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() */
60 # define NEED_QUOTA_DEFS
61 /* # include <sys/quota.h> - this causes complaints about caddr_t */
62 # include <sys/stat.h>
63 # include <linux/lustre/lustre_fiemap.h>
64 #endif /* __KERNEL__ */
66 /* Handle older distros */
67 #ifndef __ALIGN_KERNEL
68 # define __ALIGN_KERNEL(x, a) __ALIGN_KERNEL_MASK(x, (typeof(x))(a) - 1)
69 # define __ALIGN_KERNEL_MASK(x, mask) (((x) + (mask)) & ~(mask))
72 #if defined(__cplusplus)
77 * This is a temporary solution of adding quota type.
78 * Should be removed as soon as system header is updated.
81 #define LL_MAXQUOTAS 3
83 #define INITQFNAMES { \
84 "user", /* USRQUOTA */ \
85 "group", /* GRPQUOTA */ \
86 "project", /* PRJQUOTA */ \
100 * We need to always use 64bit version because the structure
101 * is shared across entire cluster where 32bit and 64bit machines
104 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
105 typedef struct stat64 lstat_t;
106 #define lstat_f lstat64
107 #define fstat_f fstat64
108 #define fstatat_f fstatat64
110 typedef struct stat lstat_t;
111 #define lstat_f lstat
112 #define fstat_f fstat
113 #define fstatat_f fstatat
116 #define HAVE_LOV_USER_MDS_DATA
118 #define LUSTRE_EOF 0xffffffffffffffffULL
121 #define LL_SUPER_MAGIC 0x0BD00BD0
123 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
125 /* FIEMAP flags supported by Lustre */
126 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
128 enum obd_statfs_state {
129 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
130 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
131 OS_STATE_NOPRECREATE = 0x00000004, /**< no object precreation */
132 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
133 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
134 OS_STATE_SUM = 0x00000100, /**< aggregated for all tagrets */
135 OS_STATE_NONROT = 0x00000200, /**< non-rotational device */
138 /** filesystem statistics/attributes for target device */
140 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
141 __u64 os_blocks; /* total size in #os_bsize blocks */
142 __u64 os_bfree; /* number of unused blocks */
143 __u64 os_bavail; /* blocks available for allocation */
144 __u64 os_files; /* total number of objects */
145 __u64 os_ffree; /* # objects that could be created */
146 __u8 os_fsid[40]; /* identifier for filesystem */
147 __u32 os_bsize; /* block size in bytes for os_blocks */
148 __u32 os_namelen; /* maximum length of filename in bytes*/
149 __u64 os_maxbytes; /* maximum object size in bytes */
150 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
151 __u32 os_fprecreated; /* objs available now to the caller */
152 /* used in QoS code to find preferred
154 __u32 os_granted; /* space granted for MDS */
155 __u32 os_spare3; /* Unused padding fields. Remember */
156 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
167 * FID is a cluster-wide unique identifier of a file or an object (stripe).
168 * FIDs are never reused.
172 * FID sequence. Sequence is a unit of migration: all files (objects)
173 * with FIDs from a given sequence are stored on the same server.
174 * Lustre should support 2^64 objects, so even if each sequence
175 * has only a single object we can still enumerate 2^64 objects.
178 /* FID number within sequence. */
181 * FID version, used to distinguish different versions (in the sense
182 * of snapshots, etc.) of the same file system object. Not currently
188 static inline bool fid_is_zero(const struct lu_fid *fid)
190 return fid->f_seq == 0 && fid->f_oid == 0;
193 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
194 * MDT-object's FID::f_ver, instead it is the OST-object index in its
195 * parent MDT-object's layout EA. */
196 #define f_stripe_idx f_ver
199 __u32 ol_stripe_size;
200 __u32 ol_stripe_count;
204 } __attribute__((packed));
206 /* The filter_fid structure has changed several times over its lifetime.
207 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
208 * stripe_index and the "self FID" (objid/seq) to be able to recover the
209 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
210 * the OST, the "trusted.lma" xattr was added to the OST objects to store
211 * the "self FID" to be consistent with the MDT on-disk format, and the
212 * filter_fid only stored the MDT inode parent FID and stripe index.
214 * In 2.10, the addition of PFL composite layouts required more information
215 * to be stored into the filter_fid in order to be able to identify which
216 * component the OST object belonged. As well, the stripe size may vary
217 * between components, so it was no longer safe to assume the stripe size
218 * or stripe_count of a file. This is also more robust for plain layouts.
220 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
221 * enough space to store both the filter_fid and LMA in the inode, so they
222 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
223 * an extra seek for every OST object access.
225 * In 2.11, FLR mirror layouts also need to store the layout version and
226 * range so that writes to old versions of the layout are not allowed.
227 * That ensures that mirrored objects are not modified by evicted clients,
228 * and ensures that the components are correctly marked stale on the MDT.
230 struct filter_fid_18_23 {
231 struct lu_fid ff_parent; /* stripe_idx in f_ver */
236 struct filter_fid_24_29 {
237 struct lu_fid ff_parent; /* stripe_idx in f_ver */
240 struct filter_fid_210 {
241 struct lu_fid ff_parent; /* stripe_idx in f_ver */
242 struct ost_layout ff_layout;
246 struct lu_fid ff_parent; /* stripe_idx in f_ver */
247 struct ost_layout ff_layout;
248 __u32 ff_layout_version;
249 __u32 ff_range; /* range of layout version that
250 * write are allowed */
251 } __attribute__((packed));
253 /* Userspace should treat lu_fid as opaque, and only use the following methods
254 * to print or parse them. Other functions (e.g. compare, swab) could be moved
255 * here from lustre_idl.h if needed. */
259 LMAC_HSM = 0x00000001,
260 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
261 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
262 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
263 * under /O/<seq>/d<x>. */
264 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
265 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
266 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
270 * Masks for all features that should be supported by a Lustre version to
271 * access a specific file.
272 * This information is stored in lustre_mdt_attrs::lma_incompat.
275 LMAI_RELEASED = 0x00000001, /* file is released */
276 LMAI_AGENT = 0x00000002, /* agent inode */
277 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
278 is on the remote MDT */
279 LMAI_STRIPED = 0x00000008, /* striped directory inode */
280 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
281 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
282 LMAI_STRIPED | LMAI_ORPHAN)
287 * Following struct for object attributes, that will be kept inode's EA.
288 * Introduced in 2.0 release (please see b15993, for details)
289 * Added to all objects since Lustre 2.4 as contains self FID
291 struct lustre_mdt_attrs {
293 * Bitfield for supported data in this structure. From enum lma_compat.
294 * lma_self_fid and lma_flags are always available.
298 * Per-file incompat feature list. Lustre version should support all
299 * flags set in this field. The supported feature mask is available in
303 /** FID of this inode */
304 struct lu_fid lma_self_fid;
307 struct lustre_ost_attrs {
308 /* Use lustre_mdt_attrs directly for now, need a common header
309 * structure if want to change lustre_mdt_attrs in future. */
310 struct lustre_mdt_attrs loa_lma;
312 /* Below five elements are for OST-object's PFID EA, the
313 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
314 * and the stripe_index (low 16 bits), the size should not exceed
315 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
316 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
317 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
318 * loa_comp_* elements are valid. */
319 struct lu_fid loa_parent_fid;
320 __u32 loa_stripe_size;
322 __u64 loa_comp_start;
327 * Prior to 2.4, the LMA structure also included SOM attributes which has since
328 * been moved to a dedicated xattr
329 * lma_flags was also removed because of lma_compat/incompat fields.
331 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
333 enum lustre_som_flags {
334 /* Unknow or no SoM data, must get size from OSTs. */
335 SOM_FL_UNKNOWN = 0x0000,
336 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
337 SOM_FL_STRICT = 0x0001,
338 /* Known stale - was right at some point in the past, but it is
339 * known (or likely) to be incorrect now (e.g. opened for write). */
340 SOM_FL_STALE = 0x0002,
341 /* Approximate, may never have been strictly correct,
342 * need to sync SOM data to achieve eventual consistency. */
343 SOM_FL_LAZY = 0x0004,
346 struct lustre_som_attrs {
348 __u16 lsa_reserved[3];
354 * OST object IDentifier.
362 struct lu_fid oi_fid;
366 #define DOSTID "%#llx:%llu"
367 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
368 ((unsigned long long)ostid_id(oi))
370 struct ll_futimes_3 {
372 __u64 lfu_atime_nsec;
374 __u64 lfu_mtime_nsec;
376 __u64 lfu_ctime_nsec;
380 * Maximum number of mirrors currently implemented.
382 #define LUSTRE_MIRROR_COUNT_MAX 16
384 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
386 LL_LEASE_RDLCK = 0x01,
387 LL_LEASE_WRLCK = 0x02,
388 LL_LEASE_UNLCK = 0x04,
391 enum ll_lease_flags {
392 LL_LEASE_RESYNC = 0x1,
393 LL_LEASE_RESYNC_DONE = 0x2,
394 LL_LEASE_LAYOUT_MERGE = 0x4,
395 LL_LEASE_LAYOUT_SPLIT = 0x8,
396 LL_LEASE_PCC_ATTACH = 0x10,
399 #define IOC_IDS_MAX 4096
400 struct ll_ioc_lease {
407 struct ll_ioc_lease_id {
418 * The ioctl naming rules:
419 * LL_* - works on the currently opened filehandle instead of parent dir
420 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
421 * *_MDC_* - gets/sets data related to MDC
422 * *_LOV_* - gets/sets data related to OSC/LOV
423 * *FILE* - called on parent dir and passes in a filename
424 * *STRIPE* - set/get lov_user_md
425 * *INFO - set/get lov_user_mds_data
427 /* lustre_ioctl.h 101-150 */
428 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
429 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
430 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
431 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
432 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
433 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
434 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
435 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
436 /* LL_IOC_RECREATE_OBJ 157 obsolete */
437 /* LL_IOC_RECREATE_FID 157 obsolete */
438 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
439 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
440 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
441 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
442 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
443 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
444 /* IOC_LOV_GETINFO 165 obsolete */
445 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
446 /* LL_IOC_RMTACL 167 obsolete */
447 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
448 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
449 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
450 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
451 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
452 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
453 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
454 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
455 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
456 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
457 /* lustre_ioctl.h 177-210 */
458 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
459 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
460 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
461 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
462 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
463 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
464 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
465 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
466 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
467 struct lustre_swap_layouts)
468 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
469 struct hsm_current_action)
470 /* lustre_ioctl.h 221-232 */
471 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
472 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
473 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
474 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
475 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
476 #define LL_IOC_GET_LEASE _IO('f', 244)
477 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
478 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
479 #define LL_IOC_MIGRATE _IOR('f', 247, int)
480 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
481 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
482 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
483 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
484 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
485 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
486 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
487 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
489 #ifndef FS_IOC_FSGETXATTR
491 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
494 __u32 fsx_xflags; /* xflags field value (get/set) */
495 __u32 fsx_extsize; /* extsize field value (get/set)*/
496 __u32 fsx_nextents; /* nextents field value (get) */
497 __u32 fsx_projid; /* project identifier (get/set) */
498 unsigned char fsx_pad[12];
500 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
501 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
503 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
504 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
505 #ifndef FS_XFLAG_PROJINHERIT
506 #define FS_XFLAG_PROJINHERIT 0x00000200
510 #define LL_STATFS_LMV 1
511 #define LL_STATFS_LOV 2
512 #define LL_STATFS_NODELAY 4
514 #define IOC_MDC_TYPE 'i'
515 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
516 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
517 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
518 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
520 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
522 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
523 * files, but are unlikely to be used in practice and are not harmful if
524 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
525 * devices and are safe for use on new files. See LU-4209. */
526 /* To be compatible with old statically linked binary we keep the check for
527 * the older 0100000000 flag. This is already removed upstream. LU-812. */
528 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
530 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
532 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
533 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
534 O_LOV_DELAY_CREATE_MASK)
536 #define LL_FILE_IGNORE_LOCK 0x00000001
537 #define LL_FILE_GROUP_LOCKED 0x00000002
538 #define LL_FILE_READAHEA 0x00000004
539 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
540 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
541 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
543 #define LOV_USER_MAGIC_V1 0x0BD10BD0
544 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
545 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
546 #define LOV_USER_MAGIC_V3 0x0BD30BD0
547 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
548 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
549 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
550 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
552 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
553 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
554 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
556 #define LOV_PATTERN_NONE 0x000
557 #define LOV_PATTERN_RAID0 0x001
558 #define LOV_PATTERN_RAID1 0x002
559 #define LOV_PATTERN_MDT 0x100
560 #define LOV_PATTERN_OVERSTRIPING 0x200
562 #define LOV_PATTERN_F_MASK 0xffff0000
563 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
564 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
565 #define LOV_PATTERN_DEFAULT 0xffffffff
567 #define LOV_OFFSET_DEFAULT ((__u16)-1)
569 static inline bool lov_pattern_supported(__u32 pattern)
571 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
572 (pattern & ~LOV_PATTERN_F_RELEASED) ==
573 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
574 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
577 /* RELEASED and MDT patterns are not valid in many places, so rather than
578 * having many extra checks on lov_pattern_supported, we have this separate
579 * check for non-released, non-DOM components
581 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
583 return pattern == LOV_PATTERN_RAID0 ||
584 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
588 #define LOV_MAXPOOLNAME 15
589 #define LOV_POOLNAMEF "%.15s"
591 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
592 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
593 #define LOV_MAX_STRIPE_COUNT_OLD 160
594 /* This calculation is crafted so that input of 4096 will result in 160
595 * which in turn is equal to old maximal stripe count.
596 * XXX: In fact this is too simpified for now, what it also need is to get
597 * ea_type argument to clearly know how much space each stripe consumes.
599 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
600 * allocation that is sufficient for the current generation of systems.
602 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
603 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
604 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
605 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
607 #define XATTR_LUSTRE_PREFIX "lustre."
608 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
610 #define lov_user_ost_data lov_user_ost_data_v1
611 struct lov_user_ost_data_v1 { /* per-stripe data structure */
612 struct ost_id l_ost_oi; /* OST object ID */
613 __u32 l_ost_gen; /* generation of this OST index */
614 __u32 l_ost_idx; /* OST index in LOV */
615 } __attribute__((packed));
617 #define lov_user_md lov_user_md_v1
618 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
619 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
620 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
621 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
622 __u32 lmm_stripe_size; /* size of stripe in bytes */
623 __u16 lmm_stripe_count; /* num stripes in use for this object */
625 __u16 lmm_stripe_offset; /* starting stripe offset in
626 * lmm_objects, use when writing */
627 __u16 lmm_layout_gen; /* layout generation number
628 * used when reading */
630 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
631 } __attribute__((packed, __may_alias__));
633 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
634 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
635 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
636 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
637 __u32 lmm_stripe_size; /* size of stripe in bytes */
638 __u16 lmm_stripe_count; /* num stripes in use for this object */
640 __u16 lmm_stripe_offset; /* starting stripe offset in
641 * lmm_objects, use when writing */
642 __u16 lmm_layout_gen; /* layout generation number
643 * used when reading */
645 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
646 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
647 } __attribute__((packed));
649 struct lov_foreign_md {
650 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
651 __u32 lfm_length; /* length of lfm_value */
652 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
653 __u32 lfm_flags; /* flags, type specific */
657 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
658 offsetof(struct lov_foreign_md, lfm_value))
660 #define foreign_size_le(lfm) \
661 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
662 offsetof(struct lov_foreign_md, lfm_value))
669 #define DEXT "[%#llx, %#llx)"
670 #define PEXT(ext) (ext)->e_start, (ext)->e_end
672 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
673 struct lu_extent *e2)
675 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
678 static inline bool lu_extent_is_whole(struct lu_extent *e)
680 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
683 enum lov_comp_md_entry_flags {
684 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
685 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
686 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
687 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
688 LCME_FL_OFFLINE = 0x00000008, /* Not used */
689 LCME_FL_INIT = 0x00000010, /* instantiated */
690 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
691 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
692 won't be stored on disk */
695 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
696 LCME_FL_PREF_RW | LCME_FL_NOSYNC)
697 /* The flags can be set by users at mirror creation time. */
698 #define LCME_USER_FLAGS (LCME_FL_PREF_RW)
700 /* The flags are for mirrors */
701 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
703 /* These flags have meaning when set in a default layout and will be inherited
704 * from the default/template layout set on a directory.
706 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC)
708 /* the highest bit in obdo::o_layout_version is used to mark if the file is
710 #define LU_LAYOUT_RESYNC LCME_FL_NEG
712 /* lcme_id can be specified as certain flags, and the the first
713 * bit of lcme_id is used to indicate that the ID is representing
714 * certain LCME_FL_* but not a real ID. Which implies we can have
715 * at most 31 flags (see LCME_FL_XXX). */
718 LCME_ID_MAX = 0x7FFFFFFF,
719 LCME_ID_ALL = 0xFFFFFFFF,
720 LCME_ID_NOT_ID = LCME_FL_NEG
723 #define LCME_ID_MASK LCME_ID_MAX
725 struct lov_comp_md_entry_v1 {
726 __u32 lcme_id; /* unique id of component */
727 __u32 lcme_flags; /* LCME_FL_XXX */
728 struct lu_extent lcme_extent; /* file extent for component */
729 __u32 lcme_offset; /* offset of component blob,
730 start from lov_comp_md_v1 */
731 __u32 lcme_size; /* size of component blob */
732 __u32 lcme_layout_gen;
733 __u64 lcme_timestamp; /* snapshot time if applicable*/
734 __u32 lcme_padding_1;
735 } __attribute__((packed));
737 #define SEQ_ID_MAX 0x0000FFFF
738 #define SEQ_ID_MASK SEQ_ID_MAX
739 /* bit 30:16 of lcme_id is used to store mirror id */
740 #define MIRROR_ID_MASK 0x7FFF0000
741 #define MIRROR_ID_NEG 0x8000
742 #define MIRROR_ID_SHIFT 16
744 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
746 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
749 static inline __u16 mirror_id_of(__u32 id)
751 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
755 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
757 enum lov_comp_md_flags {
758 /* the least 2 bits are used by FLR to record file state */
761 LCM_FL_WRITE_PENDING = 2,
762 LCM_FL_SYNC_PENDING = 3,
763 LCM_FL_FLR_MASK = 0x3,
766 struct lov_comp_md_v1 {
767 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
768 __u32 lcm_size; /* overall size including this struct */
769 __u32 lcm_layout_gen;
771 __u16 lcm_entry_count;
772 /* lcm_mirror_count stores the number of actual mirrors minus 1,
773 * so that non-flr files will have value 0 meaning 1 mirror. */
774 __u16 lcm_mirror_count;
775 __u16 lcm_padding1[3];
777 struct lov_comp_md_entry_v1 lcm_entries[0];
778 } __attribute__((packed));
780 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
782 if (stripes == (__u16)-1)
785 if (lmm_magic == LOV_USER_MAGIC_V1)
786 return sizeof(struct lov_user_md_v1) +
787 stripes * sizeof(struct lov_user_ost_data_v1);
788 return sizeof(struct lov_user_md_v3) +
789 stripes * sizeof(struct lov_user_ost_data_v1);
792 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
793 * use this. It is unsafe to #define those values in this header as it
794 * is possible the application has already #included <sys/stat.h>. */
795 #ifdef HAVE_LOV_USER_MDS_DATA
796 #define lov_user_mds_data lov_user_mds_data_v1
797 struct lov_user_mds_data_v1 {
798 lstat_t lmd_st; /* MDS stat struct */
799 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
800 } __attribute__((packed));
803 struct lmv_user_mds_data {
804 struct lu_fid lum_fid;
810 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
811 LMV_HASH_TYPE_ALL_CHARS = 1,
812 LMV_HASH_TYPE_FNV_1A_64 = 2,
816 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
818 #define LMV_HASH_NAME_ALL_CHARS "all_char"
819 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
821 /* not real hash type, but exposed to user as "space" hash type */
822 #define LMV_HASH_NAME_SPACE "space"
824 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
825 * and the higher part will be the flag to indicate the status of object,
826 * for example the object is being migrated. And the hash function
827 * might be interpreted differently with different flags. */
828 #define LMV_HASH_TYPE_MASK 0x0000ffff
830 /* once this is set on a plain directory default layout, newly created
831 * subdirectories will be distributed on all MDTs by space usage.
833 #define LMV_HASH_FLAG_SPACE 0x08000000
835 /* The striped directory has ever lost its master LMV EA, then LFSCK
836 * re-generated it. This flag is used to indicate such case. It is an
838 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
840 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
841 #define LMV_HASH_FLAG_MIGRATION 0x80000000
843 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
845 struct lustre_foreign_type {
847 const char *lft_name;
851 * LOV/LMV foreign types
853 enum lustre_foreign_types {
854 LU_FOREIGN_TYPE_NONE = 0,
855 LU_FOREIGN_TYPE_DAOS = 0xda05,
856 /* must be the max/last one */
857 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
860 extern struct lustre_foreign_type lu_foreign_types[];
862 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
863 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
864 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
865 #define lmv_user_md lmv_user_md_v1
866 struct lmv_user_md_v1 {
867 __u32 lum_magic; /* must be the first field */
868 __u32 lum_stripe_count; /* dirstripe count */
869 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
870 __u32 lum_hash_type; /* Dir stripe policy */
871 __u32 lum_type; /* LMV type: default */
875 char lum_pool_name[LOV_MAXPOOLNAME + 1];
876 struct lmv_user_mds_data lum_objects[0];
877 } __attribute__((packed));
879 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
881 if (size <= sizeof(struct lmv_user_md))
884 size -= sizeof(struct lmv_user_md);
885 return (size + sizeof(struct lmv_user_mds_data) - 1) /
886 sizeof(struct lmv_user_mds_data);
890 * NB, historically default layout didn't set type, but use XATTR name to differ
891 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
892 * and still use the same method.
895 LMV_TYPE_DEFAULT = 0x0000,
898 static inline int lmv_user_md_size(int stripes, int lmm_magic)
900 int size = sizeof(struct lmv_user_md);
902 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
903 size += stripes * sizeof(struct lmv_user_mds_data);
908 struct ll_recreate_obj {
914 __u64 id; /* holds object id */
915 __u32 generation; /* holds object generation */
916 __u32 f_type; /* holds object type or stripe idx when passing it to
917 * OST for saving into EA. */
925 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
926 const struct obd_uuid *u2)
928 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
931 static inline int obd_uuid_empty(struct obd_uuid *uuid)
933 return uuid->uuid[0] == '\0';
936 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
938 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
939 uuid->uuid[sizeof(*uuid) - 1] = '\0';
942 /* For printf's only, make sure uuid is terminated */
943 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
948 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
949 /* Obviously not safe, but for printfs, no real harm done...
950 we're always null-terminated, even in a race. */
951 static char temp[sizeof(*uuid)];
952 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
953 temp[sizeof(*uuid) - 1] = '\0';
956 return (char *)(uuid->uuid);
959 #define LUSTRE_MAXFSNAME 8
961 /* Extract fsname from uuid (or target name) of a target
962 e.g. (myfs-OST0007_UUID -> myfs)
963 see also deuuidify. */
964 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
968 strncpy(buf, uuid, buflen - 1);
969 buf[buflen - 1] = '\0';
970 p = strrchr(buf, '-');
975 /* printf display format for Lustre FIDs
976 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
977 #define FID_NOBRACE_LEN 40
978 #define FID_LEN (FID_NOBRACE_LEN + 2)
979 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
980 #define DFID "["DFID_NOBRACE"]"
981 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
983 /* scanf input parse format for fids in DFID_NOBRACE format
984 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
985 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
986 #define SFID "0x%llx:0x%x:0x%x"
987 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
989 /********* Quotas **********/
991 #define LUSTRE_QUOTABLOCK_BITS 10
992 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
994 static inline __u64 lustre_stoqb(size_t space)
996 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
999 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1000 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1001 #define Q_GETOINFO 0x800102 /* get obd quota info */
1002 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1003 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1005 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1006 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1007 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1008 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1009 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1010 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1011 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1012 /* lustre-specific control commands */
1013 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1014 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1015 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1016 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1018 /* In the current Lustre implementation, the grace time is either the time
1019 * or the timestamp to be used after some quota ID exceeds the soft limt,
1020 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1022 #define LQUOTA_GRACE_BITS 48
1023 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1024 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1025 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1026 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1027 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1029 /* different quota flags */
1031 /* the default quota flag, the corresponding quota ID will use the default
1032 * quota setting, the hardlimit and softlimit of its quota record in the global
1033 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1034 * and high 16 bits will contain this flag (see above comment).
1036 #define LQUOTA_FLAG_DEFAULT 0x0001
1038 #define ALLQUOTA 255 /* set all quota */
1039 static inline char *qtype_name(int qtype)
1052 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1053 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1056 #define N_PERMS_MAX 64
1058 struct perm_downcall_data {
1064 struct identity_downcall_data {
1071 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1072 __u32 idd_groups[0];
1075 struct sepol_downcall_data {
1077 time_t sdd_sepol_mtime;
1078 __u16 sdd_sepol_len;
1082 #ifdef NEED_QUOTA_DEFS
1084 #define QIF_BLIMITS 1
1086 #define QIF_ILIMITS 4
1087 #define QIF_INODES 8
1088 #define QIF_BTIME 16
1089 #define QIF_ITIME 32
1090 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1091 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1092 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1093 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1096 #endif /* !__KERNEL__ */
1098 /* lustre volatile file support
1099 * file name header: ".^L^S^T^R:volatile"
1101 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1102 #define LUSTRE_VOLATILE_HDR_LEN 14
1104 enum lustre_quota_version {
1108 /* XXX: same as if_dqinfo struct in kernel */
1116 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1118 __u64 dqb_bhardlimit;
1119 __u64 dqb_bsoftlimit;
1121 __u64 dqb_ihardlimit;
1122 __u64 dqb_isoftlimit;
1123 __u64 dqb_curinodes;
1137 struct if_quotactl {
1144 struct obd_dqinfo qc_dqinfo;
1145 struct obd_dqblk qc_dqblk;
1147 struct obd_uuid obd_uuid;
1150 /* swap layout flags */
1151 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1152 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1153 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1154 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1155 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1157 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1158 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1159 struct lustre_swap_layouts {
1167 /** Bit-mask of valid attributes */
1168 /* The LA_* flags are written to disk as part of the ChangeLog records
1169 * so they are part of the on-disk and network protocol, and cannot be changed.
1170 * Only the first 12 bits are currently saved.
1173 LA_ATIME = 1 << 0, /* 0x00001 */
1174 LA_MTIME = 1 << 1, /* 0x00002 */
1175 LA_CTIME = 1 << 2, /* 0x00004 */
1176 LA_SIZE = 1 << 3, /* 0x00008 */
1177 LA_MODE = 1 << 4, /* 0x00010 */
1178 LA_UID = 1 << 5, /* 0x00020 */
1179 LA_GID = 1 << 6, /* 0x00040 */
1180 LA_BLOCKS = 1 << 7, /* 0x00080 */
1181 LA_TYPE = 1 << 8, /* 0x00100 */
1182 LA_FLAGS = 1 << 9, /* 0x00200 */
1183 LA_NLINK = 1 << 10, /* 0x00400 */
1184 LA_RDEV = 1 << 11, /* 0x00800 */
1185 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1186 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1187 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1188 LA_PROJID = 1 << 15, /* 0x08000 */
1189 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1190 LA_LSIZE = 1 << 17, /* 0x20000 */
1191 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1193 * Attributes must be transmitted to OST objects
1195 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1198 #define MDS_FMODE_READ 00000001
1199 #define MDS_FMODE_WRITE 00000002
1201 #define MDS_FMODE_CLOSED 00000000
1202 #define MDS_FMODE_EXEC 00000004
1203 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1204 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1205 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1207 #define MDS_OPEN_CREATED 00000010
1208 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1210 #define MDS_OPEN_CREAT 00000100
1211 #define MDS_OPEN_EXCL 00000200
1212 #define MDS_OPEN_TRUNC 00001000
1213 #define MDS_OPEN_APPEND 00002000
1214 #define MDS_OPEN_SYNC 00010000
1215 #define MDS_OPEN_DIRECTORY 00200000
1217 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1218 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1219 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1220 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1221 * We do not support JOIN FILE
1222 * anymore, reserve this flags
1223 * just for preventing such bit
1226 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1227 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1228 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1229 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1230 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1232 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1234 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1235 * delegation, succeed if it's not
1236 * being opened with conflict mode.
1238 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1240 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1241 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1242 * for newly created file */
1244 /* lustre internal open flags, which should not be set from user space */
1245 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1246 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1247 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1248 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1252 /********* Changelogs **********/
1253 /** Changelog record types */
1254 enum changelog_rec_type {
1257 CL_CREATE = 1, /* namespace */
1258 CL_MKDIR = 2, /* namespace */
1259 CL_HARDLINK = 3, /* namespace */
1260 CL_SOFTLINK = 4, /* namespace */
1261 CL_MKNOD = 5, /* namespace */
1262 CL_UNLINK = 6, /* namespace */
1263 CL_RMDIR = 7, /* namespace */
1264 CL_RENAME = 8, /* namespace */
1265 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1266 CL_OPEN = 10, /* not currently used */
1267 CL_CLOSE = 11, /* may be written to log only with mtime change */
1268 CL_LAYOUT = 12, /* file layout/striping modified */
1272 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1273 CL_HSM = 16, /* HSM specific events, see flags */
1274 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1278 CL_FLRW = 21, /* FLR: file was firstly written */
1279 CL_RESYNC = 22, /* FLR: file was resync-ed */
1281 CL_DN_OPEN = 24, /* denied open */
1285 static inline const char *changelog_type2str(int type) {
1286 static const char *changelog_str[] = {
1287 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1288 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1289 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1290 "FLRW", "RESYNC","GXATR", "NOPEN",
1293 if (type >= 0 && type < CL_LAST)
1294 return changelog_str[type];
1298 /* 12 bits of per-record data can be stored in the bottom of the flags */
1299 #define CLF_FLAGSHIFT 12
1300 enum changelog_rec_flags {
1301 CLF_VERSION = 0x1000,
1302 CLF_RENAME = 0x2000,
1304 CLF_EXTRA_FLAGS = 0x8000,
1305 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1307 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1308 CLF_VERMASK = ~CLF_FLAGMASK,
1312 /* Anything under the flagmask may be per-type (if desired) */
1313 /* Flags for unlink */
1314 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1315 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1316 /* HSM cleaning needed */
1317 /* Flags for rename */
1318 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1320 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1321 * has an archive in backend */
1324 /* 12b used (from high weight to low weight):
1329 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1330 #define CLF_HSM_ERR_H 6
1331 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1332 #define CLF_HSM_EVENT_H 9
1333 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1334 #define CLF_HSM_FLAG_H 11
1335 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1336 #define CLF_HSM_SPARE_H 15
1337 #define CLF_HSM_LAST 15
1339 /* Remove bits higher than _h, then extract the value
1340 * between _h and _l by shifting lower weigth to bit 0. */
1341 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1342 >> (CLF_HSM_LAST - _h + _l))
1344 #define CLF_HSM_SUCCESS 0x00
1345 #define CLF_HSM_MAXERROR 0x7E
1346 #define CLF_HSM_ERROVERFLOW 0x7F
1348 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1350 /* 3 bits field => 8 values allowed */
1362 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1364 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1368 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1371 *clf_flags |= (he << CLF_HSM_EVENT_L);
1374 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1376 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1379 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1382 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1385 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1387 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1390 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1393 *clf_flags |= (error << CLF_HSM_ERR_L);
1396 enum changelog_rec_extra_flags {
1398 CLFE_UIDGID = 0x0001,
1401 CLFE_XATTR = 0x0008,
1402 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1405 enum changelog_send_flag {
1406 /* Not yet implemented */
1407 CHANGELOG_FLAG_FOLLOW = 0x01,
1408 /* Blocking IO makes sense in case of slow user parsing of the records,
1409 * but it also prevents us from cleaning up if the records are not
1411 CHANGELOG_FLAG_BLOCK = 0x02,
1412 /* Pack jobid into the changelog records if available. */
1413 CHANGELOG_FLAG_JOBID = 0x04,
1414 /* Pack additional flag bits into the changelog record */
1415 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1418 enum changelog_send_extra_flag {
1419 /* Pack uid/gid into the changelog record */
1420 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1421 /* Pack nid into the changelog record */
1422 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1423 /* Pack open mode into the changelog record */
1424 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1425 /* Pack xattr name into the changelog record */
1426 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1429 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1430 changelog_rec_offset(CLF_SUPPORTED, \
1433 /* 31 usable bytes string + null terminator. */
1434 #define LUSTRE_JOBID_SIZE 32
1436 /* This is the minimal changelog record. It can contain extensions
1437 * such as rename fields or process jobid. Its exact content is described
1438 * by the cr_flags and cr_extra_flags.
1440 * Extensions are packed in the same order as their corresponding flags,
1441 * then in the same order as their corresponding extra flags.
1443 struct changelog_rec {
1445 __u16 cr_flags; /**< \a changelog_rec_flags */
1446 __u32 cr_type; /**< \a changelog_rec_type */
1447 __u64 cr_index; /**< changelog record number */
1448 __u64 cr_prev; /**< last index for this target fid */
1451 struct lu_fid cr_tfid; /**< target fid */
1452 __u32 cr_markerflags; /**< CL_MARK flags */
1454 struct lu_fid cr_pfid; /**< parent fid */
1457 /* Changelog extension for RENAME. */
1458 struct changelog_ext_rename {
1459 struct lu_fid cr_sfid; /**< source fid, or zero */
1460 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1463 /* Changelog extension to include JOBID. */
1464 struct changelog_ext_jobid {
1465 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1468 /* Changelog extension to include additional flags. */
1469 struct changelog_ext_extra_flags {
1470 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1473 /* Changelog extra extension to include UID/GID. */
1474 struct changelog_ext_uidgid {
1479 /* Changelog extra extension to include NID. */
1480 struct changelog_ext_nid {
1481 /* have __u64 instead of lnet_nid_t type for use by client api */
1483 /* for use when IPv6 support is added */
1488 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1489 struct changelog_ext_openmode {
1493 /* Changelog extra extension to include xattr */
1494 struct changelog_ext_xattr {
1495 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1498 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1499 const struct changelog_rec *rec);
1501 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1502 enum changelog_rec_extra_flags cref)
1504 size_t size = sizeof(struct changelog_rec);
1506 if (crf & CLF_RENAME)
1507 size += sizeof(struct changelog_ext_rename);
1509 if (crf & CLF_JOBID)
1510 size += sizeof(struct changelog_ext_jobid);
1512 if (crf & CLF_EXTRA_FLAGS) {
1513 size += sizeof(struct changelog_ext_extra_flags);
1514 if (cref & CLFE_UIDGID)
1515 size += sizeof(struct changelog_ext_uidgid);
1516 if (cref & CLFE_NID)
1517 size += sizeof(struct changelog_ext_nid);
1518 if (cref & CLFE_OPEN)
1519 size += sizeof(struct changelog_ext_openmode);
1520 if (cref & CLFE_XATTR)
1521 size += sizeof(struct changelog_ext_xattr);
1527 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
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;
1534 return changelog_rec_offset(rec->cr_flags, cref);
1537 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1539 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1543 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1545 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1547 return (struct changelog_ext_rename *)((char *)rec +
1548 changelog_rec_offset(crf,
1552 /* The jobid follows the rename extension, if present */
1554 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1556 enum changelog_rec_flags crf = rec->cr_flags &
1557 (CLF_VERSION | CLF_RENAME);
1559 return (struct changelog_ext_jobid *)((char *)rec +
1560 changelog_rec_offset(crf,
1564 /* The additional flags follow the rename and jobid extensions, if present */
1566 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1567 const struct changelog_rec *rec)
1569 enum changelog_rec_flags crf = rec->cr_flags &
1570 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1572 return (struct changelog_ext_extra_flags *)((char *)rec +
1573 changelog_rec_offset(crf,
1577 /* The uid/gid is the first extra extension */
1579 struct changelog_ext_uidgid *changelog_rec_uidgid(
1580 const struct changelog_rec *rec)
1582 enum changelog_rec_flags crf = rec->cr_flags &
1583 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1585 return (struct changelog_ext_uidgid *)((char *)rec +
1586 changelog_rec_offset(crf,
1590 /* The nid is the second extra extension */
1592 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1594 enum changelog_rec_flags crf = rec->cr_flags &
1595 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1596 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1598 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1599 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1602 return (struct changelog_ext_nid *)((char *)rec +
1603 changelog_rec_offset(crf, cref));
1606 /* The OPEN mode is the third extra extension */
1608 struct changelog_ext_openmode *changelog_rec_openmode(
1609 const struct changelog_rec *rec)
1611 enum changelog_rec_flags crf = rec->cr_flags &
1612 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1613 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1615 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1616 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1617 (CLFE_UIDGID | CLFE_NID);
1619 return (struct changelog_ext_openmode *)((char *)rec +
1620 changelog_rec_offset(crf, cref));
1623 /* The xattr name is the fourth extra extension */
1625 struct changelog_ext_xattr *changelog_rec_xattr(
1626 const struct changelog_rec *rec)
1628 enum changelog_rec_flags crf = rec->cr_flags &
1629 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1630 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1632 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1633 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1634 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1636 return (struct changelog_ext_xattr *)((char *)rec +
1637 changelog_rec_offset(crf, cref));
1640 /* The name follows the rename, jobid and extra flags extns, if present */
1641 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1643 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1645 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1646 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1648 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1649 cref & CLFE_SUPPORTED);
1652 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1654 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1657 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1659 char *cr_name = changelog_rec_name(rec);
1661 return cr_name + strlen(cr_name) + 1;
1665 * Remap a record to the desired format as specified by the crf flags.
1666 * The record must be big enough to contain the final remapped version.
1667 * Superfluous extension fields are removed and missing ones are added
1668 * and zeroed. The flags of the record are updated accordingly.
1670 * The jobid and rename extensions can be added to a record, to match the
1671 * format an application expects, typically. In this case, the newly added
1672 * fields will be zeroed.
1673 * The Jobid field can be removed, to guarantee compatibility with older
1674 * clients that don't expect this field in the records they process.
1676 * The following assumptions are being made:
1677 * - CLF_RENAME will not be removed
1678 * - CLF_JOBID will not be added without CLF_RENAME being added too
1679 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1681 * @param[in,out] rec The record to remap.
1682 * @param[in] crf_wanted Flags describing the desired extensions.
1683 * @param[in] cref_want Flags describing the desired extra extensions.
1685 static inline void changelog_remap_rec(struct changelog_rec *rec,
1686 enum changelog_rec_flags crf_wanted,
1687 enum changelog_rec_extra_flags cref_want)
1689 char *xattr_mov = NULL;
1690 char *omd_mov = NULL;
1691 char *nid_mov = NULL;
1692 char *uidgid_mov = NULL;
1696 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1698 crf_wanted &= CLF_SUPPORTED;
1699 cref_want &= CLFE_SUPPORTED;
1701 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1702 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1703 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1704 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1710 /* First move the variable-length name field */
1711 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1712 changelog_rec_name(rec), rec->cr_namelen);
1714 /* Locations of extensions in the remapped record */
1715 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1716 xattr_mov = (char *)rec +
1717 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1718 cref_want & ~CLFE_XATTR);
1719 omd_mov = (char *)rec +
1720 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1721 cref_want & ~(CLFE_OPEN |
1723 nid_mov = (char *)rec +
1724 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1725 cref_want & ~(CLFE_NID |
1728 uidgid_mov = (char *)rec +
1729 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1730 cref_want & ~(CLFE_UIDGID |
1734 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1737 ef_mov = (char *)rec +
1738 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1740 jid_mov = (char *)rec +
1741 changelog_rec_offset(crf_wanted &
1742 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1744 rnm_mov = (char *)rec +
1745 changelog_rec_offset(crf_wanted &
1751 /* Move the extension fields to the desired positions */
1752 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1753 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1754 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1755 memmove(xattr_mov, changelog_rec_xattr(rec),
1756 sizeof(struct changelog_ext_xattr));
1758 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1759 memmove(omd_mov, changelog_rec_openmode(rec),
1760 sizeof(struct changelog_ext_openmode));
1762 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1763 memmove(nid_mov, changelog_rec_nid(rec),
1764 sizeof(struct changelog_ext_nid));
1766 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1767 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1768 sizeof(struct changelog_ext_uidgid));
1770 memmove(ef_mov, changelog_rec_extra_flags(rec),
1771 sizeof(struct changelog_ext_extra_flags));
1774 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1775 memmove(jid_mov, changelog_rec_jobid(rec),
1776 sizeof(struct changelog_ext_jobid));
1778 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1779 memmove(rnm_mov, changelog_rec_rename(rec),
1780 sizeof(struct changelog_ext_rename));
1782 /* Clear newly added fields */
1783 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1784 !(cref & CLFE_XATTR))
1785 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1787 if (omd_mov && (cref_want & CLFE_OPEN) &&
1788 !(cref & CLFE_OPEN))
1789 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1791 if (nid_mov && (cref_want & CLFE_NID) &&
1793 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1795 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1796 !(cref & CLFE_UIDGID))
1797 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1799 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1800 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1801 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1803 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1804 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1806 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1807 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1809 /* Update the record's flags accordingly */
1810 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1811 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1812 changelog_rec_extra_flags(rec)->cr_extra_flags =
1813 changelog_rec_extra_flags(rec)->cr_extra_flags |
1817 enum changelog_message_type {
1818 CL_RECORD = 10, /* message is a changelog_rec */
1819 CL_EOF = 11, /* at end of current changelog */
1822 /********* Misc **********/
1824 struct ioc_data_version {
1826 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1827 __u32 idv_flags; /* enum ioc_data_version_flags */
1830 enum ioc_data_version_flags {
1831 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1832 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1836 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1839 #define dot_lustre_name ".lustre"
1842 /********* HSM **********/
1844 /** HSM per-file state
1845 * See HSM_FLAGS below.
1848 HS_NONE = 0x00000000,
1849 HS_EXISTS = 0x00000001,
1850 HS_DIRTY = 0x00000002,
1851 HS_RELEASED = 0x00000004,
1852 HS_ARCHIVED = 0x00000008,
1853 HS_NORELEASE = 0x00000010,
1854 HS_NOARCHIVE = 0x00000020,
1855 HS_LOST = 0x00000040,
1858 /* HSM user-setable flags. */
1859 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1861 /* Other HSM flags. */
1862 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1865 * All HSM-related possible flags that could be applied to a file.
1866 * This should be kept in sync with hsm_states.
1868 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1871 * HSM request progress state
1873 enum hsm_progress_states {
1880 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1883 case HPS_WAITING: return "waiting";
1884 case HPS_RUNNING: return "running";
1885 case HPS_DONE: return "done";
1886 default: return "unknown";
1893 } __attribute__((packed));
1896 * Current HSM states of a Lustre file.
1898 * This structure purpose is to be sent to user-space mainly. It describes the
1899 * current HSM flags and in-progress action.
1901 struct hsm_user_state {
1902 /** Current HSM states, from enum hsm_states. */
1904 __u32 hus_archive_id;
1905 /** The current undergoing action, if there is one */
1906 __u32 hus_in_progress_state;
1907 __u32 hus_in_progress_action;
1908 struct hsm_extent hus_in_progress_location;
1909 char hus_extended_info[];
1912 struct hsm_state_set_ioc {
1913 struct lu_fid hssi_fid;
1915 __u64 hssi_clearmask;
1919 * This structure describes the current in-progress action for a file.
1920 * it is retuned to user space and send over the wire
1922 struct hsm_current_action {
1923 /** The current undergoing action, if there is one */
1924 /* state is one of hsm_progress_states */
1926 /* action is one of hsm_user_action */
1928 struct hsm_extent hca_location;
1931 /***** HSM user requests ******/
1932 /* User-generated (lfs/ioctl) request types */
1933 enum hsm_user_action {
1934 HUA_NONE = 1, /* no action (noop) */
1935 HUA_ARCHIVE = 10, /* copy to hsm */
1936 HUA_RESTORE = 11, /* prestage */
1937 HUA_RELEASE = 12, /* drop ost objects */
1938 HUA_REMOVE = 13, /* remove from archive */
1939 HUA_CANCEL = 14 /* cancel a request */
1942 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1945 case HUA_NONE: return "NOOP";
1946 case HUA_ARCHIVE: return "ARCHIVE";
1947 case HUA_RESTORE: return "RESTORE";
1948 case HUA_RELEASE: return "RELEASE";
1949 case HUA_REMOVE: return "REMOVE";
1950 case HUA_CANCEL: return "CANCEL";
1951 default: return "UNKNOWN";
1956 * List of hr_flags (bit field)
1958 #define HSM_FORCE_ACTION 0x0001
1959 /* used by CT, cannot be set by user */
1960 #define HSM_GHOST_COPY 0x0002
1963 * Contains all the fixed part of struct hsm_user_request.
1966 struct hsm_request {
1967 __u32 hr_action; /* enum hsm_user_action */
1968 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1969 __u64 hr_flags; /* request flags */
1970 __u32 hr_itemcount; /* item count in hur_user_item vector */
1974 struct hsm_user_item {
1975 struct lu_fid hui_fid;
1976 struct hsm_extent hui_extent;
1977 } __attribute__((packed));
1979 struct hsm_user_request {
1980 struct hsm_request hur_request;
1981 struct hsm_user_item hur_user_item[0];
1982 /* extra data blob at end of struct (after all
1983 * hur_user_items), only use helpers to access it
1985 } __attribute__((packed));
1987 /** Return pointer to data field in a hsm user request */
1988 static inline void *hur_data(struct hsm_user_request *hur)
1990 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1994 * Compute the current length of the provided hsm_user_request. This returns -1
1995 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1997 * return -1 on bounds check error.
1999 static inline ssize_t hur_len(struct hsm_user_request *hur)
2003 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2004 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2005 (__u64)hur->hur_request.hr_itemcount *
2006 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2008 if (size != (ssize_t)size)
2014 /****** HSM RPCs to copytool *****/
2015 /* Message types the copytool may receive */
2016 enum hsm_message_type {
2017 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2020 /* Actions the copytool may be instructed to take for a given action_item */
2021 enum hsm_copytool_action {
2022 HSMA_NONE = 10, /* no action */
2023 HSMA_ARCHIVE = 20, /* arbitrary offset */
2029 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2032 case HSMA_NONE: return "NOOP";
2033 case HSMA_ARCHIVE: return "ARCHIVE";
2034 case HSMA_RESTORE: return "RESTORE";
2035 case HSMA_REMOVE: return "REMOVE";
2036 case HSMA_CANCEL: return "CANCEL";
2037 default: return "UNKNOWN";
2041 /* Copytool item action description */
2042 struct hsm_action_item {
2043 __u32 hai_len; /* valid size of this struct */
2044 __u32 hai_action; /* hsm_copytool_action, but use known size */
2045 struct lu_fid hai_fid; /* Lustre FID to operate on */
2046 struct lu_fid hai_dfid; /* fid used for data access */
2047 struct hsm_extent hai_extent; /* byte range to operate on */
2048 __u64 hai_cookie; /* action cookie from coordinator */
2049 __u64 hai_gid; /* grouplock id */
2050 char hai_data[0]; /* variable length */
2051 } __attribute__((packed));
2054 * helper function which print in hexa the first bytes of
2057 * \param hai [IN] record to print
2058 * \param buffer [IN,OUT] buffer to write the hex string to
2059 * \param len [IN] max buffer length
2063 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2064 char *buffer, size_t len)
2071 data_len = hai->hai_len - sizeof(*hai);
2072 for (i = 0; (i < data_len) && (len > 2); i++) {
2073 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2083 /* Copytool action list */
2084 #define HAL_VERSION 1
2085 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2086 struct hsm_action_list {
2088 __u32 hal_count; /* number of hai's to follow */
2089 __u64 hal_compound_id; /* returned by coordinator, ignored */
2091 __u32 hal_archive_id; /* which archive backend */
2093 char hal_fsname[0]; /* null-terminated */
2094 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2095 boundaries. See hai_zero */
2096 } __attribute__((packed));
2098 /* Return pointer to first hai in action list */
2099 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2101 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2103 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2106 /* Return pointer to next hai */
2107 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2109 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2111 return (struct hsm_action_item *)((char *)hai + offset);
2114 /* Return size of an hsm_action_list */
2115 static inline size_t hal_size(struct hsm_action_list *hal)
2119 struct hsm_action_item *hai;
2121 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2122 hai = hai_first(hal);
2123 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2124 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2130 * describe the attributes to be set on imported file
2132 struct hsm_user_import {
2141 __u32 hui_archive_id;
2144 /* Copytool progress reporting */
2145 #define HP_FLAG_COMPLETED 0x01
2146 #define HP_FLAG_RETRY 0x02
2148 struct hsm_progress {
2149 struct lu_fid hp_fid;
2151 struct hsm_extent hp_extent;
2153 __u16 hp_errval; /* positive val */
2158 __u64 hc_data_version;
2160 __u16 hc_errval; /* positive val */
2162 struct hsm_action_item hc_hai;
2166 enum llapi_json_types {
2167 LLAPI_JSON_INTEGER = 1,
2173 struct llapi_json_item {
2182 struct llapi_json_item *lji_next;
2185 struct llapi_json_item_list {
2186 int ljil_item_count;
2187 struct llapi_json_item *ljil_items;
2190 enum lu_ladvise_type {
2191 LU_LADVISE_INVALID = 0,
2192 LU_LADVISE_WILLREAD = 1,
2193 LU_LADVISE_DONTNEED = 2,
2194 LU_LADVISE_LOCKNOEXPAND = 3,
2195 LU_LADVISE_LOCKAHEAD = 4,
2199 #define LU_LADVISE_NAMES { \
2200 [LU_LADVISE_WILLREAD] = "willread", \
2201 [LU_LADVISE_DONTNEED] = "dontneed", \
2202 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2203 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2206 /* This is the userspace argument for ladvise. It is currently the same as
2207 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2208 * may need info which is only used locally. */
2209 struct llapi_lu_ladvise {
2210 __u16 lla_advice; /* advice type */
2211 __u16 lla_value1; /* values for different advice types */
2213 __u64 lla_start; /* first byte of extent for advice */
2214 __u64 lla_end; /* last byte of extent for advice */
2220 LF_ASYNC = 0x00000001,
2221 LF_UNSET = 0x00000002,
2224 #define LADVISE_MAGIC 0x1ADF1CE0
2225 /* Masks of valid flags for each advice */
2226 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2227 /* Flags valid for all advices not explicitly specified */
2228 #define LF_DEFAULT_MASK LF_ASYNC
2230 #define LF_MASK (LF_ASYNC | LF_UNSET)
2232 #define lla_lockahead_mode lla_value1
2233 #define lla_peradvice_flags lla_value2
2234 #define lla_lockahead_result lla_value3
2236 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2237 * is used on the wire. It is defined separately as we may need info which is
2238 * only used locally. */
2239 struct llapi_ladvise_hdr {
2240 __u32 lah_magic; /* LADVISE_MAGIC */
2241 __u32 lah_count; /* number of advices */
2242 __u64 lah_flags; /* from enum ladvise_flag */
2243 __u32 lah_value1; /* unused */
2244 __u32 lah_value2; /* unused */
2245 __u64 lah_value3; /* unused */
2246 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2249 #define LAH_COUNT_MAX (1024)
2253 SK_CRYPT_INVALID = -1,
2255 SK_CRYPT_AES256_CTR = 1,
2259 SK_HMAC_INVALID = -1,
2265 struct sk_crypt_type {
2266 const char *sct_name;
2270 struct sk_hmac_type {
2271 const char *sht_name;
2275 enum lock_mode_user {
2281 #define LOCK_MODE_NAMES { \
2282 [MODE_READ_USER] = "READ",\
2283 [MODE_WRITE_USER] = "WRITE"\
2286 enum lockahead_results {
2287 LLA_RESULT_SENT = 0,
2288 LLA_RESULT_DIFFERENT,
2292 enum lu_heat_flag_bit {
2293 LU_HEAT_FLAG_BIT_INVALID = 0,
2294 LU_HEAT_FLAG_BIT_OFF,
2295 LU_HEAT_FLAG_BIT_CLEAR,
2299 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2300 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2303 enum obd_heat_type {
2304 OBD_HEAT_READSAMPLE = 0,
2305 OBD_HEAT_WRITESAMPLE = 1,
2306 OBD_HEAT_READBYTE = 2,
2307 OBD_HEAT_WRITEBYTE = 3,
2311 #define LU_HEAT_NAMES { \
2312 [OBD_HEAT_READSAMPLE] = "readsample", \
2313 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2314 [OBD_HEAT_READBYTE] = "readbyte", \
2315 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2330 static inline const char *pcc_type2string(enum lu_pcc_type type)
2335 case LU_PCC_READWRITE:
2342 struct lu_pcc_attach {
2343 __u32 pcca_type; /* PCC type */
2344 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2347 enum lu_pcc_detach_opts {
2348 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2349 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2352 struct lu_pcc_detach_fid {
2353 /* fid of the file to detach */
2354 struct lu_fid pccd_fid;
2358 struct lu_pcc_detach {
2362 enum lu_pcc_state_flags {
2363 PCC_STATE_FL_NONE = 0x0,
2364 /* The inode attr is cached locally */
2365 PCC_STATE_FL_ATTR_VALID = 0x01,
2366 /* The file is being attached into PCC */
2367 PCC_STATE_FL_ATTACHING = 0x02,
2368 /* Allow to auto attach at open */
2369 PCC_STATE_FL_OPEN_ATTACH = 0x04,
2372 struct lu_pcc_state {
2373 __u32 pccs_type; /* enum lu_pcc_type */
2374 __u32 pccs_open_count;
2375 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2377 char pccs_path[PATH_MAX];
2380 #if defined(__cplusplus)
2384 /** @} lustreuser */
2386 #endif /* _LUSTRE_USER_H */