user requests ******/
1352 /* User-generated (lfs/ioctl) request types */
1353 enum hsm_user_action {
1354 HUA_NONE = 1, /* no action (noop) */
1355 HUA_ARCHIVE = 10, /* copy to hsm */
1356 HUA_RESTORE = 11, /* prestage */
1357 HUA_RELEASE = 12, /* drop ost objects */
1358 HUA_REMOVE = 13, /* remove from archive */
1359 HUA_CANCEL = 14 /* cancel a request */
1362 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1365 case HUA_NONE: return "NOOP";
1366 case HUA_ARCHIVE: return "ARCHIVE";
1367 case HUA_RESTORE: return "RESTORE";
1368 case HUA_RELEASE: return "RELEASE";
1369 case HUA_REMOVE: return "REMOVE";
1370 case HUA_CANCEL: return "CANCEL";
1371 default: return "UNKNOWN";
1376 * List of hr_flags (bit field)
1378 #define HSM_FORCE_ACTION 0x0001
1379 /* used by CT, cannot be set by user */
1380 #define HSM_GHOST_COPY 0x0002
1383 * Contains all the fixed part of struct hsm_user_request.
1386 struct hsm_request {
1387 __u32 hr_action; /* enum hsm_user_action */
1388 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1389 __u64 hr_flags; /* request flags */
1390 __u32 hr_itemcount; /* item count in hur_user_item vector */
1394 struct hsm_user_item {
1396 struct hsm_extent hui_extent;
1397 } __attribute__((packed));
1399 struct hsm_user_request {
1400 struct hsm_request hur_request;
1401 struct hsm_user_item hur_user_item[0];
1402 /* extra data blob at end of struct (after all
1403 * hur_user_items), only use helpers to access it
1405 } __attribute__((packed));
1407 /** Return pointer to data field in a hsm user request */
1408 static inline void *hur_data(struct hsm_user_request *hur)
1410 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1414 * Compute the current length of the provided hsm_user_request. This returns -1
1415 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1417 * return -1 on bounds check error.
1419 static inline ssize_t hur_len(struct hsm_user_request *hur)
1423 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1424 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1425 (__u64)hur->hur_request.hr_itemcount *
1426 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1428 if (size != (ssize_t)size)
1434 /****** HSM RPCs to copytool *****/
1435 /* Message types the copytool may receive */
1436 enum hsm_message_type {
1437 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1440 /* Actions the copytool may be instructed to take for a given action_item */
1441 enum hsm_copytool_action {
1442 HSMA_NONE = 10, /* no action */
1443 HSMA_ARCHIVE = 20, /* arbitrary offset */
1449 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1452 case HSMA_NONE: return "NOOP";
1453 case HSMA_ARCHIVE: return "ARCHIVE";
1454 case HSMA_RESTORE: return "RESTORE";
1455 case HSMA_REMOVE: return "REMOVE";
1456 case HSMA_CANCEL: return "CANCEL";
1457 default: return "UNKNOWN";
1461 /* Copytool item action description */
1462 struct hsm_action_item {
1463 __u32 hai_len; /* valid size of this struct */
1464 __u32 hai_action; /* hsm_copytool_action, but use known size */
1465 lustre_fid hai_fid; /* Lustre FID to operate on */
1466 lustre_fid hai_dfid; /* fid used for data access */
1467 struct hsm_extent hai_extent; /* byte range to operate on */
1468 __u64 hai_cookie; /* action cookie from coordinator */
1469 __u64 hai_gid; /* grouplock id */
1470 char hai_data[0]; /* variable length */
1471 } __attribute__((packed));
1474 * helper function which print in hexa the first bytes of
1477 * \param hai [IN] record to print
1478 * \param buffer [IN,OUT] buffer to write the hex string to
1479 * \param len [IN] max buffer length
1483 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1484 char *buffer, size_t len)
1491 data_len = hai->hai_len - sizeof(*hai);
1492 for (i = 0; (i < data_len) && (len > 2); i++) {
1493 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1503 /* Copytool action list */
1504 #define HAL_VERSION 1
1505 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1506 struct hsm_action_list {
1508 __u32 hal_count; /* number of hai's to follow */
1509 __u64 hal_compound_id; /* returned by coordinator */
1511 __u32 hal_archive_id; /* which archive backend */
1513 char hal_fsname[0]; /* null-terminated */
1514 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1515 boundaries. See hai_zero */
1516 } __attribute__((packed));
1518 #ifndef HAVE_CFS_SIZE_ROUND
1519 static inline int cfs_size_round (int val)
1521 return (val + 7) & (~0x7);
1523 #define HAVE_CFS_SIZE_ROUND
1526 /* Return pointer to first hai in action list */
1527 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1529 return (struct hsm_action_item *)(hal->hal_fsname +
1530 cfs_size_round(strlen(hal-> \
1534 /* Return pointer to next hai */
1535 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1537 return (struct hsm_action_item *)((char *)hai +
1538 cfs_size_round(hai->hai_len));
1541 /* Return size of an hsm_action_list */
1542 static inline size_t hal_size(struct hsm_action_list *hal)
1546 struct hsm_action_item *hai;
1548 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1549 hai = hai_first(hal);
1550 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1551 sz += cfs_size_round(hai->hai_len);
1557 * describe the attributes to be set on imported file
1559 struct hsm_user_import {
1568 __u32 hui_archive_id;
1571 /* Copytool progress reporting */
1572 #define HP_FLAG_COMPLETED 0x01
1573 #define HP_FLAG_RETRY 0x02
1575 struct hsm_progress {
1578 struct hsm_extent hp_extent;
1580 __u16 hp_errval; /* positive val */
1585 __u64 hc_data_version;
1587 __u16 hc_errval; /* positive val */
1589 struct hsm_action_item hc_hai;
1593 enum llapi_json_types {
1594 LLAPI_JSON_INTEGER = 1,
1600 struct llapi_json_item {
1609 struct llapi_json_item *lji_next;
1612 struct llapi_json_item_list {
1613 int ljil_item_count;
1614 struct llapi_json_item *ljil_items;
1617 enum lu_ladvise_type {
1618 LU_LADVISE_INVALID = 0,
1619 LU_LADVISE_WILLREAD = 1,
1620 LU_LADVISE_DONTNEED = 2,
1621 LU_LADVISE_LOCKNOEXPAND = 3,
1622 LU_LADVISE_LOCKAHEAD = 4,
1626 #define LU_LADVISE_NAMES { \
1627 [LU_LADVISE_WILLREAD] = "willread", \
1628 [LU_LADVISE_DONTNEED] = "dontneed", \
1629 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
1630 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
1633 /* This is the userspace argument for ladvise. It is currently the same as
1634 * what goes on the wire (struct lu_ladvise), but is defined separately as we
1635 * may need info which is only used locally. */
1636 struct llapi_lu_ladvise {
1637 __u16 lla_advice; /* advice type */
1638 __u16 lla_value1; /* values for different advice types */
1640 __u64 lla_start; /* first byte of extent for advice */
1641 __u64 lla_end; /* last byte of extent for advice */
1647 LF_ASYNC = 0x00000001,
1648 LF_UNSET = 0x00000002,
1651 #define LADVISE_MAGIC 0x1ADF1CE0
1652 /* Masks of valid flags for each advice */
1653 #define LF_LOCKNOEXPAND_MASK LF_UNSET
1654 /* Flags valid for all advices not explicitly specified */
1655 #define LF_DEFAULT_MASK LF_ASYNC
1657 #define LF_MASK (LF_ASYNC | LF_UNSET)
1659 #define lla_lockahead_mode lla_value1
1660 #define lla_peradvice_flags lla_value2
1661 #define lla_lockahead_result lla_value3
1663 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
1664 * is used on the wire. It is defined separately as we may need info which is
1665 * only used locally. */
1666 struct llapi_ladvise_hdr {
1667 __u32 lah_magic; /* LADVISE_MAGIC */
1668 __u32 lah_count; /* number of advices */
1669 __u64 lah_flags; /* from enum ladvise_flag */
1670 __u32 lah_value1; /* unused */
1671 __u32 lah_value2; /* unused */
1672 __u64 lah_value3; /* unused */
1673 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
1676 #define LAH_COUNT_MAX (1024)
1680 SK_CRYPT_INVALID = -1,
1682 SK_CRYPT_AES256_CTR = 1,
1687 SK_HMAC_INVALID = -1,
1694 struct sk_crypt_type {
1699 struct sk_hmac_type {
1704 enum lock_mode_user {
1710 #define LOCK_MODE_NAMES { \
1711 [MODE_READ_USER] = "READ",\
1712 [MODE_WRITE_USER] = "WRITE"\
1715 enum lockahead_results {
1716 LLA_RESULT_SENT = 0,
1717 LLA_RESULT_DIFFERENT,
1721 /** @} lustreuser */
1723 #endif /* _LUSTRE_USER_H */