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 mirror flags can be set by users at creation time. */
715 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
717 /* The allowed flags obtained from the client at component creation time. */
718 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
720 /* The mirror flags sent by client */
721 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
723 /* These flags have meaning when set in a default layout and will be inherited
724 * from the default/template layout set on a directory.
726 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
729 /* the highest bit in obdo::o_layout_version is used to mark if the file is
731 #define LU_LAYOUT_RESYNC LCME_FL_NEG
733 /* lcme_id can be specified as certain flags, and the the first
734 * bit of lcme_id is used to indicate that the ID is representing
735 * certain LCME_FL_* but not a real ID. Which implies we can have
736 * at most 31 flags (see LCME_FL_XXX). */
739 LCME_ID_MAX = 0x7FFFFFFF,
740 LCME_ID_ALL = 0xFFFFFFFF,
741 LCME_ID_NOT_ID = LCME_FL_NEG
744 #define LCME_ID_MASK LCME_ID_MAX
746 struct lov_comp_md_entry_v1 {
747 __u32 lcme_id; /* unique id of component */
748 __u32 lcme_flags; /* LCME_FL_XXX */
749 struct lu_extent lcme_extent; /* file extent for component */
750 __u32 lcme_offset; /* offset of component blob,
751 start from lov_comp_md_v1 */
752 __u32 lcme_size; /* size of component blob */
753 __u32 lcme_layout_gen;
754 __u64 lcme_timestamp; /* snapshot time if applicable*/
755 __u32 lcme_padding_1;
756 } __attribute__((packed));
758 #define SEQ_ID_MAX 0x0000FFFF
759 #define SEQ_ID_MASK SEQ_ID_MAX
760 /* bit 30:16 of lcme_id is used to store mirror id */
761 #define MIRROR_ID_MASK 0x7FFF0000
762 #define MIRROR_ID_NEG 0x8000
763 #define MIRROR_ID_SHIFT 16
765 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
767 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
770 static inline __u16 mirror_id_of(__u32 id)
772 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
776 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
778 enum lov_comp_md_flags {
779 /* the least 2 bits are used by FLR to record file state */
782 LCM_FL_WRITE_PENDING = 2,
783 LCM_FL_SYNC_PENDING = 3,
784 LCM_FL_FLR_MASK = 0x3,
787 struct lov_comp_md_v1 {
788 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
789 __u32 lcm_size; /* overall size including this struct */
790 __u32 lcm_layout_gen;
792 __u16 lcm_entry_count;
793 /* lcm_mirror_count stores the number of actual mirrors minus 1,
794 * so that non-flr files will have value 0 meaning 1 mirror. */
795 __u16 lcm_mirror_count;
796 __u16 lcm_padding1[3];
798 struct lov_comp_md_entry_v1 lcm_entries[0];
799 } __attribute__((packed));
801 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
803 if (stripes == (__u16)-1)
806 if (lmm_magic == LOV_USER_MAGIC_V1)
807 return sizeof(struct lov_user_md_v1) +
808 stripes * sizeof(struct lov_user_ost_data_v1);
809 return sizeof(struct lov_user_md_v3) +
810 stripes * sizeof(struct lov_user_ost_data_v1);
813 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
814 * use this. It is unsafe to #define those values in this header as it
815 * is possible the application has already #included <sys/stat.h>. */
816 #ifdef HAVE_LOV_USER_MDS_DATA
817 #define lov_user_mds_data lov_user_mds_data_v1
818 struct lov_user_mds_data_v1 {
819 lstat_t lmd_st; /* MDS stat struct */
820 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
821 } __attribute__((packed));
824 struct lmv_user_mds_data {
825 struct lu_fid lum_fid;
831 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
832 LMV_HASH_TYPE_ALL_CHARS = 1,
833 LMV_HASH_TYPE_FNV_1A_64 = 2,
837 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
839 #define LMV_HASH_NAME_ALL_CHARS "all_char"
840 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
842 /* not real hash type, but exposed to user as "space" hash type */
843 #define LMV_HASH_NAME_SPACE "space"
845 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
846 * and the higher part will be the flag to indicate the status of object,
847 * for example the object is being migrated. And the hash function
848 * might be interpreted differently with different flags. */
849 #define LMV_HASH_TYPE_MASK 0x0000ffff
851 /* once this is set on a plain directory default layout, newly created
852 * subdirectories will be distributed on all MDTs by space usage.
854 #define LMV_HASH_FLAG_SPACE 0x08000000
856 /* The striped directory has ever lost its master LMV EA, then LFSCK
857 * re-generated it. This flag is used to indicate such case. It is an
859 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
861 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
862 #define LMV_HASH_FLAG_MIGRATION 0x80000000
864 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
866 struct lustre_foreign_type {
868 const char *lft_name;
872 * LOV/LMV foreign types
874 enum lustre_foreign_types {
875 LU_FOREIGN_TYPE_NONE = 0,
876 LU_FOREIGN_TYPE_DAOS = 0xda05,
877 /* must be the max/last one */
878 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
881 extern struct lustre_foreign_type lu_foreign_types[];
883 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
884 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
885 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
886 #define lmv_user_md lmv_user_md_v1
887 struct lmv_user_md_v1 {
888 __u32 lum_magic; /* must be the first field */
889 __u32 lum_stripe_count; /* dirstripe count */
890 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
891 __u32 lum_hash_type; /* Dir stripe policy */
892 __u32 lum_type; /* LMV type: default */
896 char lum_pool_name[LOV_MAXPOOLNAME + 1];
897 struct lmv_user_mds_data lum_objects[0];
898 } __attribute__((packed));
900 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
902 if (size <= sizeof(struct lmv_user_md))
905 size -= sizeof(struct lmv_user_md);
906 return (size + sizeof(struct lmv_user_mds_data) - 1) /
907 sizeof(struct lmv_user_mds_data);
911 * NB, historically default layout didn't set type, but use XATTR name to differ
912 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
913 * and still use the same method.
916 LMV_TYPE_DEFAULT = 0x0000,
919 static inline int lmv_user_md_size(int stripes, int lmm_magic)
921 int size = sizeof(struct lmv_user_md);
923 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
924 size += stripes * sizeof(struct lmv_user_mds_data);
929 struct ll_recreate_obj {
935 __u64 id; /* holds object id */
936 __u32 generation; /* holds object generation */
937 __u32 f_type; /* holds object type or stripe idx when passing it to
938 * OST for saving into EA. */
946 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
947 const struct obd_uuid *u2)
949 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
952 static inline int obd_uuid_empty(struct obd_uuid *uuid)
954 return uuid->uuid[0] == '\0';
957 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
959 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
960 uuid->uuid[sizeof(*uuid) - 1] = '\0';
963 /* For printf's only, make sure uuid is terminated */
964 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
969 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
970 /* Obviously not safe, but for printfs, no real harm done...
971 we're always null-terminated, even in a race. */
972 static char temp[sizeof(*uuid)];
973 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
974 temp[sizeof(*uuid) - 1] = '\0';
977 return (char *)(uuid->uuid);
980 #define LUSTRE_MAXFSNAME 8
982 /* Extract fsname from uuid (or target name) of a target
983 e.g. (myfs-OST0007_UUID -> myfs)
984 see also deuuidify. */
985 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
989 strncpy(buf, uuid, buflen - 1);
990 buf[buflen - 1] = '\0';
991 p = strrchr(buf, '-');
996 /* printf display format for Lustre FIDs
997 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
998 #define FID_NOBRACE_LEN 40
999 #define FID_LEN (FID_NOBRACE_LEN + 2)
1000 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1001 #define DFID "["DFID_NOBRACE"]"
1002 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1004 /* scanf input parse format for fids in DFID_NOBRACE format
1005 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1006 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1007 #define SFID "0x%llx:0x%x:0x%x"
1008 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1010 /********* Quotas **********/
1012 #define LUSTRE_QUOTABLOCK_BITS 10
1013 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1015 static inline __u64 lustre_stoqb(size_t space)
1017 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1020 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1021 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1022 #define Q_GETOINFO 0x800102 /* get obd quota info */
1023 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1024 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1026 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1027 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1028 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1029 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1030 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1031 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1032 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1033 /* lustre-specific control commands */
1034 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1035 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1036 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1037 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1039 /* In the current Lustre implementation, the grace time is either the time
1040 * or the timestamp to be used after some quota ID exceeds the soft limt,
1041 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1043 #define LQUOTA_GRACE_BITS 48
1044 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1045 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1046 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1047 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1048 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1050 /* different quota flags */
1052 /* the default quota flag, the corresponding quota ID will use the default
1053 * quota setting, the hardlimit and softlimit of its quota record in the global
1054 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1055 * and high 16 bits will contain this flag (see above comment).
1057 #define LQUOTA_FLAG_DEFAULT 0x0001
1059 #define ALLQUOTA 255 /* set all quota */
1060 static inline char *qtype_name(int qtype)
1073 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1074 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1077 #define N_PERMS_MAX 64
1079 struct perm_downcall_data {
1085 struct identity_downcall_data {
1092 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1093 __u32 idd_groups[0];
1096 struct sepol_downcall_data {
1098 time_t sdd_sepol_mtime;
1099 __u16 sdd_sepol_len;
1103 #ifdef NEED_QUOTA_DEFS
1105 #define QIF_BLIMITS 1
1107 #define QIF_ILIMITS 4
1108 #define QIF_INODES 8
1109 #define QIF_BTIME 16
1110 #define QIF_ITIME 32
1111 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1112 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1113 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1114 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1117 #endif /* !__KERNEL__ */
1119 /* lustre volatile file support
1120 * file name header: ".^L^S^T^R:volatile"
1122 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1123 #define LUSTRE_VOLATILE_HDR_LEN 14
1125 enum lustre_quota_version {
1129 /* XXX: same as if_dqinfo struct in kernel */
1137 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1139 __u64 dqb_bhardlimit;
1140 __u64 dqb_bsoftlimit;
1142 __u64 dqb_ihardlimit;
1143 __u64 dqb_isoftlimit;
1144 __u64 dqb_curinodes;
1158 struct if_quotactl {
1165 struct obd_dqinfo qc_dqinfo;
1166 struct obd_dqblk qc_dqblk;
1168 struct obd_uuid obd_uuid;
1171 /* swap layout flags */
1172 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1173 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1174 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1175 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1176 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1178 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1179 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1180 struct lustre_swap_layouts {
1188 /** Bit-mask of valid attributes */
1189 /* The LA_* flags are written to disk as part of the ChangeLog records
1190 * so they are part of the on-disk and network protocol, and cannot be changed.
1191 * Only the first 12 bits are currently saved.
1194 LA_ATIME = 1 << 0, /* 0x00001 */
1195 LA_MTIME = 1 << 1, /* 0x00002 */
1196 LA_CTIME = 1 << 2, /* 0x00004 */
1197 LA_SIZE = 1 << 3, /* 0x00008 */
1198 LA_MODE = 1 << 4, /* 0x00010 */
1199 LA_UID = 1 << 5, /* 0x00020 */
1200 LA_GID = 1 << 6, /* 0x00040 */
1201 LA_BLOCKS = 1 << 7, /* 0x00080 */
1202 LA_TYPE = 1 << 8, /* 0x00100 */
1203 LA_FLAGS = 1 << 9, /* 0x00200 */
1204 LA_NLINK = 1 << 10, /* 0x00400 */
1205 LA_RDEV = 1 << 11, /* 0x00800 */
1206 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1207 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1208 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1209 LA_PROJID = 1 << 15, /* 0x08000 */
1210 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1211 LA_LSIZE = 1 << 17, /* 0x20000 */
1212 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1214 * Attributes must be transmitted to OST objects
1216 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1219 #define MDS_FMODE_READ 00000001
1220 #define MDS_FMODE_WRITE 00000002
1222 #define MDS_FMODE_CLOSED 00000000
1223 #define MDS_FMODE_EXEC 00000004
1224 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1225 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1226 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1228 #define MDS_OPEN_CREATED 00000010
1229 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1231 #define MDS_OPEN_CREAT 00000100
1232 #define MDS_OPEN_EXCL 00000200
1233 #define MDS_OPEN_TRUNC 00001000
1234 #define MDS_OPEN_APPEND 00002000
1235 #define MDS_OPEN_SYNC 00010000
1236 #define MDS_OPEN_DIRECTORY 00200000
1238 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1239 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1240 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1241 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1242 * We do not support JOIN FILE
1243 * anymore, reserve this flags
1244 * just for preventing such bit
1247 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1248 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1249 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1250 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1251 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1253 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1255 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1256 * delegation, succeed if it's not
1257 * being opened with conflict mode.
1259 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1261 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1262 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1263 * for newly created file */
1265 /* lustre internal open flags, which should not be set from user space */
1266 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1267 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1268 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1269 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1273 /********* Changelogs **********/
1274 /** Changelog record types */
1275 enum changelog_rec_type {
1278 CL_CREATE = 1, /* namespace */
1279 CL_MKDIR = 2, /* namespace */
1280 CL_HARDLINK = 3, /* namespace */
1281 CL_SOFTLINK = 4, /* namespace */
1282 CL_MKNOD = 5, /* namespace */
1283 CL_UNLINK = 6, /* namespace */
1284 CL_RMDIR = 7, /* namespace */
1285 CL_RENAME = 8, /* namespace */
1286 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1287 CL_OPEN = 10, /* not currently used */
1288 CL_CLOSE = 11, /* may be written to log only with mtime change */
1289 CL_LAYOUT = 12, /* file layout/striping modified */
1293 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1294 CL_HSM = 16, /* HSM specific events, see flags */
1295 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1299 CL_FLRW = 21, /* FLR: file was firstly written */
1300 CL_RESYNC = 22, /* FLR: file was resync-ed */
1302 CL_DN_OPEN = 24, /* denied open */
1306 static inline const char *changelog_type2str(int type) {
1307 static const char *changelog_str[] = {
1308 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1309 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1310 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1311 "FLRW", "RESYNC","GXATR", "NOPEN",
1314 if (type >= 0 && type < CL_LAST)
1315 return changelog_str[type];
1319 /* 12 bits of per-record data can be stored in the bottom of the flags */
1320 #define CLF_FLAGSHIFT 12
1321 enum changelog_rec_flags {
1322 CLF_VERSION = 0x1000,
1323 CLF_RENAME = 0x2000,
1325 CLF_EXTRA_FLAGS = 0x8000,
1326 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1328 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1329 CLF_VERMASK = ~CLF_FLAGMASK,
1333 /* Anything under the flagmask may be per-type (if desired) */
1334 /* Flags for unlink */
1335 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1336 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1337 /* HSM cleaning needed */
1338 /* Flags for rename */
1339 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1341 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1342 * has an archive in backend */
1345 /* 12b used (from high weight to low weight):
1350 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1351 #define CLF_HSM_ERR_H 6
1352 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1353 #define CLF_HSM_EVENT_H 9
1354 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1355 #define CLF_HSM_FLAG_H 11
1356 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1357 #define CLF_HSM_SPARE_H 15
1358 #define CLF_HSM_LAST 15
1360 /* Remove bits higher than _h, then extract the value
1361 * between _h and _l by shifting lower weigth to bit 0. */
1362 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1363 >> (CLF_HSM_LAST - _h + _l))
1365 #define CLF_HSM_SUCCESS 0x00
1366 #define CLF_HSM_MAXERROR 0x7E
1367 #define CLF_HSM_ERROVERFLOW 0x7F
1369 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1371 /* 3 bits field => 8 values allowed */
1383 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1385 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1389 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1392 *clf_flags |= (he << CLF_HSM_EVENT_L);
1395 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1397 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1400 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1403 *clf_flags |= (bits << CLF_HSM_FLAG_L);
1406 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1408 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1411 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1414 *clf_flags |= (error << CLF_HSM_ERR_L);
1417 enum changelog_rec_extra_flags {
1419 CLFE_UIDGID = 0x0001,
1422 CLFE_XATTR = 0x0008,
1423 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1426 enum changelog_send_flag {
1427 /* Not yet implemented */
1428 CHANGELOG_FLAG_FOLLOW = 0x01,
1429 /* Blocking IO makes sense in case of slow user parsing of the records,
1430 * but it also prevents us from cleaning up if the records are not
1432 CHANGELOG_FLAG_BLOCK = 0x02,
1433 /* Pack jobid into the changelog records if available. */
1434 CHANGELOG_FLAG_JOBID = 0x04,
1435 /* Pack additional flag bits into the changelog record */
1436 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1439 enum changelog_send_extra_flag {
1440 /* Pack uid/gid into the changelog record */
1441 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1442 /* Pack nid into the changelog record */
1443 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1444 /* Pack open mode into the changelog record */
1445 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1446 /* Pack xattr name into the changelog record */
1447 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1450 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1451 changelog_rec_offset(CLF_SUPPORTED, \
1454 /* 31 usable bytes string + null terminator. */
1455 #define LUSTRE_JOBID_SIZE 32
1457 /* This is the minimal changelog record. It can contain extensions
1458 * such as rename fields or process jobid. Its exact content is described
1459 * by the cr_flags and cr_extra_flags.
1461 * Extensions are packed in the same order as their corresponding flags,
1462 * then in the same order as their corresponding extra flags.
1464 struct changelog_rec {
1466 __u16 cr_flags; /**< \a changelog_rec_flags */
1467 __u32 cr_type; /**< \a changelog_rec_type */
1468 __u64 cr_index; /**< changelog record number */
1469 __u64 cr_prev; /**< last index for this target fid */
1472 struct lu_fid cr_tfid; /**< target fid */
1473 __u32 cr_markerflags; /**< CL_MARK flags */
1475 struct lu_fid cr_pfid; /**< parent fid */
1478 /* Changelog extension for RENAME. */
1479 struct changelog_ext_rename {
1480 struct lu_fid cr_sfid; /**< source fid, or zero */
1481 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1484 /* Changelog extension to include JOBID. */
1485 struct changelog_ext_jobid {
1486 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1489 /* Changelog extension to include additional flags. */
1490 struct changelog_ext_extra_flags {
1491 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1494 /* Changelog extra extension to include UID/GID. */
1495 struct changelog_ext_uidgid {
1500 /* Changelog extra extension to include NID. */
1501 struct changelog_ext_nid {
1502 /* have __u64 instead of lnet_nid_t type for use by client api */
1504 /* for use when IPv6 support is added */
1509 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1510 struct changelog_ext_openmode {
1514 /* Changelog extra extension to include xattr */
1515 struct changelog_ext_xattr {
1516 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1519 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1520 const struct changelog_rec *rec);
1522 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1523 enum changelog_rec_extra_flags cref)
1525 size_t size = sizeof(struct changelog_rec);
1527 if (crf & CLF_RENAME)
1528 size += sizeof(struct changelog_ext_rename);
1530 if (crf & CLF_JOBID)
1531 size += sizeof(struct changelog_ext_jobid);
1533 if (crf & CLF_EXTRA_FLAGS) {
1534 size += sizeof(struct changelog_ext_extra_flags);
1535 if (cref & CLFE_UIDGID)
1536 size += sizeof(struct changelog_ext_uidgid);
1537 if (cref & CLFE_NID)
1538 size += sizeof(struct changelog_ext_nid);
1539 if (cref & CLFE_OPEN)
1540 size += sizeof(struct changelog_ext_openmode);
1541 if (cref & CLFE_XATTR)
1542 size += sizeof(struct changelog_ext_xattr);
1548 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1550 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1552 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1553 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1555 return changelog_rec_offset(rec->cr_flags, cref);
1558 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1560 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1564 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1566 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1568 return (struct changelog_ext_rename *)((char *)rec +
1569 changelog_rec_offset(crf,
1573 /* The jobid follows the rename extension, if present */
1575 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1577 enum changelog_rec_flags crf = rec->cr_flags &
1578 (CLF_VERSION | CLF_RENAME);
1580 return (struct changelog_ext_jobid *)((char *)rec +
1581 changelog_rec_offset(crf,
1585 /* The additional flags follow the rename and jobid extensions, if present */
1587 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1588 const struct changelog_rec *rec)
1590 enum changelog_rec_flags crf = rec->cr_flags &
1591 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1593 return (struct changelog_ext_extra_flags *)((char *)rec +
1594 changelog_rec_offset(crf,
1598 /* The uid/gid is the first extra extension */
1600 struct changelog_ext_uidgid *changelog_rec_uidgid(
1601 const struct changelog_rec *rec)
1603 enum changelog_rec_flags crf = rec->cr_flags &
1604 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1606 return (struct changelog_ext_uidgid *)((char *)rec +
1607 changelog_rec_offset(crf,
1611 /* The nid is the second extra extension */
1613 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1615 enum changelog_rec_flags crf = rec->cr_flags &
1616 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1617 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1619 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1620 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1623 return (struct changelog_ext_nid *)((char *)rec +
1624 changelog_rec_offset(crf, cref));
1627 /* The OPEN mode is the third extra extension */
1629 struct changelog_ext_openmode *changelog_rec_openmode(
1630 const struct changelog_rec *rec)
1632 enum changelog_rec_flags crf = rec->cr_flags &
1633 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1634 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1636 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1637 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1638 (CLFE_UIDGID | CLFE_NID);
1640 return (struct changelog_ext_openmode *)((char *)rec +
1641 changelog_rec_offset(crf, cref));
1644 /* The xattr name is the fourth extra extension */
1646 struct changelog_ext_xattr *changelog_rec_xattr(
1647 const struct changelog_rec *rec)
1649 enum changelog_rec_flags crf = rec->cr_flags &
1650 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1651 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1653 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1654 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1655 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN);
1657 return (struct changelog_ext_xattr *)((char *)rec +
1658 changelog_rec_offset(crf, cref));
1661 /* The name follows the rename, jobid and extra flags extns, if present */
1662 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1664 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1666 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1667 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1669 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1670 cref & CLFE_SUPPORTED);
1673 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1675 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1678 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1680 char *cr_name = changelog_rec_name(rec);
1682 return cr_name + strlen(cr_name) + 1;
1686 * Remap a record to the desired format as specified by the crf flags.
1687 * The record must be big enough to contain the final remapped version.
1688 * Superfluous extension fields are removed and missing ones are added
1689 * and zeroed. The flags of the record are updated accordingly.
1691 * The jobid and rename extensions can be added to a record, to match the
1692 * format an application expects, typically. In this case, the newly added
1693 * fields will be zeroed.
1694 * The Jobid field can be removed, to guarantee compatibility with older
1695 * clients that don't expect this field in the records they process.
1697 * The following assumptions are being made:
1698 * - CLF_RENAME will not be removed
1699 * - CLF_JOBID will not be added without CLF_RENAME being added too
1700 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1702 * @param[in,out] rec The record to remap.
1703 * @param[in] crf_wanted Flags describing the desired extensions.
1704 * @param[in] cref_want Flags describing the desired extra extensions.
1706 static inline void changelog_remap_rec(struct changelog_rec *rec,
1707 enum changelog_rec_flags crf_wanted,
1708 enum changelog_rec_extra_flags cref_want)
1710 char *xattr_mov = NULL;
1711 char *omd_mov = NULL;
1712 char *nid_mov = NULL;
1713 char *uidgid_mov = NULL;
1717 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1719 crf_wanted &= CLF_SUPPORTED;
1720 cref_want &= CLFE_SUPPORTED;
1722 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1723 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1724 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1725 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1731 /* First move the variable-length name field */
1732 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1733 changelog_rec_name(rec), rec->cr_namelen);
1735 /* Locations of extensions in the remapped record */
1736 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1737 xattr_mov = (char *)rec +
1738 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1739 cref_want & ~CLFE_XATTR);
1740 omd_mov = (char *)rec +
1741 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1742 cref_want & ~(CLFE_OPEN |
1744 nid_mov = (char *)rec +
1745 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1746 cref_want & ~(CLFE_NID |
1749 uidgid_mov = (char *)rec +
1750 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1751 cref_want & ~(CLFE_UIDGID |
1755 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1758 ef_mov = (char *)rec +
1759 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1761 jid_mov = (char *)rec +
1762 changelog_rec_offset(crf_wanted &
1763 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1765 rnm_mov = (char *)rec +
1766 changelog_rec_offset(crf_wanted &
1772 /* Move the extension fields to the desired positions */
1773 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1774 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1775 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1776 memmove(xattr_mov, changelog_rec_xattr(rec),
1777 sizeof(struct changelog_ext_xattr));
1779 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1780 memmove(omd_mov, changelog_rec_openmode(rec),
1781 sizeof(struct changelog_ext_openmode));
1783 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1784 memmove(nid_mov, changelog_rec_nid(rec),
1785 sizeof(struct changelog_ext_nid));
1787 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1788 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1789 sizeof(struct changelog_ext_uidgid));
1791 memmove(ef_mov, changelog_rec_extra_flags(rec),
1792 sizeof(struct changelog_ext_extra_flags));
1795 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1796 memmove(jid_mov, changelog_rec_jobid(rec),
1797 sizeof(struct changelog_ext_jobid));
1799 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1800 memmove(rnm_mov, changelog_rec_rename(rec),
1801 sizeof(struct changelog_ext_rename));
1803 /* Clear newly added fields */
1804 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1805 !(cref & CLFE_XATTR))
1806 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1808 if (omd_mov && (cref_want & CLFE_OPEN) &&
1809 !(cref & CLFE_OPEN))
1810 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1812 if (nid_mov && (cref_want & CLFE_NID) &&
1814 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1816 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1817 !(cref & CLFE_UIDGID))
1818 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1820 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1821 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1822 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1824 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1825 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1827 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1828 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1830 /* Update the record's flags accordingly */
1831 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1832 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1833 changelog_rec_extra_flags(rec)->cr_extra_flags =
1834 changelog_rec_extra_flags(rec)->cr_extra_flags |
1838 enum changelog_message_type {
1839 CL_RECORD = 10, /* message is a changelog_rec */
1840 CL_EOF = 11, /* at end of current changelog */
1843 /********* Misc **********/
1845 struct ioc_data_version {
1847 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1848 __u32 idv_flags; /* enum ioc_data_version_flags */
1851 enum ioc_data_version_flags {
1852 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1853 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1857 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1860 #define dot_lustre_name ".lustre"
1863 /********* HSM **********/
1865 /** HSM per-file state
1866 * See HSM_FLAGS below.
1869 HS_NONE = 0x00000000,
1870 HS_EXISTS = 0x00000001,
1871 HS_DIRTY = 0x00000002,
1872 HS_RELEASED = 0x00000004,
1873 HS_ARCHIVED = 0x00000008,
1874 HS_NORELEASE = 0x00000010,
1875 HS_NOARCHIVE = 0x00000020,
1876 HS_LOST = 0x00000040,
1879 /* HSM user-setable flags. */
1880 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1882 /* Other HSM flags. */
1883 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1886 * All HSM-related possible flags that could be applied to a file.
1887 * This should be kept in sync with hsm_states.
1889 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1892 * HSM request progress state
1894 enum hsm_progress_states {
1901 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1904 case HPS_WAITING: return "waiting";
1905 case HPS_RUNNING: return "running";
1906 case HPS_DONE: return "done";
1907 default: return "unknown";
1914 } __attribute__((packed));
1917 * Current HSM states of a Lustre file.
1919 * This structure purpose is to be sent to user-space mainly. It describes the
1920 * current HSM flags and in-progress action.
1922 struct hsm_user_state {
1923 /** Current HSM states, from enum hsm_states. */
1925 __u32 hus_archive_id;
1926 /** The current undergoing action, if there is one */
1927 __u32 hus_in_progress_state;
1928 __u32 hus_in_progress_action;
1929 struct hsm_extent hus_in_progress_location;
1930 char hus_extended_info[];
1933 struct hsm_state_set_ioc {
1934 struct lu_fid hssi_fid;
1936 __u64 hssi_clearmask;
1940 * This structure describes the current in-progress action for a file.
1941 * it is retuned to user space and send over the wire
1943 struct hsm_current_action {
1944 /** The current undergoing action, if there is one */
1945 /* state is one of hsm_progress_states */
1947 /* action is one of hsm_user_action */
1949 struct hsm_extent hca_location;
1952 /***** HSM user requests ******/
1953 /* User-generated (lfs/ioctl) request types */
1954 enum hsm_user_action {
1955 HUA_NONE = 1, /* no action (noop) */
1956 HUA_ARCHIVE = 10, /* copy to hsm */
1957 HUA_RESTORE = 11, /* prestage */
1958 HUA_RELEASE = 12, /* drop ost objects */
1959 HUA_REMOVE = 13, /* remove from archive */
1960 HUA_CANCEL = 14 /* cancel a request */
1963 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1966 case HUA_NONE: return "NOOP";
1967 case HUA_ARCHIVE: return "ARCHIVE";
1968 case HUA_RESTORE: return "RESTORE";
1969 case HUA_RELEASE: return "RELEASE";
1970 case HUA_REMOVE: return "REMOVE";
1971 case HUA_CANCEL: return "CANCEL";
1972 default: return "UNKNOWN";
1977 * List of hr_flags (bit field)
1979 #define HSM_FORCE_ACTION 0x0001
1980 /* used by CT, cannot be set by user */
1981 #define HSM_GHOST_COPY 0x0002
1984 * Contains all the fixed part of struct hsm_user_request.
1987 struct hsm_request {
1988 __u32 hr_action; /* enum hsm_user_action */
1989 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1990 __u64 hr_flags; /* request flags */
1991 __u32 hr_itemcount; /* item count in hur_user_item vector */
1995 struct hsm_user_item {
1996 struct lu_fid hui_fid;
1997 struct hsm_extent hui_extent;
1998 } __attribute__((packed));
2000 struct hsm_user_request {
2001 struct hsm_request hur_request;
2002 struct hsm_user_item hur_user_item[0];
2003 /* extra data blob at end of struct (after all
2004 * hur_user_items), only use helpers to access it
2006 } __attribute__((packed));
2008 /** Return pointer to data field in a hsm user request */
2009 static inline void *hur_data(struct hsm_user_request *hur)
2011 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2015 * Compute the current length of the provided hsm_user_request. This returns -1
2016 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
2018 * return -1 on bounds check error.
2020 static inline ssize_t hur_len(struct hsm_user_request *hur)
2024 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2025 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2026 (__u64)hur->hur_request.hr_itemcount *
2027 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2029 if (size != (ssize_t)size)
2035 /****** HSM RPCs to copytool *****/
2036 /* Message types the copytool may receive */
2037 enum hsm_message_type {
2038 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2041 /* Actions the copytool may be instructed to take for a given action_item */
2042 enum hsm_copytool_action {
2043 HSMA_NONE = 10, /* no action */
2044 HSMA_ARCHIVE = 20, /* arbitrary offset */
2050 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2053 case HSMA_NONE: return "NOOP";
2054 case HSMA_ARCHIVE: return "ARCHIVE";
2055 case HSMA_RESTORE: return "RESTORE";
2056 case HSMA_REMOVE: return "REMOVE";
2057 case HSMA_CANCEL: return "CANCEL";
2058 default: return "UNKNOWN";
2062 /* Copytool item action description */
2063 struct hsm_action_item {
2064 __u32 hai_len; /* valid size of this struct */
2065 __u32 hai_action; /* hsm_copytool_action, but use known size */
2066 struct lu_fid hai_fid; /* Lustre FID to operate on */
2067 struct lu_fid hai_dfid; /* fid used for data access */
2068 struct hsm_extent hai_extent; /* byte range to operate on */
2069 __u64 hai_cookie; /* action cookie from coordinator */
2070 __u64 hai_gid; /* grouplock id */
2071 char hai_data[0]; /* variable length */
2072 } __attribute__((packed));
2075 * helper function which print in hexa the first bytes of
2078 * \param hai [IN] record to print
2079 * \param buffer [IN,OUT] buffer to write the hex string to
2080 * \param len [IN] max buffer length
2084 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2085 char *buffer, size_t len)
2092 data_len = hai->hai_len - sizeof(*hai);
2093 for (i = 0; (i < data_len) && (len > 2); i++) {
2094 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2104 /* Copytool action list */
2105 #define HAL_VERSION 1
2106 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2107 struct hsm_action_list {
2109 __u32 hal_count; /* number of hai's to follow */
2110 __u64 hal_compound_id; /* returned by coordinator, ignored */
2112 __u32 hal_archive_id; /* which archive backend */
2114 char hal_fsname[0]; /* null-terminated */
2115 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2116 boundaries. See hai_zero */
2117 } __attribute__((packed));
2119 /* Return pointer to first hai in action list */
2120 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2122 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2124 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2127 /* Return pointer to next hai */
2128 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2130 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2132 return (struct hsm_action_item *)((char *)hai + offset);
2135 /* Return size of an hsm_action_list */
2136 static inline size_t hal_size(struct hsm_action_list *hal)
2140 struct hsm_action_item *hai;
2142 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2143 hai = hai_first(hal);
2144 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2145 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2151 * describe the attributes to be set on imported file
2153 struct hsm_user_import {
2162 __u32 hui_archive_id;
2165 /* Copytool progress reporting */
2166 #define HP_FLAG_COMPLETED 0x01
2167 #define HP_FLAG_RETRY 0x02
2169 struct hsm_progress {
2170 struct lu_fid hp_fid;
2172 struct hsm_extent hp_extent;
2174 __u16 hp_errval; /* positive val */
2179 __u64 hc_data_version;
2181 __u16 hc_errval; /* positive val */
2183 struct hsm_action_item hc_hai;
2187 enum llapi_json_types {
2188 LLAPI_JSON_INTEGER = 1,
2194 struct llapi_json_item {
2203 struct llapi_json_item *lji_next;
2206 struct llapi_json_item_list {
2207 int ljil_item_count;
2208 struct llapi_json_item *ljil_items;
2211 enum lu_ladvise_type {
2212 LU_LADVISE_INVALID = 0,
2213 LU_LADVISE_WILLREAD = 1,
2214 LU_LADVISE_DONTNEED = 2,
2215 LU_LADVISE_LOCKNOEXPAND = 3,
2216 LU_LADVISE_LOCKAHEAD = 4,
2220 #define LU_LADVISE_NAMES { \
2221 [LU_LADVISE_WILLREAD] = "willread", \
2222 [LU_LADVISE_DONTNEED] = "dontneed", \
2223 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2224 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2227 /* This is the userspace argument for ladvise. It is currently the same as
2228 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2229 * may need info which is only used locally. */
2230 struct llapi_lu_ladvise {
2231 __u16 lla_advice; /* advice type */
2232 __u16 lla_value1; /* values for different advice types */
2234 __u64 lla_start; /* first byte of extent for advice */
2235 __u64 lla_end; /* last byte of extent for advice */
2241 LF_ASYNC = 0x00000001,
2242 LF_UNSET = 0x00000002,
2245 #define LADVISE_MAGIC 0x1ADF1CE0
2246 /* Masks of valid flags for each advice */
2247 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2248 /* Flags valid for all advices not explicitly specified */
2249 #define LF_DEFAULT_MASK LF_ASYNC
2251 #define LF_MASK (LF_ASYNC | LF_UNSET)
2253 #define lla_lockahead_mode lla_value1
2254 #define lla_peradvice_flags lla_value2
2255 #define lla_lockahead_result lla_value3
2257 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2258 * is used on the wire. It is defined separately as we may need info which is
2259 * only used locally. */
2260 struct llapi_ladvise_hdr {
2261 __u32 lah_magic; /* LADVISE_MAGIC */
2262 __u32 lah_count; /* number of advices */
2263 __u64 lah_flags; /* from enum ladvise_flag */
2264 __u32 lah_value1; /* unused */
2265 __u32 lah_value2; /* unused */
2266 __u64 lah_value3; /* unused */
2267 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2270 #define LAH_COUNT_MAX (1024)
2274 SK_CRYPT_INVALID = -1,
2276 SK_CRYPT_AES256_CTR = 1,
2280 SK_HMAC_INVALID = -1,
2286 struct sk_crypt_type {
2287 const char *sct_name;
2291 struct sk_hmac_type {
2292 const char *sht_name;
2296 enum lock_mode_user {
2302 #define LOCK_MODE_NAMES { \
2303 [MODE_READ_USER] = "READ",\
2304 [MODE_WRITE_USER] = "WRITE"\
2307 enum lockahead_results {
2308 LLA_RESULT_SENT = 0,
2309 LLA_RESULT_DIFFERENT,
2313 enum lu_heat_flag_bit {
2314 LU_HEAT_FLAG_BIT_INVALID = 0,
2315 LU_HEAT_FLAG_BIT_OFF,
2316 LU_HEAT_FLAG_BIT_CLEAR,
2320 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2321 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2324 enum obd_heat_type {
2325 OBD_HEAT_READSAMPLE = 0,
2326 OBD_HEAT_WRITESAMPLE = 1,
2327 OBD_HEAT_READBYTE = 2,
2328 OBD_HEAT_WRITEBYTE = 3,
2332 #define LU_HEAT_NAMES { \
2333 [OBD_HEAT_READSAMPLE] = "readsample", \
2334 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2335 [OBD_HEAT_READBYTE] = "readbyte", \
2336 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2351 static inline const char *pcc_type2string(enum lu_pcc_type type)
2356 case LU_PCC_READWRITE:
2363 struct lu_pcc_attach {
2364 __u32 pcca_type; /* PCC type */
2365 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2368 enum lu_pcc_detach_opts {
2369 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2370 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2373 struct lu_pcc_detach_fid {
2374 /* fid of the file to detach */
2375 struct lu_fid pccd_fid;
2379 struct lu_pcc_detach {
2383 enum lu_pcc_state_flags {
2384 PCC_STATE_FL_NONE = 0x0,
2385 /* The inode attr is cached locally */
2386 PCC_STATE_FL_ATTR_VALID = 0x01,
2387 /* The file is being attached into PCC */
2388 PCC_STATE_FL_ATTACHING = 0x02,
2389 /* Allow to auto attach at open */
2390 PCC_STATE_FL_OPEN_ATTACH = 0x04,
2393 struct lu_pcc_state {
2394 __u32 pccs_type; /* enum lu_pcc_type */
2395 __u32 pccs_open_count;
2396 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2398 char pccs_path[PATH_MAX];
2401 #if defined(__cplusplus)
2405 /** @} lustreuser */
2407 #endif /* _LUSTRE_USER_H */