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, 2015, 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
46 # include <linux/quota.h>
47 # include <linux/sched/signal.h>
48 # include <linux/string.h> /* snprintf() */
49 # include <linux/version.h>
50 #else /* !__KERNEL__ */
51 # define NEED_QUOTA_DEFS
52 # include <stdio.h> /* snprintf() */
54 # include <sys/quota.h>
55 # include <sys/stat.h>
56 #endif /* __KERNEL__ */
57 #include "ll_fiemap.h"
60 * We need to always use 64bit version because the structure
61 * is shared across entire cluster where 32bit and 64bit machines
64 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
65 typedef struct stat64 lstat_t
;
66 #define lstat_f lstat64
67 #define fstat_f fstat64
68 #define fstatat_f fstatat64
70 typedef struct stat lstat_t
;
73 #define fstatat_f fstatat
76 #define HAVE_LOV_USER_MDS_DATA
78 #define LUSTRE_EOF 0xffffffffffffffffULL
81 #define LL_SUPER_MAGIC 0x0BD00BD0
83 #ifndef FSFILT_IOC_GETFLAGS
84 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
85 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
86 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
87 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
88 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
89 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
92 /* FIEMAP flags supported by Lustre */
93 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
95 enum obd_statfs_state
{
96 OS_STATE_DEGRADED
= 0x00000001, /**< RAID degraded/rebuilding */
97 OS_STATE_READONLY
= 0x00000002, /**< filesystem is read-only */
98 OS_STATE_RDONLY_1
= 0x00000004, /**< obsolete 1.6, was EROFS=30 */
99 OS_STATE_RDONLY_2
= 0x00000008, /**< obsolete 1.6, was EROFS=30 */
100 OS_STATE_RDONLY_3
= 0x00000010, /**< obsolete 1.6, was EROFS=30 */
114 __u32 os_state
; /**< obd_statfs_state OS_STATE_* flag */
115 __u32 os_fprecreated
; /* objs available now to the caller */
116 /* used in QoS code to find preferred OSTs */
130 * FID is a cluster-wide unique identifier of a file or an object (stripe).
131 * FIDs are never reused.
135 * FID sequence. Sequence is a unit of migration: all files (objects)
136 * with FIDs from a given sequence are stored on the same server.
137 * Lustre should support 2^64 objects, so even if each sequence
138 * has only a single object we can still enumerate 2^64 objects.
141 /* FID number within sequence. */
144 * FID version, used to distinguish different versions (in the sense
145 * of snapshots, etc.) of the same file system object. Not currently
151 static inline bool fid_is_zero(const struct lu_fid
*fid
)
153 return !fid
->f_seq
&& !fid
->f_oid
;
157 struct lu_fid ff_parent
; /* ff_parent.f_ver == file stripe number */
160 /* keep this one for compatibility */
161 struct filter_fid_old
{
162 struct lu_fid ff_parent
;
167 /* Userspace should treat lu_fid as opaque, and only use the following methods
168 * to print or parse them. Other functions (e.g. compare, swab) could be moved
169 * here from lustre_idl.h if needed.
174 * Following struct for object attributes, that will be kept inode's EA.
175 * Introduced in 2.0 release (please see b15993, for details)
176 * Added to all objects since Lustre 2.4 as contains self FID
178 struct lustre_mdt_attrs
{
180 * Bitfield for supported data in this structure. From enum lma_compat.
181 * lma_self_fid and lma_flags are always available.
185 * Per-file incompat feature list. Lustre version should support all
186 * flags set in this field. The supported feature mask is available in
190 /** FID of this inode */
191 struct lu_fid lma_self_fid
;
195 * Prior to 2.4, the LMA structure also included SOM attributes which has since
196 * been moved to a dedicated xattr
197 * lma_flags was also removed because of lma_compat/incompat fields.
199 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
202 * OST object IDentifier.
210 struct lu_fid oi_fid
;
214 #define DOSTID "%#llx:%llu"
215 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
218 * The ioctl naming rules:
219 * LL_* - works on the currently opened filehandle instead of parent dir
220 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
221 * *_MDC_* - gets/sets data related to MDC
222 * *_LOV_* - gets/sets data related to OSC/LOV
223 * *FILE* - called on parent dir and passes in a filename
224 * *STRIPE* - set/get lov_user_md
225 * *INFO - set/get lov_user_mds_data
227 /* lustre_ioctl.h 101-150 */
228 #define LL_IOC_GETFLAGS _IOR('f', 151, long)
229 #define LL_IOC_SETFLAGS _IOW('f', 152, long)
230 #define LL_IOC_CLRFLAGS _IOW('f', 153, long)
231 #define LL_IOC_LOV_SETSTRIPE _IOW('f', 154, long)
232 #define LL_IOC_LOV_GETSTRIPE _IOW('f', 155, long)
233 #define LL_IOC_LOV_SETEA _IOW('f', 156, long)
234 /* LL_IOC_RECREATE_OBJ 157 obsolete */
235 /* LL_IOC_RECREATE_FID 158 obsolete */
236 #define LL_IOC_GROUP_LOCK _IOW('f', 158, long)
237 #define LL_IOC_GROUP_UNLOCK _IOW('f', 159, long)
238 /* #define LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
239 /* #define LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
240 /* #define LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
241 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
242 /* IOC_LOV_GETINFO 165 obsolete */
243 #define LL_IOC_FLUSHCTX _IOW('f', 166, long)
244 /* LL_IOC_RMTACL 167 obsolete */
245 #define LL_IOC_GETOBDCOUNT _IOR('f', 168, long)
246 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
247 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
248 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
249 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
250 #define LL_IOC_PATH2FID _IOR('f', 173, long)
251 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
252 #define LL_IOC_GET_MDTIDX _IOR('f', 175, int)
254 /* lustre_ioctl.h 177-210 */
255 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
256 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
257 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
258 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
259 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
260 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
261 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
262 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
263 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
264 struct lustre_swap_layouts)
265 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
266 struct hsm_current_action)
267 /* see <lustre_lib.h> for ioctl numbers 221-232 */
269 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
270 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
271 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
272 #define LL_IOC_GET_LEASE _IO('f', 244)
273 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
274 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
275 #define LL_IOC_MIGRATE _IOR('f', 247, int)
276 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
277 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
279 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
281 LL_LEASE_RDLCK
= 0x1,
282 LL_LEASE_WRLCK
= 0x2,
283 LL_LEASE_UNLCK
= 0x4,
286 #define LL_STATFS_LMV 1
287 #define LL_STATFS_LOV 2
288 #define LL_STATFS_NODELAY 4
290 #define IOC_MDC_TYPE 'i'
291 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
292 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
293 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
294 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
296 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
298 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
299 * files, but are unlikely to be used in practice and are not harmful if
300 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
301 * devices and are safe for use on new files (See LU-812, LU-4209).
303 #define O_LOV_DELAY_CREATE (O_NOCTTY | FASYNC)
305 #define LL_FILE_IGNORE_LOCK 0x00000001
306 #define LL_FILE_GROUP_LOCKED 0x00000002
307 #define LL_FILE_READAHEA 0x00000004
308 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
309 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
310 #define LL_FILE_RMTACL 0x00000020
312 #define LOV_USER_MAGIC_V1 0x0BD10BD0
313 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
314 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
315 #define LOV_USER_MAGIC_V3 0x0BD30BD0
316 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
317 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
319 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
321 #define LOV_PATTERN_RAID0 0x001
322 #define LOV_PATTERN_RAID1 0x002
323 #define LOV_PATTERN_FIRST 0x100
324 #define LOV_PATTERN_CMOBD 0x200
326 #define LOV_PATTERN_F_MASK 0xffff0000
327 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
328 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
330 #define LOV_MAXPOOLNAME 15
331 #define LOV_POOLNAMEF "%.15s"
333 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
334 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
335 #define LOV_MAX_STRIPE_COUNT_OLD 160
336 /* This calculation is crafted so that input of 4096 will result in 160
337 * which in turn is equal to old maximal stripe count.
338 * XXX: In fact this is too simplified for now, what it also need is to get
339 * ea_type argument to clearly know how much space each stripe consumes.
341 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
342 * allocation that is sufficient for the current generation of systems.
344 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1)
346 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
347 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
348 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
350 #define XATTR_LUSTRE_PREFIX "lustre."
351 #define XATTR_LUSTRE_LOV "lustre.lov"
353 #define lov_user_ost_data lov_user_ost_data_v1
354 struct lov_user_ost_data_v1
{ /* per-stripe data structure */
355 struct ost_id l_ost_oi
; /* OST object ID */
356 __u32 l_ost_gen
; /* generation of this OST index */
357 __u32 l_ost_idx
; /* OST index in LOV */
360 #define lov_user_md lov_user_md_v1
361 struct lov_user_md_v1
{ /* LOV EA user data (host-endian) */
362 __u32 lmm_magic
; /* magic number = LOV_USER_MAGIC_V1 */
363 __u32 lmm_pattern
; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
364 struct ost_id lmm_oi
; /* LOV object ID */
365 __u32 lmm_stripe_size
; /* size of stripe in bytes */
366 __u16 lmm_stripe_count
; /* num stripes in use for this object */
368 __u16 lmm_stripe_offset
; /* starting stripe offset in
369 * lmm_objects, use when writing
371 __u16 lmm_layout_gen
; /* layout generation number
375 struct lov_user_ost_data_v1 lmm_objects
[0]; /* per-stripe data */
376 } __attribute__((packed
, __may_alias__
));
378 struct lov_user_md_v3
{ /* LOV EA user data (host-endian) */
379 __u32 lmm_magic
; /* magic number = LOV_USER_MAGIC_V3 */
380 __u32 lmm_pattern
; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
381 struct ost_id lmm_oi
; /* LOV object ID */
382 __u32 lmm_stripe_size
; /* size of stripe in bytes */
383 __u16 lmm_stripe_count
; /* num stripes in use for this object */
385 __u16 lmm_stripe_offset
; /* starting stripe offset in
386 * lmm_objects, use when writing
388 __u16 lmm_layout_gen
; /* layout generation number
392 char lmm_pool_name
[LOV_MAXPOOLNAME
+ 1]; /* pool name */
393 struct lov_user_ost_data_v1 lmm_objects
[0]; /* per-stripe data */
396 static inline __u32
lov_user_md_size(__u16 stripes
, __u32 lmm_magic
)
398 if (lmm_magic
== LOV_USER_MAGIC_V1
)
399 return sizeof(struct lov_user_md_v1
) +
400 stripes
* sizeof(struct lov_user_ost_data_v1
);
401 return sizeof(struct lov_user_md_v3
) +
402 stripes
* sizeof(struct lov_user_ost_data_v1
);
405 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
406 * use this. It is unsafe to #define those values in this header as it
407 * is possible the application has already #included <sys/stat.h>. */
408 #ifdef HAVE_LOV_USER_MDS_DATA
409 #define lov_user_mds_data lov_user_mds_data_v1
410 struct lov_user_mds_data_v1
{
411 lstat_t lmd_st
; /* MDS stat struct */
412 struct lov_user_md_v1 lmd_lmm
; /* LOV EA V1 user data */
415 struct lov_user_mds_data_v3
{
416 lstat_t lmd_st
; /* MDS stat struct */
417 struct lov_user_md_v3 lmd_lmm
; /* LOV EA V3 user data */
421 struct lmv_user_mds_data
{
422 struct lu_fid lum_fid
;
428 LMV_HASH_TYPE_UNKNOWN
= 0, /* 0 is reserved for testing purpose */
429 LMV_HASH_TYPE_ALL_CHARS
= 1,
430 LMV_HASH_TYPE_FNV_1A_64
= 2,
433 #define LMV_HASH_NAME_ALL_CHARS "all_char"
434 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
437 * Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
438 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data)
440 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
441 #define lmv_user_md lmv_user_md_v1
442 struct lmv_user_md_v1
{
443 __u32 lum_magic
; /* must be the first field */
444 __u32 lum_stripe_count
; /* dirstripe count */
445 __u32 lum_stripe_offset
; /* MDT idx for default dirstripe */
446 __u32 lum_hash_type
; /* Dir stripe policy */
447 __u32 lum_type
; /* LMV type: default or normal */
451 char lum_pool_name
[LOV_MAXPOOLNAME
+ 1];
452 struct lmv_user_mds_data lum_objects
[0];
455 static inline int lmv_user_md_size(int stripes
, int lmm_magic
)
457 return sizeof(struct lmv_user_md
) +
458 stripes
* sizeof(struct lmv_user_mds_data
);
461 struct ll_recreate_obj
{
467 __u64 id
; /* holds object id */
468 __u32 generation
; /* holds object generation */
469 __u32 f_type
; /* holds object type or stripe idx when passing it to
470 * OST for saving into EA. */
478 static inline bool obd_uuid_equals(const struct obd_uuid
*u1
,
479 const struct obd_uuid
*u2
)
481 return strcmp((char *)u1
->uuid
, (char *)u2
->uuid
) == 0;
484 static inline int obd_uuid_empty(struct obd_uuid
*uuid
)
486 return uuid
->uuid
[0] == '\0';
489 static inline void obd_str2uuid(struct obd_uuid
*uuid
, const char *tmp
)
491 strncpy((char *)uuid
->uuid
, tmp
, sizeof(*uuid
));
492 uuid
->uuid
[sizeof(*uuid
) - 1] = '\0';
495 /* For printf's only, make sure uuid is terminated */
496 static inline char *obd_uuid2str(const struct obd_uuid
*uuid
)
501 if (uuid
->uuid
[sizeof(*uuid
) - 1] != '\0') {
502 /* Obviously not safe, but for printfs, no real harm done...
503 * we're always null-terminated, even in a race.
505 static char temp
[sizeof(*uuid
)];
507 memcpy(temp
, uuid
->uuid
, sizeof(*uuid
) - 1);
508 temp
[sizeof(*uuid
) - 1] = '\0';
511 return (char *)(uuid
->uuid
);
514 /* Extract fsname from uuid (or target name) of a target
515 * e.g. (myfs-OST0007_UUID -> myfs)
516 * see also deuuidify.
518 static inline void obd_uuid2fsname(char *buf
, char *uuid
, int buflen
)
522 strncpy(buf
, uuid
, buflen
- 1);
523 buf
[buflen
- 1] = '\0';
524 p
= strrchr(buf
, '-');
529 /* printf display format
530 * * usage: printf("file FID is "DFID"\n", PFID(fid));
532 #define FID_NOBRACE_LEN 40
533 #define FID_LEN (FID_NOBRACE_LEN + 2)
534 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
535 #define DFID "["DFID_NOBRACE"]"
536 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
538 /* scanf input parse format for fids in DFID_NOBRACE format
539 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
540 * usage: sscanf(fidstr, SFID, RFID(&fid));
542 #define SFID "0x%llx:0x%x:0x%x"
543 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
545 /********* Quotas **********/
547 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
548 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
549 #define Q_GETOINFO 0x800102 /* get obd quota info */
550 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
551 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
553 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
554 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
555 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
556 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
557 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
558 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
559 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
560 /* lustre-specific control commands */
561 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
562 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
564 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
566 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
569 #define N_PERMS_MAX 64
571 struct perm_downcall_data
{
577 struct identity_downcall_data
{
584 struct perm_downcall_data idd_perms
[N_PERMS_MAX
];
588 /* lustre volatile file support
589 * file name header: .^L^S^T^R:volatile"
591 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
592 #define LUSTRE_VOLATILE_HDR_LEN 14
593 /* hdr + MDT index */
594 #define LUSTRE_VOLATILE_IDX LUSTRE_VOLATILE_HDR":%.4X:"
596 enum lustre_quota_version
{
600 /* XXX: same as if_dqinfo struct in kernel */
608 /* XXX: same as if_dqblk struct in kernel, plus one padding */
610 __u64 dqb_bhardlimit
;
611 __u64 dqb_bsoftlimit
;
613 __u64 dqb_ihardlimit
;
614 __u64 dqb_isoftlimit
;
636 struct obd_dqinfo qc_dqinfo
;
637 struct obd_dqblk qc_dqblk
;
639 struct obd_uuid obd_uuid
;
642 /* swap layout flags */
643 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
644 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
645 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
646 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
647 #define SWAP_LAYOUTS_CLOSE BIT(4)
649 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
650 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
651 struct lustre_swap_layouts
{
659 /********* Changelogs **********/
660 /** Changelog record types */
661 enum changelog_rec_type
{
663 CL_CREATE
= 1, /* namespace */
664 CL_MKDIR
= 2, /* namespace */
665 CL_HARDLINK
= 3, /* namespace */
666 CL_SOFTLINK
= 4, /* namespace */
667 CL_MKNOD
= 5, /* namespace */
668 CL_UNLINK
= 6, /* namespace */
669 CL_RMDIR
= 7, /* namespace */
670 CL_RENAME
= 8, /* namespace */
671 CL_EXT
= 9, /* namespace extended record (2nd half of rename) */
672 CL_OPEN
= 10, /* not currently used */
673 CL_CLOSE
= 11, /* may be written to log only with mtime change */
674 CL_LAYOUT
= 12, /* file layout/striping modified */
678 CL_HSM
= 16, /* HSM specific events, see flags */
679 CL_MTIME
= 17, /* Precedence: setattr > mtime > ctime > atime */
685 static inline const char *changelog_type2str(int type
)
687 static const char *changelog_str
[] = {
688 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
689 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
690 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
693 if (type
>= 0 && type
< CL_LAST
)
694 return changelog_str
[type
];
698 /* per-record flags */
699 #define CLF_FLAGSHIFT 12
700 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
701 #define CLF_VERMASK (~CLF_FLAGMASK)
702 enum changelog_rec_flags
{
703 CLF_VERSION
= 0x1000,
706 CLF_SUPPORTED
= CLF_VERSION
| CLF_RENAME
| CLF_JOBID
709 /* Anything under the flagmask may be per-type (if desired) */
710 /* Flags for unlink */
711 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
712 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
713 /* HSM cleaning needed */
714 /* Flags for rename */
715 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink of
718 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
719 * has an archive in backend
723 /* 12b used (from high weight to low weight):
728 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
729 #define CLF_HSM_ERR_H 6
730 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
731 #define CLF_HSM_EVENT_H 9
732 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
733 #define CLF_HSM_FLAG_H 11
734 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
735 #define CLF_HSM_SPARE_H 15
736 #define CLF_HSM_LAST 15
738 /* Remove bits higher than _h, then extract the value
739 * between _h and _l by shifting lower weigth to bit 0.
741 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
742 >> (CLF_HSM_LAST - _h + _l))
744 #define CLF_HSM_SUCCESS 0x00
745 #define CLF_HSM_MAXERROR 0x7E
746 #define CLF_HSM_ERROVERFLOW 0x7F
748 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
750 /* 3 bits field => 8 values allowed */
762 static inline enum hsm_event
hsm_get_cl_event(__u16 flags
)
764 return CLF_GET_BITS(flags
, CLF_HSM_EVENT_H
, CLF_HSM_EVENT_L
);
767 static inline void hsm_set_cl_event(int *flags
, enum hsm_event he
)
769 *flags
|= (he
<< CLF_HSM_EVENT_L
);
772 static inline __u16
hsm_get_cl_flags(int flags
)
774 return CLF_GET_BITS(flags
, CLF_HSM_FLAG_H
, CLF_HSM_FLAG_L
);
777 static inline void hsm_set_cl_flags(int *flags
, int bits
)
779 *flags
|= (bits
<< CLF_HSM_FLAG_L
);
782 static inline int hsm_get_cl_error(int flags
)
784 return CLF_GET_BITS(flags
, CLF_HSM_ERR_H
, CLF_HSM_ERR_L
);
787 static inline void hsm_set_cl_error(int *flags
, int error
)
789 *flags
|= (error
<< CLF_HSM_ERR_L
);
792 enum changelog_send_flag
{
793 /* Not yet implemented */
794 CHANGELOG_FLAG_FOLLOW
= BIT(0),
796 * Blocking IO makes sense in case of slow user parsing of the records,
797 * but it also prevents us from cleaning up if the records are not
800 CHANGELOG_FLAG_BLOCK
= BIT(1),
801 /* Pack jobid into the changelog records if available. */
802 CHANGELOG_FLAG_JOBID
= BIT(2),
805 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
806 changelog_rec_offset(CLF_SUPPORTED))
808 /* 31 usable bytes string + null terminator. */
809 #define LUSTRE_JOBID_SIZE 32
812 * This is the minimal changelog record. It can contain extensions
813 * such as rename fields or process jobid. Its exact content is described
816 * Extensions are packed in the same order as their corresponding flags.
818 struct changelog_rec
{
820 __u16 cr_flags
; /**< \a changelog_rec_flags */
821 __u32 cr_type
; /**< \a changelog_rec_type */
822 __u64 cr_index
; /**< changelog record number */
823 __u64 cr_prev
; /**< last index for this target fid */
826 struct lu_fid cr_tfid
; /**< target fid */
827 __u32 cr_markerflags
; /**< CL_MARK flags */
829 struct lu_fid cr_pfid
; /**< parent fid */
832 /* Changelog extension for RENAME. */
833 struct changelog_ext_rename
{
834 struct lu_fid cr_sfid
; /**< source fid, or zero */
835 struct lu_fid cr_spfid
; /**< source parent fid, or zero */
838 /* Changelog extension to include JOBID. */
839 struct changelog_ext_jobid
{
840 char cr_jobid
[LUSTRE_JOBID_SIZE
]; /**< zero-terminated string. */
843 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf
)
845 size_t size
= sizeof(struct changelog_rec
);
847 if (crf
& CLF_RENAME
)
848 size
+= sizeof(struct changelog_ext_rename
);
851 size
+= sizeof(struct changelog_ext_jobid
);
856 static inline size_t changelog_rec_size(struct changelog_rec
*rec
)
858 return changelog_rec_offset(rec
->cr_flags
);
861 static inline size_t changelog_rec_varsize(struct changelog_rec
*rec
)
863 return changelog_rec_size(rec
) - sizeof(*rec
) + rec
->cr_namelen
;
867 struct changelog_ext_rename
*changelog_rec_rename(struct changelog_rec
*rec
)
869 enum changelog_rec_flags crf
= rec
->cr_flags
& CLF_VERSION
;
871 return (struct changelog_ext_rename
*)((char *)rec
+
872 changelog_rec_offset(crf
));
875 /* The jobid follows the rename extension, if present */
877 struct changelog_ext_jobid
*changelog_rec_jobid(struct changelog_rec
*rec
)
879 enum changelog_rec_flags crf
= rec
->cr_flags
&
880 (CLF_VERSION
| CLF_RENAME
);
882 return (struct changelog_ext_jobid
*)((char *)rec
+
883 changelog_rec_offset(crf
));
886 /* The name follows the rename and jobid extensions, if present */
887 static inline char *changelog_rec_name(struct changelog_rec
*rec
)
889 return (char *)rec
+ changelog_rec_offset(rec
->cr_flags
&
893 static inline size_t changelog_rec_snamelen(struct changelog_rec
*rec
)
895 return rec
->cr_namelen
- strlen(changelog_rec_name(rec
)) - 1;
898 static inline char *changelog_rec_sname(struct changelog_rec
*rec
)
900 char *cr_name
= changelog_rec_name(rec
);
902 return cr_name
+ strlen(cr_name
) + 1;
906 * Remap a record to the desired format as specified by the crf flags.
907 * The record must be big enough to contain the final remapped version.
908 * Superfluous extension fields are removed and missing ones are added
909 * and zeroed. The flags of the record are updated accordingly.
911 * The jobid and rename extensions can be added to a record, to match the
912 * format an application expects, typically. In this case, the newly added
913 * fields will be zeroed.
914 * The Jobid field can be removed, to guarantee compatibility with older
915 * clients that don't expect this field in the records they process.
917 * The following assumptions are being made:
918 * - CLF_RENAME will not be removed
919 * - CLF_JOBID will not be added without CLF_RENAME being added too
921 * @param[in,out] rec The record to remap.
922 * @param[in] crf_wanted Flags describing the desired extensions.
924 static inline void changelog_remap_rec(struct changelog_rec
*rec
,
925 enum changelog_rec_flags crf_wanted
)
927 char *jid_mov
, *rnm_mov
;
929 crf_wanted
&= CLF_SUPPORTED
;
931 if ((rec
->cr_flags
& CLF_SUPPORTED
) == crf_wanted
)
934 /* First move the variable-length name field */
935 memmove((char *)rec
+ changelog_rec_offset(crf_wanted
),
936 changelog_rec_name(rec
), rec
->cr_namelen
);
938 /* Locations of jobid and rename extensions in the remapped record */
939 jid_mov
= (char *)rec
+
940 changelog_rec_offset(crf_wanted
& ~CLF_JOBID
);
941 rnm_mov
= (char *)rec
+
942 changelog_rec_offset(crf_wanted
& ~(CLF_JOBID
| CLF_RENAME
));
944 /* Move the extension fields to the desired positions */
945 if ((crf_wanted
& CLF_JOBID
) && (rec
->cr_flags
& CLF_JOBID
))
946 memmove(jid_mov
, changelog_rec_jobid(rec
),
947 sizeof(struct changelog_ext_jobid
));
949 if ((crf_wanted
& CLF_RENAME
) && (rec
->cr_flags
& CLF_RENAME
))
950 memmove(rnm_mov
, changelog_rec_rename(rec
),
951 sizeof(struct changelog_ext_rename
));
953 /* Clear newly added fields */
954 if ((crf_wanted
& CLF_JOBID
) && !(rec
->cr_flags
& CLF_JOBID
))
955 memset(jid_mov
, 0, sizeof(struct changelog_ext_jobid
));
957 if ((crf_wanted
& CLF_RENAME
) && !(rec
->cr_flags
& CLF_RENAME
))
958 memset(rnm_mov
, 0, sizeof(struct changelog_ext_rename
));
960 /* Update the record's flags accordingly */
961 rec
->cr_flags
= (rec
->cr_flags
& CLF_FLAGMASK
) | crf_wanted
;
964 struct ioc_changelog
{
971 enum changelog_message_type
{
972 CL_RECORD
= 10, /* message is a changelog_rec */
973 CL_EOF
= 11, /* at end of current changelog */
976 /********* Misc **********/
978 struct ioc_data_version
{
980 __u64 idv_flags
; /* See LL_DV_xxx */
983 #define LL_DV_RD_FLUSH BIT(0) /* Flush dirty pages from clients */
984 #define LL_DV_WR_FLUSH BIT(1) /* Flush all caching pages from clients */
987 # define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
990 #define dot_lustre_name ".lustre"
992 /********* HSM **********/
994 /** HSM per-file state
995 * See HSM_FLAGS below.
998 HS_NONE
= 0x00000000,
999 HS_EXISTS
= 0x00000001,
1000 HS_DIRTY
= 0x00000002,
1001 HS_RELEASED
= 0x00000004,
1002 HS_ARCHIVED
= 0x00000008,
1003 HS_NORELEASE
= 0x00000010,
1004 HS_NOARCHIVE
= 0x00000020,
1005 HS_LOST
= 0x00000040,
1008 /* HSM user-setable flags. */
1009 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1011 /* Other HSM flags. */
1012 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1015 * All HSM-related possible flags that could be applied to a file.
1016 * This should be kept in sync with hsm_states.
1018 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1021 * HSMÂ request progress state
1023 enum hsm_progress_states
{
1031 static inline char *hsm_progress_state2name(enum hsm_progress_states s
)
1034 case HPS_WAITING
: return "waiting";
1035 case HPS_RUNNING
: return "running";
1036 case HPS_DONE
: return "done";
1037 default: return "unknown";
1047 * Current HSM states of a Lustre file.
1049 * This structure purpose is to be sent to user-space mainly. It describes the
1050 * current HSM flags and in-progress action.
1052 struct hsm_user_state
{
1053 /** Current HSM states, from enum hsm_states. */
1055 __u32 hus_archive_id
;
1056 /** The current undergoing action, if there is one */
1057 __u32 hus_in_progress_state
;
1058 __u32 hus_in_progress_action
;
1059 struct hsm_extent hus_in_progress_location
;
1060 char hus_extended_info
[];
1063 struct hsm_state_set_ioc
{
1064 struct lu_fid hssi_fid
;
1066 __u64 hssi_clearmask
;
1070 * This structure describes the current in-progress action for a file.
1071 * it is returned to user space and send over the wire
1073 struct hsm_current_action
{
1074 /** The current undergoing action, if there is one */
1075 /* state is one of hsm_progress_states */
1077 /* action is one of hsm_user_action */
1079 struct hsm_extent hca_location
;
1082 /***** HSM user requests ******/
1083 /* User-generated (lfs/ioctl) request types */
1084 enum hsm_user_action
{
1085 HUA_NONE
= 1, /* no action (noop) */
1086 HUA_ARCHIVE
= 10, /* copy to hsm */
1087 HUA_RESTORE
= 11, /* prestage */
1088 HUA_RELEASE
= 12, /* drop ost objects */
1089 HUA_REMOVE
= 13, /* remove from archive */
1090 HUA_CANCEL
= 14 /* cancel a request */
1093 static inline char *hsm_user_action2name(enum hsm_user_action a
)
1096 case HUA_NONE
: return "NOOP";
1097 case HUA_ARCHIVE
: return "ARCHIVE";
1098 case HUA_RESTORE
: return "RESTORE";
1099 case HUA_RELEASE
: return "RELEASE";
1100 case HUA_REMOVE
: return "REMOVE";
1101 case HUA_CANCEL
: return "CANCEL";
1102 default: return "UNKNOWN";
1107 * List of hr_flags (bit field)
1109 #define HSM_FORCE_ACTION 0x0001
1110 /* used by CT, connot be set by user */
1111 #define HSM_GHOST_COPY 0x0002
1114 * Contains all the fixed part of struct hsm_user_request.
1117 struct hsm_request
{
1118 __u32 hr_action
; /* enum hsm_user_action */
1119 __u32 hr_archive_id
; /* archive id, used only with HUA_ARCHIVE */
1120 __u64 hr_flags
; /* request flags */
1121 __u32 hr_itemcount
; /* item count in hur_user_item vector */
1125 struct hsm_user_item
{
1126 struct lu_fid hui_fid
;
1127 struct hsm_extent hui_extent
;
1130 struct hsm_user_request
{
1131 struct hsm_request hur_request
;
1132 struct hsm_user_item hur_user_item
[0];
1133 /* extra data blob at end of struct (after all
1134 * hur_user_items), only use helpers to access it
1138 /** Return pointer to data field in a hsm user request */
1139 static inline void *hur_data(struct hsm_user_request
*hur
)
1141 return &hur
->hur_user_item
[hur
->hur_request
.hr_itemcount
];
1145 * Compute the current length of the provided hsm_user_request. This returns -1
1146 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1148 * return -1 on bounds check error.
1150 static inline ssize_t
hur_len(struct hsm_user_request
*hur
)
1154 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1155 size
= offsetof(struct hsm_user_request
, hur_user_item
[0]) +
1156 (__u64
)hur
->hur_request
.hr_itemcount
*
1157 sizeof(hur
->hur_user_item
[0]) + hur
->hur_request
.hr_data_len
;
1159 if (size
!= (ssize_t
)size
)
1165 /****** HSM RPCs to copytool *****/
1166 /* Message types the copytool may receive */
1167 enum hsm_message_type
{
1168 HMT_ACTION_LIST
= 100, /* message is a hsm_action_list */
1171 /* Actions the copytool may be instructed to take for a given action_item */
1172 enum hsm_copytool_action
{
1173 HSMA_NONE
= 10, /* no action */
1174 HSMA_ARCHIVE
= 20, /* arbitrary offset */
1180 static inline char *hsm_copytool_action2name(enum hsm_copytool_action a
)
1183 case HSMA_NONE
: return "NOOP";
1184 case HSMA_ARCHIVE
: return "ARCHIVE";
1185 case HSMA_RESTORE
: return "RESTORE";
1186 case HSMA_REMOVE
: return "REMOVE";
1187 case HSMA_CANCEL
: return "CANCEL";
1188 default: return "UNKNOWN";
1192 /* Copytool item action description */
1193 struct hsm_action_item
{
1194 __u32 hai_len
; /* valid size of this struct */
1195 __u32 hai_action
; /* hsm_copytool_action, but use known size */
1196 struct lu_fid hai_fid
; /* Lustre FID to operated on */
1197 struct lu_fid hai_dfid
; /* fid used for data access */
1198 struct hsm_extent hai_extent
; /* byte range to operate on */
1199 __u64 hai_cookie
; /* action cookie from coordinator */
1200 __u64 hai_gid
; /* grouplock id */
1201 char hai_data
[0]; /* variable length */
1205 * helper function which print in hexa the first bytes of
1207 * \param hai [IN] record to print
1208 * \param buffer [OUT] output buffer
1209 * \param len [IN] max buffer len
1212 static inline char *hai_dump_data_field(struct hsm_action_item
*hai
,
1213 char *buffer
, size_t len
)
1219 data_len
= hai
->hai_len
- sizeof(*hai
);
1220 for (i
= 0; (i
< data_len
) && (len
> 2); i
++) {
1221 snprintf(ptr
, 3, "%02X", (unsigned char)hai
->hai_data
[i
]);
1231 /* Copytool action list */
1232 #define HAL_VERSION 1
1233 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1234 struct hsm_action_list
{
1236 __u32 hal_count
; /* number of hai's to follow */
1237 __u64 hal_compound_id
; /* returned by coordinator */
1239 __u32 hal_archive_id
; /* which archive backend */
1241 char hal_fsname
[0]; /* null-terminated */
1242 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1243 * boundaries. See hai_first
1247 #ifndef HAVE_CFS_SIZE_ROUND
1248 static inline int cfs_size_round(int val
)
1250 return (val
+ 7) & (~0x7);
1253 #define HAVE_CFS_SIZE_ROUND
1256 /* Return pointer to first hai in action list */
1257 static inline struct hsm_action_item
*hai_first(struct hsm_action_list
*hal
)
1259 return (struct hsm_action_item
*)(hal
->hal_fsname
+
1260 cfs_size_round(strlen(hal
-> \
1265 /* Return pointer to next hai */
1266 static inline struct hsm_action_item
*hai_next(struct hsm_action_item
*hai
)
1268 return (struct hsm_action_item
*)((char *)hai
+
1269 cfs_size_round(hai
->hai_len
));
1272 /* Return size of an hsm_action_list */
1273 static inline int hal_size(struct hsm_action_list
*hal
)
1276 struct hsm_action_item
*hai
;
1278 sz
= sizeof(*hal
) + cfs_size_round(strlen(hal
->hal_fsname
) + 1);
1279 hai
= hai_first(hal
);
1280 for (i
= 0; i
< hal
->hal_count
; i
++, hai
= hai_next(hai
))
1281 sz
+= cfs_size_round(hai
->hai_len
);
1287 * describe the attributes to be set on imported file
1289 struct hsm_user_import
{
1298 __u32 hui_archive_id
;
1301 /* Copytool progress reporting */
1302 #define HP_FLAG_COMPLETED 0x01
1303 #define HP_FLAG_RETRY 0x02
1305 struct hsm_progress
{
1306 struct lu_fid hp_fid
;
1308 struct hsm_extent hp_extent
;
1310 __u16 hp_errval
; /* positive val */
1315 __u64 hc_data_version
;
1317 __u16 hc_errval
; /* positive val */
1319 struct hsm_action_item hc_hai
;
1322 /** @} lustreuser */
1324 #endif /* _LUSTRE_USER_H */