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() */
164 /** additional filesystem attributes for target device */
165 struct obd_statfs_info {
166 __u32 os_reserved_mb_low; /* reserved mb low */
167 __u32 os_reserved_mb_high; /* reserved mb high */
168 bool os_enable_pre; /* enable pre create logic */
174 * FID is a cluster-wide unique identifier of a file or an object (stripe).
175 * FIDs are never reused.
179 * FID sequence. Sequence is a unit of migration: all files (objects)
180 * with FIDs from a given sequence are stored on the same server.
181 * Lustre should support 2^64 objects, so even if each sequence
182 * has only a single object we can still enumerate 2^64 objects.
185 /* FID number within sequence. */
188 * FID version, used to distinguish different versions (in the sense
189 * of snapshots, etc.) of the same file system object. Not currently
195 static inline bool fid_is_zero(const struct lu_fid *fid)
197 return fid->f_seq == 0 && fid->f_oid == 0;
200 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
201 * MDT-object's FID::f_ver, instead it is the OST-object index in its
202 * parent MDT-object's layout EA. */
203 #define f_stripe_idx f_ver
206 __u32 ol_stripe_size;
207 __u32 ol_stripe_count;
211 } __attribute__((packed));
213 /* The filter_fid structure has changed several times over its lifetime.
214 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
215 * stripe_index and the "self FID" (objid/seq) to be able to recover the
216 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
217 * the OST, the "trusted.lma" xattr was added to the OST objects to store
218 * the "self FID" to be consistent with the MDT on-disk format, and the
219 * filter_fid only stored the MDT inode parent FID and stripe index.
221 * In 2.10, the addition of PFL composite layouts required more information
222 * to be stored into the filter_fid in order to be able to identify which
223 * component the OST object belonged. As well, the stripe size may vary
224 * between components, so it was no longer safe to assume the stripe size
225 * or stripe_count of a file. This is also more robust for plain layouts.
227 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
228 * enough space to store both the filter_fid and LMA in the inode, so they
229 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
230 * an extra seek for every OST object access.
232 * In 2.11, FLR mirror layouts also need to store the layout version and
233 * range so that writes to old versions of the layout are not allowed.
234 * That ensures that mirrored objects are not modified by evicted clients,
235 * and ensures that the components are correctly marked stale on the MDT.
237 struct filter_fid_18_23 {
238 struct lu_fid ff_parent; /* stripe_idx in f_ver */
243 struct filter_fid_24_29 {
244 struct lu_fid ff_parent; /* stripe_idx in f_ver */
247 struct filter_fid_210 {
248 struct lu_fid ff_parent; /* stripe_idx in f_ver */
249 struct ost_layout ff_layout;
253 struct lu_fid ff_parent; /* stripe_idx in f_ver */
254 struct ost_layout ff_layout;
255 __u32 ff_layout_version;
256 __u32 ff_range; /* range of layout version that
257 * write are allowed */
258 } __attribute__((packed));
260 /* Userspace should treat lu_fid as opaque, and only use the following methods
261 * to print or parse them. Other functions (e.g. compare, swab) could be moved
262 * here from lustre_idl.h if needed. */
266 LMAC_HSM = 0x00000001,
267 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
268 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
269 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
270 * under /O/<seq>/d<x>. */
271 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
272 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
273 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
277 * Masks for all features that should be supported by a Lustre version to
278 * access a specific file.
279 * This information is stored in lustre_mdt_attrs::lma_incompat.
282 LMAI_RELEASED = 0x00000001, /* file is released */
283 LMAI_AGENT = 0x00000002, /* agent inode */
284 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
285 is on the remote MDT */
286 LMAI_STRIPED = 0x00000008, /* striped directory inode */
287 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
288 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
289 LMAI_STRIPED | LMAI_ORPHAN)
294 * Following struct for object attributes, that will be kept inode's EA.
295 * Introduced in 2.0 release (please see b15993, for details)
296 * Added to all objects since Lustre 2.4 as contains self FID
298 struct lustre_mdt_attrs {
300 * Bitfield for supported data in this structure. From enum lma_compat.
301 * lma_self_fid and lma_flags are always available.
305 * Per-file incompat feature list. Lustre version should support all
306 * flags set in this field. The supported feature mask is available in
310 /** FID of this inode */
311 struct lu_fid lma_self_fid;
314 struct lustre_ost_attrs {
315 /* Use lustre_mdt_attrs directly for now, need a common header
316 * structure if want to change lustre_mdt_attrs in future. */
317 struct lustre_mdt_attrs loa_lma;
319 /* Below five elements are for OST-object's PFID EA, the
320 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
321 * and the stripe_index (low 16 bits), the size should not exceed
322 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
323 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
324 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
325 * loa_comp_* elements are valid. */
326 struct lu_fid loa_parent_fid;
327 __u32 loa_stripe_size;
329 __u64 loa_comp_start;
334 * Prior to 2.4, the LMA structure also included SOM attributes which has since
335 * been moved to a dedicated xattr
336 * lma_flags was also removed because of lma_compat/incompat fields.
338 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
340 enum lustre_som_flags {
341 /* Unknow or no SoM data, must get size from OSTs. */
342 SOM_FL_UNKNOWN = 0x0000,
343 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
344 SOM_FL_STRICT = 0x0001,
345 /* Known stale - was right at some point in the past, but it is
346 * known (or likely) to be incorrect now (e.g. opened for write). */
347 SOM_FL_STALE = 0x0002,
348 /* Approximate, may never have been strictly correct,
349 * need to sync SOM data to achieve eventual consistency. */
350 SOM_FL_LAZY = 0x0004,
353 struct lustre_som_attrs {
355 __u16 lsa_reserved[3];
361 * OST object IDentifier.
369 struct lu_fid oi_fid;
373 #define DOSTID "%#llx:%llu"
374 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
375 ((unsigned long long)ostid_id(oi))
377 struct ll_futimes_3 {
379 __u64 lfu_atime_nsec;
381 __u64 lfu_mtime_nsec;
383 __u64 lfu_ctime_nsec;
387 * Maximum number of mirrors currently implemented.
389 #define LUSTRE_MIRROR_COUNT_MAX 16
391 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
393 LL_LEASE_RDLCK = 0x01,
394 LL_LEASE_WRLCK = 0x02,
395 LL_LEASE_UNLCK = 0x04,
398 enum ll_lease_flags {
399 LL_LEASE_RESYNC = 0x1,
400 LL_LEASE_RESYNC_DONE = 0x2,
401 LL_LEASE_LAYOUT_MERGE = 0x4,
402 LL_LEASE_LAYOUT_SPLIT = 0x8,
403 LL_LEASE_PCC_ATTACH = 0x10,
406 #define IOC_IDS_MAX 4096
407 struct ll_ioc_lease {
414 struct ll_ioc_lease_id {
425 * The ioctl naming rules:
426 * LL_* - works on the currently opened filehandle instead of parent dir
427 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
428 * *_MDC_* - gets/sets data related to MDC
429 * *_LOV_* - gets/sets data related to OSC/LOV
430 * *FILE* - called on parent dir and passes in a filename
431 * *STRIPE* - set/get lov_user_md
432 * *INFO - set/get lov_user_mds_data
434 /* lustre_ioctl.h 101-150 */
435 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
436 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
437 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
438 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
439 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
440 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
441 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
442 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
443 /* LL_IOC_RECREATE_OBJ 157 obsolete */
444 /* LL_IOC_RECREATE_FID 157 obsolete */
445 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
446 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
447 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
448 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
449 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
450 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
451 /* IOC_LOV_GETINFO 165 obsolete */
452 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
453 /* LL_IOC_RMTACL 167 obsolete */
454 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
455 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
456 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
457 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
458 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
459 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
460 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
461 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
462 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
463 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
464 /* lustre_ioctl.h 177-210 */
465 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
466 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
467 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
468 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
469 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
470 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
471 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
472 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
473 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
474 struct lustre_swap_layouts)
475 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
476 struct hsm_current_action)
477 /* lustre_ioctl.h 221-232 */
478 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
479 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
480 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
481 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
482 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
483 #define LL_IOC_GET_LEASE _IO('f', 244)
484 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
485 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
486 #define LL_IOC_MIGRATE _IOR('f', 247, int)
487 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
488 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
489 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
490 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
491 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
492 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
493 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
494 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
496 #ifndef FS_IOC_FSGETXATTR
498 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
501 __u32 fsx_xflags; /* xflags field value (get/set) */
502 __u32 fsx_extsize; /* extsize field value (get/set)*/
503 __u32 fsx_nextents; /* nextents field value (get) */
504 __u32 fsx_projid; /* project identifier (get/set) */
505 unsigned char fsx_pad[12];
507 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
508 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
510 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
511 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
512 #ifndef FS_XFLAG_PROJINHERIT
513 #define FS_XFLAG_PROJINHERIT 0x00000200
517 #define LL_STATFS_LMV 1
518 #define LL_STATFS_LOV 2
519 #define LL_STATFS_NODELAY 4
521 #define IOC_MDC_TYPE 'i'
522 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
523 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
524 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
525 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
527 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
529 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
530 * files, but are unlikely to be used in practice and are not harmful if
531 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
532 * devices and are safe for use on new files. See LU-4209. */
533 /* To be compatible with old statically linked binary we keep the check for
534 * the older 0100000000 flag. This is already removed upstream. LU-812. */
535 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
537 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
539 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
540 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
541 O_LOV_DELAY_CREATE_MASK)
543 #define LL_FILE_IGNORE_LOCK 0x00000001
544 #define LL_FILE_GROUP_LOCKED 0x00000002
545 #define LL_FILE_READAHEA 0x00000004
546 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
547 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
548 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
550 #define LOV_USER_MAGIC_V1 0x0BD10BD0
551 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
552 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
553 #define LOV_USER_MAGIC_V3 0x0BD30BD0
554 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
555 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
556 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
557 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
559 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
560 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
561 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
563 #define LOV_PATTERN_NONE 0x000
564 #define LOV_PATTERN_RAID0 0x001
565 #define LOV_PATTERN_RAID1 0x002
566 #define LOV_PATTERN_MDT 0x100
567 #define LOV_PATTERN_OVERSTRIPING 0x200
569 #define LOV_PATTERN_F_MASK 0xffff0000
570 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
571 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
572 #define LOV_PATTERN_DEFAULT 0xffffffff
574 #define LOV_OFFSET_DEFAULT ((__u16)-1)
576 static inline bool lov_pattern_supported(__u32 pattern)
578 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
579 (pattern & ~LOV_PATTERN_F_RELEASED) ==
580 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
581 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
584 /* RELEASED and MDT patterns are not valid in many places, so rather than
585 * having many extra checks on lov_pattern_supported, we have this separate
586 * check for non-released, non-DOM components
588 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
590 return pattern == LOV_PATTERN_RAID0 ||
591 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
595 #define LOV_MAXPOOLNAME 15
596 #define LOV_POOLNAMEF "%.15s"
598 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
599 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
600 #define LOV_MAX_STRIPE_COUNT_OLD 160
601 /* This calculation is crafted so that input of 4096 will result in 160
602 * which in turn is equal to old maximal stripe count.
603 * XXX: In fact this is too simpified for now, what it also need is to get
604 * ea_type argument to clearly know how much space each stripe consumes.
606 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
607 * allocation that is sufficient for the current generation of systems.
609 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
610 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
611 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
612 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
614 #define XATTR_LUSTRE_PREFIX "lustre."
615 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
617 #define lov_user_ost_data lov_user_ost_data_v1
618 struct lov_user_ost_data_v1 { /* per-stripe data structure */
619 struct ost_id l_ost_oi; /* OST object ID */
620 __u32 l_ost_gen; /* generation of this OST index */
621 __u32 l_ost_idx; /* OST index in LOV */
622 } __attribute__((packed));
624 #define lov_user_md lov_user_md_v1
625 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
626 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
627 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
628 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
629 __u32 lmm_stripe_size; /* size of stripe in bytes */
630 __u16 lmm_stripe_count; /* num stripes in use for this object */
632 __u16 lmm_stripe_offset; /* starting stripe offset in
633 * lmm_objects, use when writing */
634 __u16 lmm_layout_gen; /* layout generation number
635 * used when reading */
637 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
638 } __attribute__((packed, __may_alias__));
640 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
641 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
642 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
643 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
644 __u32 lmm_stripe_size; /* size of stripe in bytes */
645 __u16 lmm_stripe_count; /* num stripes in use for this object */
647 __u16 lmm_stripe_offset; /* starting stripe offset in
648 * lmm_objects, use when writing */
649 __u16 lmm_layout_gen; /* layout generation number
650 * used when reading */
652 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
653 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
654 } __attribute__((packed));
656 struct lov_foreign_md {
657 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
658 __u32 lfm_length; /* length of lfm_value */
659 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
660 __u32 lfm_flags; /* flags, type specific */
664 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
665 offsetof(struct lov_foreign_md, lfm_value))
667 #define foreign_size_le(lfm) \
668 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
669 offsetof(struct lov_foreign_md, lfm_value))
672 * The stripe size fields are shared for the extension size storage, however
673 * the extension size is stored in KB, not bytes.
675 #define SEL_UNIT_SIZE 1024llu
682 #define DEXT "[%#llx, %#llx)"
683 #define PEXT(ext) (ext)->e_start, (ext)->e_end
685 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
686 struct lu_extent *e2)
688 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
691 static inline bool lu_extent_is_whole(struct lu_extent *e)
693 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
696 enum lov_comp_md_entry_flags {
697 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
698 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
699 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
700 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
701 LCME_FL_OFFLINE = 0x00000008, /* Not used */
702 LCME_FL_INIT = 0x00000010, /* instantiated */
703 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
704 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
705 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
706 * won't be stored on disk
710 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
711 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
714 /* The component flags can be set by users at creation/modification time. */
715 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
718 /* The mirror flags can be set by users at creation time. */
719 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
721 /* The allowed flags obtained from the client at component creation time. */
722 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
724 /* The mirror flags sent by client */
725 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
727 /* These flags have meaning when set in a default layout and will be inherited
728 * from the default/template layout set on a directory.
730 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
733 /* the highest bit in obdo::o_layout_version is used to mark if the file is
735 #define LU_LAYOUT_RESYNC LCME_FL_NEG
737 /* lcme_id can be specified as certain flags, and the the first
738 * bit of lcme_id is used to indicate that the ID is representing
739 * certain LCME_FL_* but not a real ID. Which implies we can have
740 * at most 31 flags (see LCME_FL_XXX). */
743 LCME_ID_MAX = 0x7FFFFFFF,
744 LCME_ID_ALL = 0xFFFFFFFF,
745 LCME_ID_NOT_ID = LCME_FL_NEG
748 #define LCME_ID_MASK LCME_ID_MAX
750 struct lov_comp_md_entry_v1 {
751 __u32 lcme_id; /* unique id of component */
752 __u32 lcme_flags; /* LCME_FL_XXX */
753 struct lu_extent lcme_extent; /* file extent for component */
754 __u32 lcme_offset; /* offset of component blob,
755 start from lov_comp_md_v1 */
756 __u32 lcme_size; /* size of component blob */
757 __u32 lcme_layout_gen;
758 __u64 lcme_timestamp; /* snapshot time if applicable*/
759 __u32 lcme_padding_1;
760 } __attribute__((packed));
762 #define SEQ_ID_MAX 0x0000FFFF
763 #define SEQ_ID_MASK SEQ_ID_MAX
764 /* bit 30:16 of lcme_id is used to store mirror id */
765 #define MIRROR_ID_MASK 0x7FFF0000
766 #define MIRROR_ID_NEG 0x8000
767 #define MIRROR_ID_SHIFT 16
769 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
771 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
774 static inline __u16 mirror_id_of(__u32 id)
776 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
780 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
782 enum lov_comp_md_flags {
783 /* the least 2 bits are used by FLR to record file state */
786 LCM_FL_WRITE_PENDING = 2,
787 LCM_FL_SYNC_PENDING = 3,
788 LCM_FL_FLR_MASK = 0x3,
791 struct lov_comp_md_v1 {
792 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
793 __u32 lcm_size; /* overall size including this struct */
794 __u32 lcm_layout_gen;
796 __u16 lcm_entry_count;
797 /* lcm_mirror_count stores the number of actual mirrors minus 1,
798 * so that non-flr files will have value 0 meaning 1 mirror. */
799 __u16 lcm_mirror_count;
800 __u16 lcm_padding1[3];
802 struct lov_comp_md_entry_v1 lcm_entries[0];
803 } __attribute__((packed));
805 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
807 if (stripes == (__u16)-1)
810 if (lmm_magic == LOV_USER_MAGIC_V1)
811 return sizeof(struct lov_user_md_v1) +
812 stripes * sizeof(struct lov_user_ost_data_v1);
813 return sizeof(struct lov_user_md_v3) +
814 stripes * sizeof(struct lov_user_ost_data_v1);
817 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
818 * use this. It is unsafe to #define those values in this header as it
819 * is possible the application has already #included <sys/stat.h>. */
820 #ifdef HAVE_LOV_USER_MDS_DATA
821 #define lov_user_mds_data lov_user_mds_data_v1
822 struct lov_user_mds_data_v1 {
823 lstat_t lmd_st; /* MDS stat struct */
824 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
825 } __attribute__((packed));
828 struct lmv_user_mds_data {
829 struct lu_fid lum_fid;
835 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
836 LMV_HASH_TYPE_ALL_CHARS = 1,
837 LMV_HASH_TYPE_FNV_1A_64 = 2,
841 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
843 #define LMV_HASH_NAME_ALL_CHARS "all_char"
844 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
846 /* not real hash type, but exposed to user as "space" hash type */
847 #define LMV_HASH_NAME_SPACE "space"
849 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
850 * and the higher part will be the flag to indicate the status of object,
851 * for example the object is being migrated. And the hash function
852 * might be interpreted differently with different flags. */
853 #define LMV_HASH_TYPE_MASK 0x0000ffff
855 /* once this is set on a plain directory default layout, newly created
856 * subdirectories will be distributed on all MDTs by space usage.
858 #define LMV_HASH_FLAG_SPACE 0x08000000
860 /* The striped directory has ever lost its master LMV EA, then LFSCK
861 * re-generated it. This flag is used to indicate such case. It is an
863 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
865 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
866 #define LMV_HASH_FLAG_MIGRATION 0x80000000
868 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
870 struct lustre_foreign_type {
872 const char *lft_name;
876 * LOV/LMV foreign types
878 enum lustre_foreign_types {
879 LU_FOREIGN_TYPE_NONE = 0,
880 LU_FOREIGN_TYPE_DAOS = 0xda05,
881 /* must be the max/last one */
882 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
885 extern struct lustre_foreign_type lu_foreign_types[];
887 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
888 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
889 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
890 #define lmv_user_md lmv_user_md_v1
891 struct lmv_user_md_v1 {
892 __u32 lum_magic; /* must be the first field */
893 __u32 lum_stripe_count; /* dirstripe count */
894 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
895 __u32 lum_hash_type; /* Dir stripe policy */
896 __u32 lum_type; /* LMV type: default */
900 char lum_pool_name[LOV_MAXPOOLNAME + 1];
901 struct lmv_user_mds_data lum_objects[0];
902 } __attribute__((packed));
904 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
906 if (size <= sizeof(struct lmv_user_md))
909 size -= sizeof(struct lmv_user_md);
910 return (size + sizeof(struct lmv_user_mds_data) - 1) /
911 sizeof(struct lmv_user_mds_data);
915 * NB, historically default layout didn't set type, but use XATTR name to differ
916 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
917 * and still use the same method.
920 LMV_TYPE_DEFAULT = 0x0000,
923 static inline int lmv_user_md_size(int stripes, int lmm_magic)
925 int size = sizeof(struct lmv_user_md);
927 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
928 size += stripes * sizeof(struct lmv_user_mds_data);
933 struct ll_recreate_obj {
939 __u64 id; /* holds object id */
940 __u32 generation; /* holds object generation */
941 __u32 f_type; /* holds object type or stripe idx when passing it to
942 * OST for saving into EA. */
950 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
951 const struct obd_uuid *u2)
953 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
956 static inline int obd_uuid_empty(struct obd_uuid *uuid)
958 return uuid->uuid[0] == '\0';
961 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
963 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
964 uuid->uuid[sizeof(*uuid) - 1] = '\0';
967 /* For printf's only, make sure uuid is terminated */
968 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
973 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
974 /* Obviously not safe, but for printfs, no real harm done...
975 we're always null-terminated, even in a race. */
976 static char temp[sizeof(*uuid)];
977 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
978 temp[sizeof(*uuid) - 1] = '\0';
981 return (char *)(uuid->uuid);
984 #define LUSTRE_MAXFSNAME 8
986 /* Extract fsname from uuid (or target name) of a target
987 e.g. (myfs-OST0007_UUID -> myfs)
988 see also deuuidify. */
989 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
993 strncpy(buf, uuid, buflen - 1);
994 buf[buflen - 1] = '\0';
995 p = strrchr(buf, '-');
1000 /* printf display format for Lustre FIDs
1001 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1002 #define FID_NOBRACE_LEN 40
1003 #define FID_LEN (FID_NOBRACE_LEN + 2)
1004 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1005 #define DFID "["DFID_NOBRACE"]"
1006 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1008 /* scanf input parse format for fids in DFID_NOBRACE format
1009 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1010 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1011 #define SFID "0x%llx:0x%x:0x%x"
1012 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1014 /********* Quotas **********/
1016 #define LUSTRE_QUOTABLOCK_BITS 10
1017 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1019 static inline __u64 lustre_stoqb(size_t space)
1021 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1024 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1025 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1026 #define Q_GETOINFO 0x800102 /* get obd quota info */
1027 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1028 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1030 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1031 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1032 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1033 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1034 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1035 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1036 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1037 /* lustre-specific control commands */
1038 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1039 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1040 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1041 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1043 /* In the current Lustre implementation, the grace time is either the time
1044 * or the timestamp to be used after some quota ID exceeds the soft limt,
1045 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1047 #define LQUOTA_GRACE_BITS 48
1048 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1049 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1050 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1051 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1052 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1054 /* different quota flags */
1056 /* the default quota flag, the corresponding quota ID will use the default
1057 * quota setting, the hardlimit and softlimit of its quota record in the global
1058 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1059 * and high 16 bits will contain this flag (see above comment).
1061 #define LQUOTA_FLAG_DEFAULT 0x0001
1063 #define ALLQUOTA 255 /* set all quota */
1064 static inline char *qtype_name(int qtype)
1077 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1078 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1081 #define N_PERMS_MAX 64
1083 struct perm_downcall_data {
1089 struct identity_downcall_data {
1096 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1097 __u32 idd_groups[0];
1100 struct sepol_downcall_data {
1102 time_t sdd_sepol_mtime;
1103 __u16 sdd_sepol_len;
1107 #ifdef NEED_QUOTA_DEFS
1109 #define QIF_BLIMITS 1
1111 #define QIF_ILIMITS 4
1112 #define QIF_INODES 8
1113 #define QIF_BTIME 16
1114 #define QIF_ITIME 32
1115 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1116 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1117 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1118 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1121 #endif /* !__KERNEL__ */
1123 /* lustre volatile file support
1124 * file name header: ".^L^S^T^R:volatile"
1126 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1127 #define LUSTRE_VOLATILE_HDR_LEN 14
1129 enum lustre_quota_version {
1133 /* XXX: same as if_dqinfo struct in kernel */
1141 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1143 __u64 dqb_bhardlimit;
1144 __u64 dqb_bsoftlimit;
1146 __u64 dqb_ihardlimit;
1147 __u64 dqb_isoftlimit;
1148 __u64 dqb_curinodes;
1162 struct if_quotactl {
1169 struct obd_dqinfo qc_dqinfo;
1170 struct obd_dqblk qc_dqblk;
1172 struct obd_uuid obd_uuid;
1175 /* swap layout flags */
1176 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1177 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1178 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1179 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1180 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1182 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1183 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1184 struct lustre_swap_layouts {
1192 /** Bit-mask of valid attributes */
1193 /* The LA_* flags are written to disk as part of the ChangeLog records
1194 * so they are part of the on-disk and network protocol, and cannot be changed.
1195 * Only the first 12 bits are currently saved.
1198 LA_ATIME = 1 << 0, /* 0x00001 */
1199 LA_MTIME = 1 << 1, /* 0x00002 */
1200 LA_CTIME = 1 << 2, /* 0x00004 */
1201 LA_SIZE = 1 << 3, /* 0x00008 */
1202 LA_MODE = 1 << 4, /* 0x00010 */
1203 LA_UID = 1 << 5, /* 0x00020 */
1204 LA_GID = 1 << 6, /* 0x00040 */
1205 LA_BLOCKS = 1 << 7, /* 0x00080 */
1206 LA_TYPE = 1 << 8, /* 0x00100 */
1207 LA_FLAGS = 1 << 9, /* 0x00200 */
1208 LA_NLINK = 1 << 10, /* 0x00400 */
1209 LA_RDEV = 1 << 11, /* 0x00800 */
1210 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1211 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1212 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1213 LA_PROJID = 1 << 15, /* 0x08000 */
1214 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1215 LA_LSIZE = 1 << 17, /* 0x20000 */
1216 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1218 * Attributes must be transmitted to OST objects
1220 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1223 #define MDS_FMODE_READ 00000001
1224 #define MDS_FMODE_WRITE 00000002
1226 #define MDS_FMODE_CLOSED 00000000
1227 #define MDS_FMODE_EXEC 00000004
1228 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1229 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1230 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1232 #define MDS_OPEN_CREATED 00000010
1233 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1235 #define MDS_OPEN_CREAT 00000100
1236 #define MDS_OPEN_EXCL 00000200
1237 #define MDS_OPEN_TRUNC 00001000
1238 #define MDS_OPEN_APPEND 00002000
1239 #define MDS_OPEN_SYNC 00010000
1240 #define MDS_OPEN_DIRECTORY 00200000
1242 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1243 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1244 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1245 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1246 * We do not support JOIN FILE
1247 * anymore, reserve this flags
1248 * just for preventing such bit
1251 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1252 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1253 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1254 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1255 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1257 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1259 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1260 * delegation, succeed if it's not
1261 * being opened with conflict mode.
1263 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1265 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1266 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1267 * for newly created file */
1269 /* lustre internal open flags, which should not be set from user space */
1270 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1271 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1272 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1273 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1277 /********* Changelogs **********/
1278 /** Changelog record types */
1279 enum changelog_rec_type {
1282 CL_CREATE = 1, /* namespace */
1283 CL_MKDIR = 2, /* namespace */
1284 CL_HARDLINK = 3, /* namespace */
1285 CL_SOFTLINK = 4, /* namespace */
1286 CL_MKNOD = 5, /* namespace */
1287 CL_UNLINK = 6, /* namespace */
1288 CL_RMDIR = 7, /* namespace */
1289 CL_RENAME = 8, /* namespace */
1290 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1291 CL_OPEN = 10, /* not currently used */
1292 CL_CLOSE = 11, /* may be written to log only with mtime change */
1293 CL_LAYOUT = 12, /* file layout/striping modified */
1297 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1298 CL_HSM = 16, /* HSM specific events, see flags */
1299 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1303 CL_FLRW = 21, /* FLR: file was firstly written */
1304 CL_RESYNC = 22, /* FLR: file was resync-ed */
1306 CL_DN_OPEN = 24, /* denied open */
1310 static inline const char *changelog_type2str(int type) {
1311 static const char *changelog_str[] = {
1312 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1313 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1314 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1315 "FLRW", "RESYNC","GXATR", "NOPEN",
1318 if (type >= 0 && type < CL_LAST)
1319 return changelog_str[type];
1323 /* 12 bits of per-record data can be stored in the bottom of the flags */
1324 #define CLF_FLAGSHIFT 12
1325 enum changelog_rec_flags {
1326 CLF_VERSION = 0x1000,
1327 CLF_RENAME = 0x2000,
1329 CLF_EXTRA_FLAGS = 0x8000,
1330 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1332 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1333 CLF_VERMASK = ~CLF_FLAGMASK,
1337 /* Anything under the flagmask may be per-type (if desired) */
1338 /* Flags for unlink */
1339 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1340 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1341 /* HSM cleaning needed */
1342 /* Flags for rename */
1343 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1345 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1346 * has an archive in backend */
1349 /* 12b used (from high weight to low weight):
1354 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1355 #define CLF_HSM_ERR_H 6
1356 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1357 #define CLF_HSM_EVENT_H 9
1358 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1359 #define CLF_HSM_FLAG_H 11
1360 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1361 #define CLF_HSM_SPARE_H 15
1362 #define CLF_HSM_LAST 15
1364 /* Remove bits higher than _h, then extract the value
1365 * between _h and _l by shifting lower weigth to bit 0. */
1366 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1367 >> (CLF_HSM_LAST - _h + _l))
1369 #define CLF_HSM_SUCCESS 0x00
1370 #define CLF_HSM_MAXERROR 0x7E
1371 #define CLF_HSM_ERROVERFLOW 0x7F
1373 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1375 /* 3 bits field => 8 values allowed */
1387 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1389 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1393 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1396 *clf_flags |= (he << CLF_HSM_EVENT_L);
1399 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1401 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1404 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1407 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1410 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1412 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1415 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1418 *clf_flags |= (error << CLF_HSM_ERR_L);
1421 enum changelog_rec_extra_flags {
1423 CLFE_UIDGID = 0x0001,
1426 CLFE_XATTR = 0x0008,
1427 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1430 enum changelog_send_flag {
1431 /* Not yet implemented */
1432 CHANGELOG_FLAG_FOLLOW = 0x01,
1433 /* Blocking IO makes sense in case of slow user parsing of the records,
1434 * but it also prevents us from cleaning up if the records are not
1436 CHANGELOG_FLAG_BLOCK = 0x02,
1437 /* Pack jobid into the changelog records if available. */
1438 CHANGELOG_FLAG_JOBID = 0x04,
1439 /* Pack additional flag bits into the changelog record */
1440 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1443 enum changelog_send_extra_flag {
1444 /* Pack uid/gid into the changelog record */
1445 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1446 /* Pack nid into the changelog record */
1447 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1448 /* Pack open mode into the changelog record */
1449 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1450 /* Pack xattr name into the changelog record */
1451 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1454 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1455 changelog_rec_offset(CLF_SUPPORTED, \
1458 /* 31 usable bytes string + null terminator. */
1459 #define LUSTRE_JOBID_SIZE 32
1461 /* This is the minimal changelog record. It can contain extensions
1462 * such as rename fields or process jobid. Its exact content is described
1463 * by the cr_flags and cr_extra_flags.
1465 * Extensions are packed in the same order as their corresponding flags,
1466 * then in the same order as their corresponding extra flags.
1468 struct changelog_rec {
1470 __u16 cr_flags; /**< \a changelog_rec_flags */
1471 __u32 cr_type; /**< \a changelog_rec_type */
1472 __u64 cr_index; /**< changelog record number */
1473 __u64 cr_prev; /**< last index for this target fid */
1476 struct lu_fid cr_tfid; /**< target fid */
1477 __u32 cr_markerflags; /**< CL_MARK flags */
1479 struct lu_fid cr_pfid; /**< parent fid */
1482 /* Changelog extension for RENAME. */
1483 struct changelog_ext_rename {
1484 struct lu_fid cr_sfid; /**< source fid, or zero */
1485 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1488 /* Changelog extension to include JOBID. */
1489 struct changelog_ext_jobid {
1490 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1493 /* Changelog extension to include additional flags. */
1494 struct changelog_ext_extra_flags {
1495 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1498 /* Changelog extra extension to include UID/GID. */
1499 struct changelog_ext_uidgid {
1504 /* Changelog extra extension to include NID. */
1505 struct changelog_ext_nid {
1506 /* have __u64 instead of lnet_nid_t type for use by client api */
1508 /* for use when IPv6 support is added */
1513 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1514 struct changelog_ext_openmode {
1518 /* Changelog extra extension to include xattr */
1519 struct changelog_ext_xattr {
1520 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1523 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1524 const struct changelog_rec *rec);
1526 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1527 enum changelog_rec_extra_flags cref)
1529 size_t size = sizeof(struct changelog_rec);
1531 if (crf & CLF_RENAME)
1532 size += sizeof(struct changelog_ext_rename);
1534 if (crf & CLF_JOBID)
1535 size += sizeof(struct changelog_ext_jobid);
1537 if (crf & CLF_EXTRA_FLAGS) {
1538 size += sizeof(struct changelog_ext_extra_flags);
1539 if (cref & CLFE_UIDGID)
1540 size += sizeof(struct changelog_ext_uidgid);
1541 if (cref & CLFE_NID)
1542 size += sizeof(struct changelog_ext_nid);
1543 if (cref & CLFE_OPEN)
1544 size += sizeof(struct changelog_ext_openmode);
1545 if (cref & CLFE_XATTR)
1546 size += sizeof(struct changelog_ext_xattr);
1552 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1554 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1556 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1557 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1559 return changelog_rec_offset(rec->cr_flags, cref);
1562 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1564 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1568 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1570 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1572 return (struct changelog_ext_rename *)((char *)rec +
1573 changelog_rec_offset(crf,
1577 /* The jobid follows the rename extension, if present */
1579 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1581 enum changelog_rec_flags crf = rec->cr_flags &
1582 (CLF_VERSION | CLF_RENAME);
1584 return (struct changelog_ext_jobid *)((char *)rec +
1585 changelog_rec_offset(crf,
1589 /* The additional flags follow the rename and jobid extensions, if present */
1591 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1592 const struct changelog_rec *rec)
1594 enum changelog_rec_flags crf = rec->cr_flags &
1595 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1597 return (struct changelog_ext_extra_flags *)((char *)rec +
1598 changelog_rec_offset(crf,
1602 /* The uid/gid is the first extra extension */
1604 struct changelog_ext_uidgid *changelog_rec_uidgid(
1605 const struct changelog_rec *rec)
1607 enum changelog_rec_flags crf = rec->cr_flags &
1608 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1610 return (struct changelog_ext_uidgid *)((char *)rec +
1611 changelog_rec_offset(crf,
1615 /* The nid is the second extra extension */
1617 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1619 enum changelog_rec_flags crf = rec->cr_flags &
1620 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1621 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1623 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1624 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1627 return (struct changelog_ext_nid *)((char *)rec +
1628 changelog_rec_offset(crf, cref));
1631 /* The OPEN mode is the third extra extension */
1633 struct changelog_ext_openmode *changelog_rec_openmode(
1634 const struct changelog_rec *rec)
1636 enum changelog_rec_flags crf = rec->cr_flags &
1637 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1638 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1640 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1641 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1642 (CLFE_UIDGID | CLFE_NID);
1644 return (struct changelog_ext_openmode *)((char *)rec +
1645 changelog_rec_offset(crf, cref));
1648 /* The xattr name is the fourth extra extension */
1650 struct changelog_ext_xattr *changelog_rec_xattr(
1651 const struct changelog_rec *rec)
1653 enum changelog_rec_flags crf = rec->cr_flags &
1654 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1655 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1657 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1658 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1659 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1661 return (struct changelog_ext_xattr *)((char *)rec +
1662 changelog_rec_offset(crf, cref));
1665 /* The name follows the rename, jobid and extra flags extns, if present */
1666 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1668 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1670 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1671 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1673 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1674 cref & CLFE_SUPPORTED);
1677 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1679 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1682 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1684 char *cr_name = changelog_rec_name(rec);
1686 return cr_name + strlen(cr_name) + 1;
1690 * Remap a record to the desired format as specified by the crf flags.
1691 * The record must be big enough to contain the final remapped version.
1692 * Superfluous extension fields are removed and missing ones are added
1693 * and zeroed. The flags of the record are updated accordingly.
1695 * The jobid and rename extensions can be added to a record, to match the
1696 * format an application expects, typically. In this case, the newly added
1697 * fields will be zeroed.
1698 * The Jobid field can be removed, to guarantee compatibility with older
1699 * clients that don't expect this field in the records they process.
1701 * The following assumptions are being made:
1702 * - CLF_RENAME will not be removed
1703 * - CLF_JOBID will not be added without CLF_RENAME being added too
1704 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1706 * @param[in,out] rec The record to remap.
1707 * @param[in] crf_wanted Flags describing the desired extensions.
1708 * @param[in] cref_want Flags describing the desired extra extensions.
1710 static inline void changelog_remap_rec(struct changelog_rec *rec,
1711 enum changelog_rec_flags crf_wanted,
1712 enum changelog_rec_extra_flags cref_want)
1714 char *xattr_mov = NULL;
1715 char *omd_mov = NULL;
1716 char *nid_mov = NULL;
1717 char *uidgid_mov = NULL;
1721 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1723 crf_wanted &= CLF_SUPPORTED;
1724 cref_want &= CLFE_SUPPORTED;
1726 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1727 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1728 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1729 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1735 /* First move the variable-length name field */
1736 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1737 changelog_rec_name(rec), rec->cr_namelen);
1739 /* Locations of extensions in the remapped record */
1740 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1741 xattr_mov = (char *)rec +
1742 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1743 cref_want & ~CLFE_XATTR);
1744 omd_mov = (char *)rec +
1745 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1746 cref_want & ~(CLFE_OPEN |
1748 nid_mov = (char *)rec +
1749 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1750 cref_want & ~(CLFE_NID |
1753 uidgid_mov = (char *)rec +
1754 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1755 cref_want & ~(CLFE_UIDGID |
1759 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1762 ef_mov = (char *)rec +
1763 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1765 jid_mov = (char *)rec +
1766 changelog_rec_offset(crf_wanted &
1767 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1769 rnm_mov = (char *)rec +
1770 changelog_rec_offset(crf_wanted &
1776 /* Move the extension fields to the desired positions */
1777 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1778 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1779 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1780 memmove(xattr_mov, changelog_rec_xattr(rec),
1781 sizeof(struct changelog_ext_xattr));
1783 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1784 memmove(omd_mov, changelog_rec_openmode(rec),
1785 sizeof(struct changelog_ext_openmode));
1787 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1788 memmove(nid_mov, changelog_rec_nid(rec),
1789 sizeof(struct changelog_ext_nid));
1791 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1792 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1793 sizeof(struct changelog_ext_uidgid));
1795 memmove(ef_mov, changelog_rec_extra_flags(rec),
1796 sizeof(struct changelog_ext_extra_flags));
1799 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1800 memmove(jid_mov, changelog_rec_jobid(rec),
1801 sizeof(struct changelog_ext_jobid));
1803 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1804 memmove(rnm_mov, changelog_rec_rename(rec),
1805 sizeof(struct changelog_ext_rename));
1807 /* Clear newly added fields */
1808 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1809 !(cref & CLFE_XATTR))
1810 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1812 if (omd_mov && (cref_want & CLFE_OPEN) &&
1813 !(cref & CLFE_OPEN))
1814 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1816 if (nid_mov && (cref_want & CLFE_NID) &&
1818 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1820 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1821 !(cref & CLFE_UIDGID))
1822 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1824 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1825 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1826 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1828 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1829 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1831 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1832 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1834 /* Update the record's flags accordingly */
1835 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1836 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1837 changelog_rec_extra_flags(rec)->cr_extra_flags =
1838 changelog_rec_extra_flags(rec)->cr_extra_flags |
1842 enum changelog_message_type {
1843 CL_RECORD = 10, /* message is a changelog_rec */
1844 CL_EOF = 11, /* at end of current changelog */
1847 /********* Misc **********/
1849 struct ioc_data_version {
1851 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1852 __u32 idv_flags; /* enum ioc_data_version_flags */
1855 enum ioc_data_version_flags {
1856 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1857 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1861 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1864 #define dot_lustre_name ".lustre"
1867 /********* HSM **********/
1869 /** HSM per-file state
1870 * See HSM_FLAGS below.
1873 HS_NONE = 0x00000000,
1874 HS_EXISTS = 0x00000001,
1875 HS_DIRTY = 0x00000002,
1876 HS_RELEASED = 0x00000004,
1877 HS_ARCHIVED = 0x00000008,
1878 HS_NORELEASE = 0x00000010,
1879 HS_NOARCHIVE = 0x00000020,
1880 HS_LOST = 0x00000040,
1883 /* HSM user-setable flags. */
1884 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1886 /* Other HSM flags. */
1887 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1890 * All HSM-related possible flags that could be applied to a file.
1891 * This should be kept in sync with hsm_states.
1893 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1896 * HSM request progress state
1898 enum hsm_progress_states {
1905 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1908 case HPS_WAITING: return "waiting";
1909 case HPS_RUNNING: return "running";
1910 case HPS_DONE: return "done";
1911 default: return "unknown";
1918 } __attribute__((packed));
1921 * Current HSM states of a Lustre file.
1923 * This structure purpose is to be sent to user-space mainly. It describes the
1924 * current HSM flags and in-progress action.
1926 struct hsm_user_state {
1927 /** Current HSM states, from enum hsm_states. */
1929 __u32 hus_archive_id;
1930 /** The current undergoing action, if there is one */
1931 __u32 hus_in_progress_state;
1932 __u32 hus_in_progress_action;
1933 struct hsm_extent hus_in_progress_location;
1934 char hus_extended_info[];
1937 struct hsm_state_set_ioc {
1938 struct lu_fid hssi_fid;
1940 __u64 hssi_clearmask;
1944 * This structure describes the current in-progress action for a file.
1945 * it is retuned to user space and send over the wire
1947 struct hsm_current_action {
1948 /** The current undergoing action, if there is one */
1949 /* state is one of hsm_progress_states */
1951 /* action is one of hsm_user_action */
1953 struct hsm_extent hca_location;
1956 /***** HSM user requests ******/
1957 /* User-generated (lfs/ioctl) request types */
1958 enum hsm_user_action {
1959 HUA_NONE = 1, /* no action (noop) */
1960 HUA_ARCHIVE = 10, /* copy to hsm */
1961 HUA_RESTORE = 11, /* prestage */
1962 HUA_RELEASE = 12, /* drop ost objects */
1963 HUA_REMOVE = 13, /* remove from archive */
1964 HUA_CANCEL = 14 /* cancel a request */
1967 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1970 case HUA_NONE: return "NOOP";
1971 case HUA_ARCHIVE: return "ARCHIVE";
1972 case HUA_RESTORE: return "RESTORE";
1973 case HUA_RELEASE: return "RELEASE";
1974 case HUA_REMOVE: return "REMOVE";
1975 case HUA_CANCEL: return "CANCEL";
1976 default: return "UNKNOWN";
1981 * List of hr_flags (bit field)
1983 #define HSM_FORCE_ACTION 0x0001
1984 /* used by CT, cannot be set by user */
1985 #define HSM_GHOST_COPY 0x0002
1988 * Contains all the fixed part of struct hsm_user_request.
1991 struct hsm_request {
1992 __u32 hr_action; /* enum hsm_user_action */
1993 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1994 __u64 hr_flags; /* request flags */
1995 __u32 hr_itemcount; /* item count in hur_user_item vector */
1999 struct hsm_user_item {
2000 struct lu_fid hui_fid;
2001 struct hsm_extent hui_extent;
2002 } __attribute__((packed));
2004 struct hsm_user_request {
2005 struct hsm_request hur_request;
2006 struct hsm_user_item hur_user_item[0];
2007 /* extra data blob at end of struct (after all
2008 * hur_user_items), only use helpers to access it
2010 } __attribute__((packed));
2012 /** Return pointer to data field in a hsm user request */
2013 static inline void *hur_data(struct hsm_user_request *hur)
2015 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2019 * Compute the current length of the provided hsm_user_request. This returns -1
2020 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
2022 * return -1 on bounds check error.
2024 static inline ssize_t hur_len(struct hsm_user_request *hur)
2028 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2029 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2030 (__u64)hur->hur_request.hr_itemcount *
2031 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2033 if (size != (ssize_t)size)
2039 /****** HSM RPCs to copytool *****/
2040 /* Message types the copytool may receive */
2041 enum hsm_message_type {
2042 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2045 /* Actions the copytool may be instructed to take for a given action_item */
2046 enum hsm_copytool_action {
2047 HSMA_NONE = 10, /* no action */
2048 HSMA_ARCHIVE = 20, /* arbitrary offset */
2054 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2057 case HSMA_NONE: return "NOOP";
2058 case HSMA_ARCHIVE: return "ARCHIVE";
2059 case HSMA_RESTORE: return "RESTORE";
2060 case HSMA_REMOVE: return "REMOVE";
2061 case HSMA_CANCEL: return "CANCEL";
2062 default: return "UNKNOWN";
2066 /* Copytool item action description */
2067 struct hsm_action_item {
2068 __u32 hai_len; /* valid size of this struct */
2069 __u32 hai_action; /* hsm_copytool_action, but use known size */
2070 struct lu_fid hai_fid; /* Lustre FID to operate on */
2071 struct lu_fid hai_dfid; /* fid used for data access */
2072 struct hsm_extent hai_extent; /* byte range to operate on */
2073 __u64 hai_cookie; /* action cookie from coordinator */
2074 __u64 hai_gid; /* grouplock id */
2075 char hai_data[0]; /* variable length */
2076 } __attribute__((packed));
2079 * helper function which print in hexa the first bytes of
2082 * \param hai [IN] record to print
2083 * \param buffer [IN,OUT] buffer to write the hex string to
2084 * \param len [IN] max buffer length
2088 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2089 char *buffer, size_t len)
2096 data_len = hai->hai_len - sizeof(*hai);
2097 for (i = 0; (i < data_len) && (len > 2); i++) {
2098 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2108 /* Copytool action list */
2109 #define HAL_VERSION 1
2110 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2111 struct hsm_action_list {
2113 __u32 hal_count; /* number of hai's to follow */
2114 __u64 hal_compound_id; /* returned by coordinator, ignored */
2116 __u32 hal_archive_id; /* which archive backend */
2118 char hal_fsname[0]; /* null-terminated */
2119 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2120 boundaries. See hai_zero */
2121 } __attribute__((packed));
2123 /* Return pointer to first hai in action list */
2124 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2126 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2128 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2131 /* Return pointer to next hai */
2132 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2134 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2136 return (struct hsm_action_item *)((char *)hai + offset);
2139 /* Return size of an hsm_action_list */
2140 static inline size_t hal_size(struct hsm_action_list *hal)
2144 struct hsm_action_item *hai;
2146 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2147 hai = hai_first(hal);
2148 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2149 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2155 * describe the attributes to be set on imported file
2157 struct hsm_user_import {
2166 __u32 hui_archive_id;
2169 /* Copytool progress reporting */
2170 #define HP_FLAG_COMPLETED 0x01
2171 #define HP_FLAG_RETRY 0x02
2173 struct hsm_progress {
2174 struct lu_fid hp_fid;
2176 struct hsm_extent hp_extent;
2178 __u16 hp_errval; /* positive val */
2183 __u64 hc_data_version;
2185 __u16 hc_errval; /* positive val */
2187 struct hsm_action_item hc_hai;
2191 enum llapi_json_types {
2192 LLAPI_JSON_INTEGER = 1,
2198 struct llapi_json_item {
2207 struct llapi_json_item *lji_next;
2210 struct llapi_json_item_list {
2211 int ljil_item_count;
2212 struct llapi_json_item *ljil_items;
2215 enum lu_ladvise_type {
2216 LU_LADVISE_INVALID = 0,
2217 LU_LADVISE_WILLREAD = 1,
2218 LU_LADVISE_DONTNEED = 2,
2219 LU_LADVISE_LOCKNOEXPAND = 3,
2220 LU_LADVISE_LOCKAHEAD = 4,
2224 #define LU_LADVISE_NAMES { \
2225 [LU_LADVISE_WILLREAD] = "willread", \
2226 [LU_LADVISE_DONTNEED] = "dontneed", \
2227 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2228 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2231 /* This is the userspace argument for ladvise. It is currently the same as
2232 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2233 * may need info which is only used locally. */
2234 struct llapi_lu_ladvise {
2235 __u16 lla_advice; /* advice type */
2236 __u16 lla_value1; /* values for different advice types */
2238 __u64 lla_start; /* first byte of extent for advice */
2239 __u64 lla_end; /* last byte of extent for advice */
2245 LF_ASYNC = 0x00000001,
2246 LF_UNSET = 0x00000002,
2249 #define LADVISE_MAGIC 0x1ADF1CE0
2250 /* Masks of valid flags for each advice */
2251 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2252 /* Flags valid for all advices not explicitly specified */
2253 #define LF_DEFAULT_MASK LF_ASYNC
2255 #define LF_MASK (LF_ASYNC | LF_UNSET)
2257 #define lla_lockahead_mode lla_value1
2258 #define lla_peradvice_flags lla_value2
2259 #define lla_lockahead_result lla_value3
2261 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2262 * is used on the wire. It is defined separately as we may need info which is
2263 * only used locally. */
2264 struct llapi_ladvise_hdr {
2265 __u32 lah_magic; /* LADVISE_MAGIC */
2266 __u32 lah_count; /* number of advices */
2267 __u64 lah_flags; /* from enum ladvise_flag */
2268 __u32 lah_value1; /* unused */
2269 __u32 lah_value2; /* unused */
2270 __u64 lah_value3; /* unused */
2271 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2274 #define LAH_COUNT_MAX (1024)
2278 SK_CRYPT_INVALID = -1,
2280 SK_CRYPT_AES256_CTR = 1,
2284 SK_HMAC_INVALID = -1,
2290 struct sk_crypt_type {
2291 const char *sct_name;
2295 struct sk_hmac_type {
2296 const char *sht_name;
2300 enum lock_mode_user {
2306 #define LOCK_MODE_NAMES { \
2307 [MODE_READ_USER] = "READ",\
2308 [MODE_WRITE_USER] = "WRITE"\
2311 enum lockahead_results {
2312 LLA_RESULT_SENT = 0,
2313 LLA_RESULT_DIFFERENT,
2317 enum lu_heat_flag_bit {
2318 LU_HEAT_FLAG_BIT_INVALID = 0,
2319 LU_HEAT_FLAG_BIT_OFF,
2320 LU_HEAT_FLAG_BIT_CLEAR,
2324 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2325 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2328 enum obd_heat_type {
2329 OBD_HEAT_READSAMPLE = 0,
2330 OBD_HEAT_WRITESAMPLE = 1,
2331 OBD_HEAT_READBYTE = 2,
2332 OBD_HEAT_WRITEBYTE = 3,
2336 #define LU_HEAT_NAMES { \
2337 [OBD_HEAT_READSAMPLE] = "readsample", \
2338 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2339 [OBD_HEAT_READBYTE] = "readbyte", \
2340 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2355 static inline const char *pcc_type2string(enum lu_pcc_type type)
2360 case LU_PCC_READWRITE:
2367 struct lu_pcc_attach {
2368 __u32 pcca_type; /* PCC type */
2369 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2372 enum lu_pcc_detach_opts {
2373 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2374 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2377 struct lu_pcc_detach_fid {
2378 /* fid of the file to detach */
2379 struct lu_fid pccd_fid;
2383 struct lu_pcc_detach {
2387 enum lu_pcc_state_flags {
2388 PCC_STATE_FL_NONE = 0x0,
2389 /* The inode attr is cached locally */
2390 PCC_STATE_FL_ATTR_VALID = 0x01,
2391 /* The file is being attached into PCC */
2392 PCC_STATE_FL_ATTACHING = 0x02,
2393 /* Allow to auto attach at open */
2394 PCC_STATE_FL_OPEN_ATTACH = 0x04,
2397 struct lu_pcc_state {
2398 __u32 pccs_type; /* enum lu_pcc_type */
2399 __u32 pccs_open_count;
2400 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2402 char pccs_path[PATH_MAX];
2405 #if defined(__cplusplus)
2409 /** @} lustreuser */
2411 #endif /* _LUSTRE_USER_H */