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))
665 * The stripe size fields are shared for the extension size storage, however
666 * the extension size is stored in KB, not bytes.
668 #define SEL_UNIT_SIZE 1024llu
675 #define DEXT "[%#llx, %#llx)"
676 #define PEXT(ext) (ext)->e_start, (ext)->e_end
678 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
679 struct lu_extent *e2)
681 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
684 static inline bool lu_extent_is_whole(struct lu_extent *e)
686 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
689 enum lov_comp_md_entry_flags {
690 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
691 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
692 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
693 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
694 LCME_FL_OFFLINE = 0x00000008, /* Not used */
695 LCME_FL_INIT = 0x00000010, /* instantiated */
696 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
697 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
698 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
699 * won't be stored on disk
703 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
704 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
707 /* The mirror flags can be set by users at creation time. */
708 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
710 /* The allowed flags obtained from the client at component creation time. */
711 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
713 /* The mirror flags sent by client */
714 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
716 /* These flags have meaning when set in a default layout and will be inherited
717 * from the default/template layout set on a directory.
719 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
722 /* the highest bit in obdo::o_layout_version is used to mark if the file is
724 #define LU_LAYOUT_RESYNC LCME_FL_NEG
726 /* lcme_id can be specified as certain flags, and the the first
727 * bit of lcme_id is used to indicate that the ID is representing
728 * certain LCME_FL_* but not a real ID. Which implies we can have
729 * at most 31 flags (see LCME_FL_XXX). */
732 LCME_ID_MAX = 0x7FFFFFFF,
733 LCME_ID_ALL = 0xFFFFFFFF,
734 LCME_ID_NOT_ID = LCME_FL_NEG
737 #define LCME_ID_MASK LCME_ID_MAX
739 struct lov_comp_md_entry_v1 {
740 __u32 lcme_id; /* unique id of component */
741 __u32 lcme_flags; /* LCME_FL_XXX */
742 struct lu_extent lcme_extent; /* file extent for component */
743 __u32 lcme_offset; /* offset of component blob,
744 start from lov_comp_md_v1 */
745 __u32 lcme_size; /* size of component blob */
746 __u32 lcme_layout_gen;
747 __u64 lcme_timestamp; /* snapshot time if applicable*/
748 __u32 lcme_padding_1;
749 } __attribute__((packed));
751 #define SEQ_ID_MAX 0x0000FFFF
752 #define SEQ_ID_MASK SEQ_ID_MAX
753 /* bit 30:16 of lcme_id is used to store mirror id */
754 #define MIRROR_ID_MASK 0x7FFF0000
755 #define MIRROR_ID_NEG 0x8000
756 #define MIRROR_ID_SHIFT 16
758 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
760 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
763 static inline __u16 mirror_id_of(__u32 id)
765 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
769 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
771 enum lov_comp_md_flags {
772 /* the least 2 bits are used by FLR to record file state */
775 LCM_FL_WRITE_PENDING = 2,
776 LCM_FL_SYNC_PENDING = 3,
777 LCM_FL_FLR_MASK = 0x3,
780 struct lov_comp_md_v1 {
781 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
782 __u32 lcm_size; /* overall size including this struct */
783 __u32 lcm_layout_gen;
785 __u16 lcm_entry_count;
786 /* lcm_mirror_count stores the number of actual mirrors minus 1,
787 * so that non-flr files will have value 0 meaning 1 mirror. */
788 __u16 lcm_mirror_count;
789 __u16 lcm_padding1[3];
791 struct lov_comp_md_entry_v1 lcm_entries[0];
792 } __attribute__((packed));
794 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
796 if (stripes == (__u16)-1)
799 if (lmm_magic == LOV_USER_MAGIC_V1)
800 return sizeof(struct lov_user_md_v1) +
801 stripes * sizeof(struct lov_user_ost_data_v1);
802 return sizeof(struct lov_user_md_v3) +
803 stripes * sizeof(struct lov_user_ost_data_v1);
806 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
807 * use this. It is unsafe to #define those values in this header as it
808 * is possible the application has already #included <sys/stat.h>. */
809 #ifdef HAVE_LOV_USER_MDS_DATA
810 #define lov_user_mds_data lov_user_mds_data_v1
811 struct lov_user_mds_data_v1 {
812 lstat_t lmd_st; /* MDS stat struct */
813 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
814 } __attribute__((packed));
817 struct lmv_user_mds_data {
818 struct lu_fid lum_fid;
824 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
825 LMV_HASH_TYPE_ALL_CHARS = 1,
826 LMV_HASH_TYPE_FNV_1A_64 = 2,
830 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
832 #define LMV_HASH_NAME_ALL_CHARS "all_char"
833 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
835 /* not real hash type, but exposed to user as "space" hash type */
836 #define LMV_HASH_NAME_SPACE "space"
838 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
839 * and the higher part will be the flag to indicate the status of object,
840 * for example the object is being migrated. And the hash function
841 * might be interpreted differently with different flags. */
842 #define LMV_HASH_TYPE_MASK 0x0000ffff
844 /* once this is set on a plain directory default layout, newly created
845 * subdirectories will be distributed on all MDTs by space usage.
847 #define LMV_HASH_FLAG_SPACE 0x08000000
849 /* The striped directory has ever lost its master LMV EA, then LFSCK
850 * re-generated it. This flag is used to indicate such case. It is an
852 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
854 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
855 #define LMV_HASH_FLAG_MIGRATION 0x80000000
857 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
859 struct lustre_foreign_type {
861 const char *lft_name;
865 * LOV/LMV foreign types
867 enum lustre_foreign_types {
868 LU_FOREIGN_TYPE_NONE = 0,
869 LU_FOREIGN_TYPE_DAOS = 0xda05,
870 /* must be the max/last one */
871 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
874 extern struct lustre_foreign_type lu_foreign_types[];
876 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
877 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
878 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
879 #define lmv_user_md lmv_user_md_v1
880 struct lmv_user_md_v1 {
881 __u32 lum_magic; /* must be the first field */
882 __u32 lum_stripe_count; /* dirstripe count */
883 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
884 __u32 lum_hash_type; /* Dir stripe policy */
885 __u32 lum_type; /* LMV type: default */
889 char lum_pool_name[LOV_MAXPOOLNAME + 1];
890 struct lmv_user_mds_data lum_objects[0];
891 } __attribute__((packed));
893 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
895 if (size <= sizeof(struct lmv_user_md))
898 size -= sizeof(struct lmv_user_md);
899 return (size + sizeof(struct lmv_user_mds_data) - 1) /
900 sizeof(struct lmv_user_mds_data);
904 * NB, historically default layout didn't set type, but use XATTR name to differ
905 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
906 * and still use the same method.
909 LMV_TYPE_DEFAULT = 0x0000,
912 static inline int lmv_user_md_size(int stripes, int lmm_magic)
914 int size = sizeof(struct lmv_user_md);
916 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
917 size += stripes * sizeof(struct lmv_user_mds_data);
922 struct ll_recreate_obj {
928 __u64 id; /* holds object id */
929 __u32 generation; /* holds object generation */
930 __u32 f_type; /* holds object type or stripe idx when passing it to
931 * OST for saving into EA. */
939 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
940 const struct obd_uuid *u2)
942 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
945 static inline int obd_uuid_empty(struct obd_uuid *uuid)
947 return uuid->uuid[0] == '\0';
950 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
952 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
953 uuid->uuid[sizeof(*uuid) - 1] = '\0';
956 /* For printf's only, make sure uuid is terminated */
957 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
962 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
963 /* Obviously not safe, but for printfs, no real harm done...
964 we're always null-terminated, even in a race. */
965 static char temp[sizeof(*uuid)];
966 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
967 temp[sizeof(*uuid) - 1] = '\0';
970 return (char *)(uuid->uuid);
973 #define LUSTRE_MAXFSNAME 8
975 /* Extract fsname from uuid (or target name) of a target
976 e.g. (myfs-OST0007_UUID -> myfs)
977 see also deuuidify. */
978 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
982 strncpy(buf, uuid, buflen - 1);
983 buf[buflen - 1] = '\0';
984 p = strrchr(buf, '-');
989 /* printf display format for Lustre FIDs
990 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
991 #define FID_NOBRACE_LEN 40
992 #define FID_LEN (FID_NOBRACE_LEN + 2)
993 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
994 #define DFID "["DFID_NOBRACE"]"
995 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
997 /* scanf input parse format for fids in DFID_NOBRACE format
998 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
999 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1000 #define SFID "0x%llx:0x%x:0x%x"
1001 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1003 /********* Quotas **********/
1005 #define LUSTRE_QUOTABLOCK_BITS 10
1006 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1008 static inline __u64 lustre_stoqb(size_t space)
1010 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1013 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1014 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1015 #define Q_GETOINFO 0x800102 /* get obd quota info */
1016 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1017 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1019 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1020 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1021 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1022 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1023 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1024 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1025 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1026 /* lustre-specific control commands */
1027 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1028 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1029 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1030 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1032 /* In the current Lustre implementation, the grace time is either the time
1033 * or the timestamp to be used after some quota ID exceeds the soft limt,
1034 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1036 #define LQUOTA_GRACE_BITS 48
1037 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1038 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1039 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1040 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1041 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1043 /* different quota flags */
1045 /* the default quota flag, the corresponding quota ID will use the default
1046 * quota setting, the hardlimit and softlimit of its quota record in the global
1047 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1048 * and high 16 bits will contain this flag (see above comment).
1050 #define LQUOTA_FLAG_DEFAULT 0x0001
1052 #define ALLQUOTA 255 /* set all quota */
1053 static inline char *qtype_name(int qtype)
1066 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1067 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1070 #define N_PERMS_MAX 64
1072 struct perm_downcall_data {
1078 struct identity_downcall_data {
1085 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1086 __u32 idd_groups[0];
1089 struct sepol_downcall_data {
1091 time_t sdd_sepol_mtime;
1092 __u16 sdd_sepol_len;
1096 #ifdef NEED_QUOTA_DEFS
1098 #define QIF_BLIMITS 1
1100 #define QIF_ILIMITS 4
1101 #define QIF_INODES 8
1102 #define QIF_BTIME 16
1103 #define QIF_ITIME 32
1104 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1105 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1106 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1107 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1110 #endif /* !__KERNEL__ */
1112 /* lustre volatile file support
1113 * file name header: ".^L^S^T^R:volatile"
1115 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1116 #define LUSTRE_VOLATILE_HDR_LEN 14
1118 enum lustre_quota_version {
1122 /* XXX: same as if_dqinfo struct in kernel */
1130 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1132 __u64 dqb_bhardlimit;
1133 __u64 dqb_bsoftlimit;
1135 __u64 dqb_ihardlimit;
1136 __u64 dqb_isoftlimit;
1137 __u64 dqb_curinodes;
1151 struct if_quotactl {
1158 struct obd_dqinfo qc_dqinfo;
1159 struct obd_dqblk qc_dqblk;
1161 struct obd_uuid obd_uuid;
1164 /* swap layout flags */
1165 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1166 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1167 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1168 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1169 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1171 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1172 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1173 struct lustre_swap_layouts {
1181 /** Bit-mask of valid attributes */
1182 /* The LA_* flags are written to disk as part of the ChangeLog records
1183 * so they are part of the on-disk and network protocol, and cannot be changed.
1184 * Only the first 12 bits are currently saved.
1187 LA_ATIME = 1 << 0, /* 0x00001 */
1188 LA_MTIME = 1 << 1, /* 0x00002 */
1189 LA_CTIME = 1 << 2, /* 0x00004 */
1190 LA_SIZE = 1 << 3, /* 0x00008 */
1191 LA_MODE = 1 << 4, /* 0x00010 */
1192 LA_UID = 1 << 5, /* 0x00020 */
1193 LA_GID = 1 << 6, /* 0x00040 */
1194 LA_BLOCKS = 1 << 7, /* 0x00080 */
1195 LA_TYPE = 1 << 8, /* 0x00100 */
1196 LA_FLAGS = 1 << 9, /* 0x00200 */
1197 LA_NLINK = 1 << 10, /* 0x00400 */
1198 LA_RDEV = 1 << 11, /* 0x00800 */
1199 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1200 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1201 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1202 LA_PROJID = 1 << 15, /* 0x08000 */
1203 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1204 LA_LSIZE = 1 << 17, /* 0x20000 */
1205 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1207 * Attributes must be transmitted to OST objects
1209 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1212 #define MDS_FMODE_READ 00000001
1213 #define MDS_FMODE_WRITE 00000002
1215 #define MDS_FMODE_CLOSED 00000000
1216 #define MDS_FMODE_EXEC 00000004
1217 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1218 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1219 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1221 #define MDS_OPEN_CREATED 00000010
1222 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1224 #define MDS_OPEN_CREAT 00000100
1225 #define MDS_OPEN_EXCL 00000200
1226 #define MDS_OPEN_TRUNC 00001000
1227 #define MDS_OPEN_APPEND 00002000
1228 #define MDS_OPEN_SYNC 00010000
1229 #define MDS_OPEN_DIRECTORY 00200000
1231 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1232 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1233 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1234 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1235 * We do not support JOIN FILE
1236 * anymore, reserve this flags
1237 * just for preventing such bit
1240 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1241 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1242 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1243 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1244 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1246 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1248 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1249 * delegation, succeed if it's not
1250 * being opened with conflict mode.
1252 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1254 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1255 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1256 * for newly created file */
1258 /* lustre internal open flags, which should not be set from user space */
1259 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1260 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1261 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1262 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1266 /********* Changelogs **********/
1267 /** Changelog record types */
1268 enum changelog_rec_type {
1271 CL_CREATE = 1, /* namespace */
1272 CL_MKDIR = 2, /* namespace */
1273 CL_HARDLINK = 3, /* namespace */
1274 CL_SOFTLINK = 4, /* namespace */
1275 CL_MKNOD = 5, /* namespace */
1276 CL_UNLINK = 6, /* namespace */
1277 CL_RMDIR = 7, /* namespace */
1278 CL_RENAME = 8, /* namespace */
1279 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1280 CL_OPEN = 10, /* not currently used */
1281 CL_CLOSE = 11, /* may be written to log only with mtime change */
1282 CL_LAYOUT = 12, /* file layout/striping modified */
1286 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1287 CL_HSM = 16, /* HSM specific events, see flags */
1288 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1292 CL_FLRW = 21, /* FLR: file was firstly written */
1293 CL_RESYNC = 22, /* FLR: file was resync-ed */
1295 CL_DN_OPEN = 24, /* denied open */
1299 static inline const char *changelog_type2str(int type) {
1300 static const char *changelog_str[] = {
1301 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1302 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1303 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1304 "FLRW", "RESYNC","GXATR", "NOPEN",
1307 if (type >= 0 && type < CL_LAST)
1308 return changelog_str[type];
1312 /* 12 bits of per-record data can be stored in the bottom of the flags */
1313 #define CLF_FLAGSHIFT 12
1314 enum changelog_rec_flags {
1315 CLF_VERSION = 0x1000,
1316 CLF_RENAME = 0x2000,
1318 CLF_EXTRA_FLAGS = 0x8000,
1319 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1321 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1322 CLF_VERMASK = ~CLF_FLAGMASK,
1326 /* Anything under the flagmask may be per-type (if desired) */
1327 /* Flags for unlink */
1328 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1329 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1330 /* HSM cleaning needed */
1331 /* Flags for rename */
1332 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1334 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1335 * has an archive in backend */
1338 /* 12b used (from high weight to low weight):
1343 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1344 #define CLF_HSM_ERR_H 6
1345 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1346 #define CLF_HSM_EVENT_H 9
1347 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1348 #define CLF_HSM_FLAG_H 11
1349 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1350 #define CLF_HSM_SPARE_H 15
1351 #define CLF_HSM_LAST 15
1353 /* Remove bits higher than _h, then extract the value
1354 * between _h and _l by shifting lower weigth to bit 0. */
1355 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1356 >> (CLF_HSM_LAST - _h + _l))
1358 #define CLF_HSM_SUCCESS 0x00
1359 #define CLF_HSM_MAXERROR 0x7E
1360 #define CLF_HSM_ERROVERFLOW 0x7F
1362 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1364 /* 3 bits field => 8 values allowed */
1376 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1378 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1382 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1385 *clf_flags |= (he << CLF_HSM_EVENT_L);
1388 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1390 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1393 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1396 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1399 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1401 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1404 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1407 *clf_flags |= (error << CLF_HSM_ERR_L);
1410 enum changelog_rec_extra_flags {
1412 CLFE_UIDGID = 0x0001,
1415 CLFE_XATTR = 0x0008,
1416 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1419 enum changelog_send_flag {
1420 /* Not yet implemented */
1421 CHANGELOG_FLAG_FOLLOW = 0x01,
1422 /* Blocking IO makes sense in case of slow user parsing of the records,
1423 * but it also prevents us from cleaning up if the records are not
1425 CHANGELOG_FLAG_BLOCK = 0x02,
1426 /* Pack jobid into the changelog records if available. */
1427 CHANGELOG_FLAG_JOBID = 0x04,
1428 /* Pack additional flag bits into the changelog record */
1429 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1432 enum changelog_send_extra_flag {
1433 /* Pack uid/gid into the changelog record */
1434 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1435 /* Pack nid into the changelog record */
1436 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1437 /* Pack open mode into the changelog record */
1438 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1439 /* Pack xattr name into the changelog record */
1440 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1443 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1444 changelog_rec_offset(CLF_SUPPORTED, \
1447 /* 31 usable bytes string + null terminator. */
1448 #define LUSTRE_JOBID_SIZE 32
1450 /* This is the minimal changelog record. It can contain extensions
1451 * such as rename fields or process jobid. Its exact content is described
1452 * by the cr_flags and cr_extra_flags.
1454 * Extensions are packed in the same order as their corresponding flags,
1455 * then in the same order as their corresponding extra flags.
1457 struct changelog_rec {
1459 __u16 cr_flags; /**< \a changelog_rec_flags */
1460 __u32 cr_type; /**< \a changelog_rec_type */
1461 __u64 cr_index; /**< changelog record number */
1462 __u64 cr_prev; /**< last index for this target fid */
1465 struct lu_fid cr_tfid; /**< target fid */
1466 __u32 cr_markerflags; /**< CL_MARK flags */
1468 struct lu_fid cr_pfid; /**< parent fid */
1471 /* Changelog extension for RENAME. */
1472 struct changelog_ext_rename {
1473 struct lu_fid cr_sfid; /**< source fid, or zero */
1474 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1477 /* Changelog extension to include JOBID. */
1478 struct changelog_ext_jobid {
1479 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1482 /* Changelog extension to include additional flags. */
1483 struct changelog_ext_extra_flags {
1484 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1487 /* Changelog extra extension to include UID/GID. */
1488 struct changelog_ext_uidgid {
1493 /* Changelog extra extension to include NID. */
1494 struct changelog_ext_nid {
1495 /* have __u64 instead of lnet_nid_t type for use by client api */
1497 /* for use when IPv6 support is added */
1502 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1503 struct changelog_ext_openmode {
1507 /* Changelog extra extension to include xattr */
1508 struct changelog_ext_xattr {
1509 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1512 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1513 const struct changelog_rec *rec);
1515 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1516 enum changelog_rec_extra_flags cref)
1518 size_t size = sizeof(struct changelog_rec);
1520 if (crf & CLF_RENAME)
1521 size += sizeof(struct changelog_ext_rename);
1523 if (crf & CLF_JOBID)
1524 size += sizeof(struct changelog_ext_jobid);
1526 if (crf & CLF_EXTRA_FLAGS) {
1527 size += sizeof(struct changelog_ext_extra_flags);
1528 if (cref & CLFE_UIDGID)
1529 size += sizeof(struct changelog_ext_uidgid);
1530 if (cref & CLFE_NID)
1531 size += sizeof(struct changelog_ext_nid);
1532 if (cref & CLFE_OPEN)
1533 size += sizeof(struct changelog_ext_openmode);
1534 if (cref & CLFE_XATTR)
1535 size += sizeof(struct changelog_ext_xattr);
1541 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1543 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1545 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1546 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1548 return changelog_rec_offset(rec->cr_flags, cref);
1551 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1553 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1557 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1559 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1561 return (struct changelog_ext_rename *)((char *)rec +
1562 changelog_rec_offset(crf,
1566 /* The jobid follows the rename extension, if present */
1568 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1570 enum changelog_rec_flags crf = rec->cr_flags &
1571 (CLF_VERSION | CLF_RENAME);
1573 return (struct changelog_ext_jobid *)((char *)rec +
1574 changelog_rec_offset(crf,
1578 /* The additional flags follow the rename and jobid extensions, if present */
1580 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1581 const struct changelog_rec *rec)
1583 enum changelog_rec_flags crf = rec->cr_flags &
1584 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1586 return (struct changelog_ext_extra_flags *)((char *)rec +
1587 changelog_rec_offset(crf,
1591 /* The uid/gid is the first extra extension */
1593 struct changelog_ext_uidgid *changelog_rec_uidgid(
1594 const struct changelog_rec *rec)
1596 enum changelog_rec_flags crf = rec->cr_flags &
1597 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1599 return (struct changelog_ext_uidgid *)((char *)rec +
1600 changelog_rec_offset(crf,
1604 /* The nid is the second extra extension */
1606 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1608 enum changelog_rec_flags crf = rec->cr_flags &
1609 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1610 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1612 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1613 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1616 return (struct changelog_ext_nid *)((char *)rec +
1617 changelog_rec_offset(crf, cref));
1620 /* The OPEN mode is the third extra extension */
1622 struct changelog_ext_openmode *changelog_rec_openmode(
1623 const struct changelog_rec *rec)
1625 enum changelog_rec_flags crf = rec->cr_flags &
1626 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1627 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1629 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1630 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1631 (CLFE_UIDGID | CLFE_NID);
1633 return (struct changelog_ext_openmode *)((char *)rec +
1634 changelog_rec_offset(crf, cref));
1637 /* The xattr name is the fourth extra extension */
1639 struct changelog_ext_xattr *changelog_rec_xattr(
1640 const struct changelog_rec *rec)
1642 enum changelog_rec_flags crf = rec->cr_flags &
1643 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1644 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1646 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1647 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1648 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1650 return (struct changelog_ext_xattr *)((char *)rec +
1651 changelog_rec_offset(crf, cref));
1654 /* The name follows the rename, jobid and extra flags extns, if present */
1655 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1657 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1659 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1660 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1662 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1663 cref & CLFE_SUPPORTED);
1666 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1668 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1671 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1673 char *cr_name = changelog_rec_name(rec);
1675 return cr_name + strlen(cr_name) + 1;
1679 * Remap a record to the desired format as specified by the crf flags.
1680 * The record must be big enough to contain the final remapped version.
1681 * Superfluous extension fields are removed and missing ones are added
1682 * and zeroed. The flags of the record are updated accordingly.
1684 * The jobid and rename extensions can be added to a record, to match the
1685 * format an application expects, typically. In this case, the newly added
1686 * fields will be zeroed.
1687 * The Jobid field can be removed, to guarantee compatibility with older
1688 * clients that don't expect this field in the records they process.
1690 * The following assumptions are being made:
1691 * - CLF_RENAME will not be removed
1692 * - CLF_JOBID will not be added without CLF_RENAME being added too
1693 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1695 * @param[in,out] rec The record to remap.
1696 * @param[in] crf_wanted Flags describing the desired extensions.
1697 * @param[in] cref_want Flags describing the desired extra extensions.
1699 static inline void changelog_remap_rec(struct changelog_rec *rec,
1700 enum changelog_rec_flags crf_wanted,
1701 enum changelog_rec_extra_flags cref_want)
1703 char *xattr_mov = NULL;
1704 char *omd_mov = NULL;
1705 char *nid_mov = NULL;
1706 char *uidgid_mov = NULL;
1710 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1712 crf_wanted &= CLF_SUPPORTED;
1713 cref_want &= CLFE_SUPPORTED;
1715 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1716 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1717 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1718 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1724 /* First move the variable-length name field */
1725 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1726 changelog_rec_name(rec), rec->cr_namelen);
1728 /* Locations of extensions in the remapped record */
1729 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1730 xattr_mov = (char *)rec +
1731 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1732 cref_want & ~CLFE_XATTR);
1733 omd_mov = (char *)rec +
1734 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1735 cref_want & ~(CLFE_OPEN |
1737 nid_mov = (char *)rec +
1738 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1739 cref_want & ~(CLFE_NID |
1742 uidgid_mov = (char *)rec +
1743 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1744 cref_want & ~(CLFE_UIDGID |
1748 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1751 ef_mov = (char *)rec +
1752 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1754 jid_mov = (char *)rec +
1755 changelog_rec_offset(crf_wanted &
1756 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1758 rnm_mov = (char *)rec +
1759 changelog_rec_offset(crf_wanted &
1765 /* Move the extension fields to the desired positions */
1766 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1767 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1768 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1769 memmove(xattr_mov, changelog_rec_xattr(rec),
1770 sizeof(struct changelog_ext_xattr));
1772 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1773 memmove(omd_mov, changelog_rec_openmode(rec),
1774 sizeof(struct changelog_ext_openmode));
1776 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1777 memmove(nid_mov, changelog_rec_nid(rec),
1778 sizeof(struct changelog_ext_nid));
1780 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1781 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1782 sizeof(struct changelog_ext_uidgid));
1784 memmove(ef_mov, changelog_rec_extra_flags(rec),
1785 sizeof(struct changelog_ext_extra_flags));
1788 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1789 memmove(jid_mov, changelog_rec_jobid(rec),
1790 sizeof(struct changelog_ext_jobid));
1792 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1793 memmove(rnm_mov, changelog_rec_rename(rec),
1794 sizeof(struct changelog_ext_rename));
1796 /* Clear newly added fields */
1797 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1798 !(cref & CLFE_XATTR))
1799 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1801 if (omd_mov && (cref_want & CLFE_OPEN) &&
1802 !(cref & CLFE_OPEN))
1803 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1805 if (nid_mov && (cref_want & CLFE_NID) &&
1807 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1809 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1810 !(cref & CLFE_UIDGID))
1811 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1813 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1814 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1815 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1817 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1818 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1820 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1821 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1823 /* Update the record's flags accordingly */
1824 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1825 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1826 changelog_rec_extra_flags(rec)->cr_extra_flags =
1827 changelog_rec_extra_flags(rec)->cr_extra_flags |
1831 enum changelog_message_type {
1832 CL_RECORD = 10, /* message is a changelog_rec */
1833 CL_EOF = 11, /* at end of current changelog */
1836 /********* Misc **********/
1838 struct ioc_data_version {
1840 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1841 __u32 idv_flags; /* enum ioc_data_version_flags */
1844 enum ioc_data_version_flags {
1845 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1846 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1850 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1853 #define dot_lustre_name ".lustre"
1856 /********* HSM **********/
1858 /** HSM per-file state
1859 * See HSM_FLAGS below.
1862 HS_NONE = 0x00000000,
1863 HS_EXISTS = 0x00000001,
1864 HS_DIRTY = 0x00000002,
1865 HS_RELEASED = 0x00000004,
1866 HS_ARCHIVED = 0x00000008,
1867 HS_NORELEASE = 0x00000010,
1868 HS_NOARCHIVE = 0x00000020,
1869 HS_LOST = 0x00000040,
1872 /* HSM user-setable flags. */
1873 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1875 /* Other HSM flags. */
1876 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1879 * All HSM-related possible flags that could be applied to a file.
1880 * This should be kept in sync with hsm_states.
1882 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1885 * HSM request progress state
1887 enum hsm_progress_states {
1894 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1897 case HPS_WAITING: return "waiting";
1898 case HPS_RUNNING: return "running";
1899 case HPS_DONE: return "done";
1900 default: return "unknown";
1907 } __attribute__((packed));
1910 * Current HSM states of a Lustre file.
1912 * This structure purpose is to be sent to user-space mainly. It describes the
1913 * current HSM flags and in-progress action.
1915 struct hsm_user_state {
1916 /** Current HSM states, from enum hsm_states. */
1918 __u32 hus_archive_id;
1919 /** The current undergoing action, if there is one */
1920 __u32 hus_in_progress_state;
1921 __u32 hus_in_progress_action;
1922 struct hsm_extent hus_in_progress_location;
1923 char hus_extended_info[];
1926 struct hsm_state_set_ioc {
1927 struct lu_fid hssi_fid;
1929 __u64 hssi_clearmask;
1933 * This structure describes the current in-progress action for a file.
1934 * it is retuned to user space and send over the wire
1936 struct hsm_current_action {
1937 /** The current undergoing action, if there is one */
1938 /* state is one of hsm_progress_states */
1940 /* action is one of hsm_user_action */
1942 struct hsm_extent hca_location;
1945 /***** HSM user requests ******/
1946 /* User-generated (lfs/ioctl) request types */
1947 enum hsm_user_action {
1948 HUA_NONE = 1, /* no action (noop) */
1949 HUA_ARCHIVE = 10, /* copy to hsm */
1950 HUA_RESTORE = 11, /* prestage */
1951 HUA_RELEASE = 12, /* drop ost objects */
1952 HUA_REMOVE = 13, /* remove from archive */
1953 HUA_CANCEL = 14 /* cancel a request */
1956 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1959 case HUA_NONE: return "NOOP";
1960 case HUA_ARCHIVE: return "ARCHIVE";
1961 case HUA_RESTORE: return "RESTORE";
1962 case HUA_RELEASE: return "RELEASE";
1963 case HUA_REMOVE: return "REMOVE";
1964 case HUA_CANCEL: return "CANCEL";
1965 default: return "UNKNOWN";
1970 * List of hr_flags (bit field)
1972 #define HSM_FORCE_ACTION 0x0001
1973 /* used by CT, cannot be set by user */
1974 #define HSM_GHOST_COPY 0x0002
1977 * Contains all the fixed part of struct hsm_user_request.
1980 struct hsm_request {
1981 __u32 hr_action; /* enum hsm_user_action */
1982 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1983 __u64 hr_flags; /* request flags */
1984 __u32 hr_itemcount; /* item count in hur_user_item vector */
1988 struct hsm_user_item {
1989 struct lu_fid hui_fid;
1990 struct hsm_extent hui_extent;
1991 } __attribute__((packed));
1993 struct hsm_user_request {
1994 struct hsm_request hur_request;
1995 struct hsm_user_item hur_user_item[0];
1996 /* extra data blob at end of struct (after all
1997 * hur_user_items), only use helpers to access it
1999 } __attribute__((packed));
2001 /** Return pointer to data field in a hsm user request */
2002 static inline void *hur_data(struct hsm_user_request *hur)
2004 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2008 * Compute the current length of the provided hsm_user_request. This returns -1
2009 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
2011 * return -1 on bounds check error.
2013 static inline ssize_t hur_len(struct hsm_user_request *hur)
2017 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2018 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2019 (__u64)hur->hur_request.hr_itemcount *
2020 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2022 if (size != (ssize_t)size)
2028 /****** HSM RPCs to copytool *****/
2029 /* Message types the copytool may receive */
2030 enum hsm_message_type {
2031 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2034 /* Actions the copytool may be instructed to take for a given action_item */
2035 enum hsm_copytool_action {
2036 HSMA_NONE = 10, /* no action */
2037 HSMA_ARCHIVE = 20, /* arbitrary offset */
2043 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2046 case HSMA_NONE: return "NOOP";
2047 case HSMA_ARCHIVE: return "ARCHIVE";
2048 case HSMA_RESTORE: return "RESTORE";
2049 case HSMA_REMOVE: return "REMOVE";
2050 case HSMA_CANCEL: return "CANCEL";
2051 default: return "UNKNOWN";
2055 /* Copytool item action description */
2056 struct hsm_action_item {
2057 __u32 hai_len; /* valid size of this struct */
2058 __u32 hai_action; /* hsm_copytool_action, but use known size */
2059 struct lu_fid hai_fid; /* Lustre FID to operate on */
2060 struct lu_fid hai_dfid; /* fid used for data access */
2061 struct hsm_extent hai_extent; /* byte range to operate on */
2062 __u64 hai_cookie; /* action cookie from coordinator */
2063 __u64 hai_gid; /* grouplock id */
2064 char hai_data[0]; /* variable length */
2065 } __attribute__((packed));
2068 * helper function which print in hexa the first bytes of
2071 * \param hai [IN] record to print
2072 * \param buffer [IN,OUT] buffer to write the hex string to
2073 * \param len [IN] max buffer length
2077 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2078 char *buffer, size_t len)
2085 data_len = hai->hai_len - sizeof(*hai);
2086 for (i = 0; (i < data_len) && (len > 2); i++) {
2087 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2097 /* Copytool action list */
2098 #define HAL_VERSION 1
2099 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2100 struct hsm_action_list {
2102 __u32 hal_count; /* number of hai's to follow */
2103 __u64 hal_compound_id; /* returned by coordinator, ignored */
2105 __u32 hal_archive_id; /* which archive backend */
2107 char hal_fsname[0]; /* null-terminated */
2108 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2109 boundaries. See hai_zero */
2110 } __attribute__((packed));
2112 /* Return pointer to first hai in action list */
2113 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2115 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2117 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2120 /* Return pointer to next hai */
2121 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2123 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2125 return (struct hsm_action_item *)((char *)hai + offset);
2128 /* Return size of an hsm_action_list */
2129 static inline size_t hal_size(struct hsm_action_list *hal)
2133 struct hsm_action_item *hai;
2135 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2136 hai = hai_first(hal);
2137 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2138 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2144 * describe the attributes to be set on imported file
2146 struct hsm_user_import {
2155 __u32 hui_archive_id;
2158 /* Copytool progress reporting */
2159 #define HP_FLAG_COMPLETED 0x01
2160 #define HP_FLAG_RETRY 0x02
2162 struct hsm_progress {
2163 struct lu_fid hp_fid;
2165 struct hsm_extent hp_extent;
2167 __u16 hp_errval; /* positive val */
2172 __u64 hc_data_version;
2174 __u16 hc_errval; /* positive val */
2176 struct hsm_action_item hc_hai;
2180 enum llapi_json_types {
2181 LLAPI_JSON_INTEGER = 1,
2187 struct llapi_json_item {
2196 struct llapi_json_item *lji_next;
2199 struct llapi_json_item_list {
2200 int ljil_item_count;
2201 struct llapi_json_item *ljil_items;
2204 enum lu_ladvise_type {
2205 LU_LADVISE_INVALID = 0,
2206 LU_LADVISE_WILLREAD = 1,
2207 LU_LADVISE_DONTNEED = 2,
2208 LU_LADVISE_LOCKNOEXPAND = 3,
2209 LU_LADVISE_LOCKAHEAD = 4,
2213 #define LU_LADVISE_NAMES { \
2214 [LU_LADVISE_WILLREAD] = "willread", \
2215 [LU_LADVISE_DONTNEED] = "dontneed", \
2216 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2217 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2220 /* This is the userspace argument for ladvise. It is currently the same as
2221 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2222 * may need info which is only used locally. */
2223 struct llapi_lu_ladvise {
2224 __u16 lla_advice; /* advice type */
2225 __u16 lla_value1; /* values for different advice types */
2227 __u64 lla_start; /* first byte of extent for advice */
2228 __u64 lla_end; /* last byte of extent for advice */
2234 LF_ASYNC = 0x00000001,
2235 LF_UNSET = 0x00000002,
2238 #define LADVISE_MAGIC 0x1ADF1CE0
2239 /* Masks of valid flags for each advice */
2240 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2241 /* Flags valid for all advices not explicitly specified */
2242 #define LF_DEFAULT_MASK LF_ASYNC
2244 #define LF_MASK (LF_ASYNC | LF_UNSET)
2246 #define lla_lockahead_mode lla_value1
2247 #define lla_peradvice_flags lla_value2
2248 #define lla_lockahead_result lla_value3
2250 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2251 * is used on the wire. It is defined separately as we may need info which is
2252 * only used locally. */
2253 struct llapi_ladvise_hdr {
2254 __u32 lah_magic; /* LADVISE_MAGIC */
2255 __u32 lah_count; /* number of advices */
2256 __u64 lah_flags; /* from enum ladvise_flag */
2257 __u32 lah_value1; /* unused */
2258 __u32 lah_value2; /* unused */
2259 __u64 lah_value3; /* unused */
2260 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2263 #define LAH_COUNT_MAX (1024)
2267 SK_CRYPT_INVALID = -1,
2269 SK_CRYPT_AES256_CTR = 1,
2273 SK_HMAC_INVALID = -1,
2279 struct sk_crypt_type {
2280 const char *sct_name;
2284 struct sk_hmac_type {
2285 const char *sht_name;
2289 enum lock_mode_user {
2295 #define LOCK_MODE_NAMES { \
2296 [MODE_READ_USER] = "READ",\
2297 [MODE_WRITE_USER] = "WRITE"\
2300 enum lockahead_results {
2301 LLA_RESULT_SENT = 0,
2302 LLA_RESULT_DIFFERENT,
2306 enum lu_heat_flag_bit {
2307 LU_HEAT_FLAG_BIT_INVALID = 0,
2308 LU_HEAT_FLAG_BIT_OFF,
2309 LU_HEAT_FLAG_BIT_CLEAR,
2313 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2314 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2317 enum obd_heat_type {
2318 OBD_HEAT_READSAMPLE = 0,
2319 OBD_HEAT_WRITESAMPLE = 1,
2320 OBD_HEAT_READBYTE = 2,
2321 OBD_HEAT_WRITEBYTE = 3,
2325 #define LU_HEAT_NAMES { \
2326 [OBD_HEAT_READSAMPLE] = "readsample", \
2327 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2328 [OBD_HEAT_READBYTE] = "readbyte", \
2329 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2344 static inline const char *pcc_type2string(enum lu_pcc_type type)
2349 case LU_PCC_READWRITE:
2356 struct lu_pcc_attach {
2357 __u32 pcca_type; /* PCC type */
2358 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2361 enum lu_pcc_detach_opts {
2362 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2363 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2366 struct lu_pcc_detach_fid {
2367 /* fid of the file to detach */
2368 struct lu_fid pccd_fid;
2372 struct lu_pcc_detach {
2376 enum lu_pcc_state_flags {
2377 PCC_STATE_FL_NONE = 0x0,
2378 /* The inode attr is cached locally */
2379 PCC_STATE_FL_ATTR_VALID = 0x01,
2380 /* The file is being attached into PCC */
2381 PCC_STATE_FL_ATTACHING = 0x02,
2382 /* Allow to auto attach at open */
2383 PCC_STATE_FL_OPEN_ATTACH = 0x04,
2386 struct lu_pcc_state {
2387 __u32 pccs_type; /* enum lu_pcc_type */
2388 __u32 pccs_open_count;
2389 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2391 char pccs_path[PATH_MAX];
2394 #if defined(__cplusplus)
2398 /** @} lustreuser */
2400 #endif /* _LUSTRE_USER_H */