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/fs.h>
47 # include <linux/quota.h>
48 # include <linux/sched/signal.h>
49 # include <linux/string.h> /* snprintf() */
50 # include <linux/version.h>
51 #else /* !__KERNEL__ */
52 # define NEED_QUOTA_DEFS
53 # include <stdio.h> /* snprintf() */
55 # include <sys/quota.h>
56 # include <sys/stat.h>
57 #endif /* __KERNEL__ */
58 #include <uapi/linux/lustre/lustre_fiemap.h>
61 * We need to always use 64bit version because the structure
62 * is shared across entire cluster where 32bit and 64bit machines
65 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
66 typedef struct stat64 lstat_t;
67 #define lstat_f lstat64
68 #define fstat_f fstat64
69 #define fstatat_f fstatat64
71 typedef struct stat lstat_t;
74 #define fstatat_f fstatat
77 #define HAVE_LOV_USER_MDS_DATA
79 #define LUSTRE_EOF 0xffffffffffffffffULL
82 #define LL_SUPER_MAGIC 0x0BD00BD0
84 #ifndef FSFILT_IOC_GETFLAGS
85 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
86 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
87 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
88 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
89 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
90 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
93 /* FIEMAP flags supported by Lustre */
94 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
96 enum obd_statfs_state {
97 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
98 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
99 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
100 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
101 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
115 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
116 __u32 os_fprecreated; /* objs available now to the caller */
117 /* used in QoS code to find preferred OSTs */
131 * FID is a cluster-wide unique identifier of a file or an object (stripe).
132 * FIDs are never reused.
136 * FID sequence. Sequence is a unit of migration: all files (objects)
137 * with FIDs from a given sequence are stored on the same server.
138 * Lustre should support 2^64 objects, so even if each sequence
139 * has only a single object we can still enumerate 2^64 objects.
142 /* FID number within sequence. */
145 * FID version, used to distinguish different versions (in the sense
146 * of snapshots, etc.) of the same file system object. Not currently
152 static inline bool fid_is_zero(const struct lu_fid *fid)
154 return !fid->f_seq && !fid->f_oid;
158 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
161 /* keep this one for compatibility */
162 struct filter_fid_old {
163 struct lu_fid ff_parent;
168 /* Userspace should treat lu_fid as opaque, and only use the following methods
169 * to print or parse them. Other functions (e.g. compare, swab) could be moved
170 * here from lustre_idl.h if needed.
175 * Following struct for object attributes, that will be kept inode's EA.
176 * Introduced in 2.0 release (please see b15993, for details)
177 * Added to all objects since Lustre 2.4 as contains self FID
179 struct lustre_mdt_attrs {
181 * Bitfield for supported data in this structure. From enum lma_compat.
182 * lma_self_fid and lma_flags are always available.
186 * Per-file incompat feature list. Lustre version should support all
187 * flags set in this field. The supported feature mask is available in
191 /** FID of this inode */
192 struct lu_fid lma_self_fid;
196 * Prior to 2.4, the LMA structure also included SOM attributes which has since
197 * been moved to a dedicated xattr
198 * lma_flags was also removed because of lma_compat/incompat fields.
200 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
203 * OST object IDentifier.
211 struct lu_fid oi_fid;
215 #define DOSTID "%#llx:%llu"
216 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
219 * The ioctl naming rules:
220 * LL_* - works on the currently opened filehandle instead of parent dir
221 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
222 * *_MDC_* - gets/sets data related to MDC
223 * *_LOV_* - gets/sets data related to OSC/LOV
224 * *FILE* - called on parent dir and passes in a filename
225 * *STRIPE* - set/get lov_user_md
226 * *INFO - set/get lov_user_mds_data
228 /* lustre_ioctl.h 101-150 */
229 #define LL_IOC_GETFLAGS _IOR('f', 151, long)
230 #define LL_IOC_SETFLAGS _IOW('f', 152, long)
231 #define LL_IOC_CLRFLAGS _IOW('f', 153, long)
232 #define LL_IOC_LOV_SETSTRIPE _IOW('f', 154, long)
233 #define LL_IOC_LOV_GETSTRIPE _IOW('f', 155, long)
234 #define LL_IOC_LOV_SETEA _IOW('f', 156, long)
235 /* LL_IOC_RECREATE_OBJ 157 obsolete */
236 /* LL_IOC_RECREATE_FID 158 obsolete */
237 #define LL_IOC_GROUP_LOCK _IOW('f', 158, long)
238 #define LL_IOC_GROUP_UNLOCK _IOW('f', 159, long)
239 /* #define LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
240 /* #define LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
241 /* #define LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
242 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
243 /* IOC_LOV_GETINFO 165 obsolete */
244 #define LL_IOC_FLUSHCTX _IOW('f', 166, long)
245 /* LL_IOC_RMTACL 167 obsolete */
246 #define LL_IOC_GETOBDCOUNT _IOR('f', 168, long)
247 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
248 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
249 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
250 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
251 #define LL_IOC_PATH2FID _IOR('f', 173, long)
252 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
253 #define LL_IOC_GET_MDTIDX _IOR('f', 175, int)
255 /* lustre_ioctl.h 177-210 */
256 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
257 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
258 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
259 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
260 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
261 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
262 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
263 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
264 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
265 struct lustre_swap_layouts)
266 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
267 struct hsm_current_action)
268 /* see <lustre_lib.h> for ioctl numbers 221-232 */
270 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
271 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
272 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
273 #define LL_IOC_GET_LEASE _IO('f', 244)
274 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
275 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
276 #define LL_IOC_MIGRATE _IOR('f', 247, int)
277 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
278 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
280 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
282 LL_LEASE_RDLCK = 0x1,
283 LL_LEASE_WRLCK = 0x2,
284 LL_LEASE_UNLCK = 0x4,
287 #define LL_STATFS_LMV 1
288 #define LL_STATFS_LOV 2
289 #define LL_STATFS_NODELAY 4
291 #define IOC_MDC_TYPE 'i'
292 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
293 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
294 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
295 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
297 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
299 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
300 * files, but are unlikely to be used in practice and are not harmful if
301 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
302 * devices and are safe for use on new files (See LU-812, LU-4209).
304 #define O_LOV_DELAY_CREATE (O_NOCTTY | FASYNC)
306 #define LL_FILE_IGNORE_LOCK 0x00000001
307 #define LL_FILE_GROUP_LOCKED 0x00000002
308 #define LL_FILE_READAHEA 0x00000004
309 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
310 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
311 #define LL_FILE_RMTACL 0x00000020
313 #define LOV_USER_MAGIC_V1 0x0BD10BD0
314 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
315 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
316 #define LOV_USER_MAGIC_V3 0x0BD30BD0
317 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
318 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
320 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
322 #define LOV_PATTERN_RAID0 0x001
323 #define LOV_PATTERN_RAID1 0x002
324 #define LOV_PATTERN_FIRST 0x100
325 #define LOV_PATTERN_CMOBD 0x200
327 #define LOV_PATTERN_F_MASK 0xffff0000
328 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
329 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
331 #define LOV_MAXPOOLNAME 15
332 #define LOV_POOLNAMEF "%.15s"
334 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
335 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
336 #define LOV_MAX_STRIPE_COUNT_OLD 160
337 /* This calculation is crafted so that input of 4096 will result in 160
338 * which in turn is equal to old maximal stripe count.
339 * XXX: In fact this is too simplified for now, what it also need is to get
340 * ea_type argument to clearly know how much space each stripe consumes.
342 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
343 * allocation that is sufficient for the current generation of systems.
345 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1)
347 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
348 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
349 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
351 #define XATTR_LUSTRE_PREFIX "lustre."
352 #define XATTR_LUSTRE_LOV "lustre.lov"
354 #define lov_user_ost_data lov_user_ost_data_v1
355 struct lov_user_ost_data_v1 { /* per-stripe data structure */
356 struct ost_id l_ost_oi; /* OST object ID */
357 __u32 l_ost_gen; /* generation of this OST index */
358 __u32 l_ost_idx; /* OST index in LOV */
361 #define lov_user_md lov_user_md_v1
362 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
363 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
364 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
365 struct ost_id lmm_oi; /* LOV object ID */
366 __u32 lmm_stripe_size; /* size of stripe in bytes */
367 __u16 lmm_stripe_count; /* num stripes in use for this object */
369 __u16 lmm_stripe_offset; /* starting stripe offset in
370 * lmm_objects, use when writing
372 __u16 lmm_layout_gen; /* layout generation number
376 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
377 } __attribute__((packed, __may_alias__));
379 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
380 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
381 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
382 struct ost_id lmm_oi; /* LOV object ID */
383 __u32 lmm_stripe_size; /* size of stripe in bytes */
384 __u16 lmm_stripe_count; /* num stripes in use for this object */
386 __u16 lmm_stripe_offset; /* starting stripe offset in
387 * lmm_objects, use when writing
389 __u16 lmm_layout_gen; /* layout generation number
393 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
394 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
397 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
399 if (lmm_magic == LOV_USER_MAGIC_V1)
400 return sizeof(struct lov_user_md_v1) +
401 stripes * sizeof(struct lov_user_ost_data_v1);
402 return sizeof(struct lov_user_md_v3) +
403 stripes * sizeof(struct lov_user_ost_data_v1);
406 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
407 * use this. It is unsafe to #define those values in this header as it
408 * is possible the application has already #included <sys/stat.h>. */
409 #ifdef HAVE_LOV_USER_MDS_DATA
410 #define lov_user_mds_data lov_user_mds_data_v1
411 struct lov_user_mds_data_v1 {
412 lstat_t lmd_st; /* MDS stat struct */
413 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
416 struct lov_user_mds_data_v3 {
417 lstat_t lmd_st; /* MDS stat struct */
418 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
422 struct lmv_user_mds_data {
423 struct lu_fid lum_fid;
429 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
430 LMV_HASH_TYPE_ALL_CHARS = 1,
431 LMV_HASH_TYPE_FNV_1A_64 = 2,
434 #define LMV_HASH_NAME_ALL_CHARS "all_char"
435 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
438 * Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
439 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data)
441 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
442 #define lmv_user_md lmv_user_md_v1
443 struct lmv_user_md_v1 {
444 __u32 lum_magic; /* must be the first field */
445 __u32 lum_stripe_count; /* dirstripe count */
446 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
447 __u32 lum_hash_type; /* Dir stripe policy */
448 __u32 lum_type; /* LMV type: default or normal */
452 char lum_pool_name[LOV_MAXPOOLNAME + 1];
453 struct lmv_user_mds_data lum_objects[0];
456 static inline int lmv_user_md_size(int stripes, int lmm_magic)
458 return sizeof(struct lmv_user_md) +
459 stripes * sizeof(struct lmv_user_mds_data);
462 struct ll_recreate_obj {
468 __u64 id; /* holds object id */
469 __u32 generation; /* holds object generation */
470 __u32 f_type; /* holds object type or stripe idx when passing it to
471 * OST for saving into EA. */
479 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
480 const struct obd_uuid *u2)
482 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
485 static inline int obd_uuid_empty(struct obd_uuid *uuid)
487 return uuid->uuid[0] == '\0';
490 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
492 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
493 uuid->uuid[sizeof(*uuid) - 1] = '\0';
496 /* For printf's only, make sure uuid is terminated */
497 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
502 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
503 /* Obviously not safe, but for printfs, no real harm done...
504 * we're always null-terminated, even in a race.
506 static char temp[sizeof(*uuid)];
508 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
509 temp[sizeof(*uuid) - 1] = '\0';
512 return (char *)(uuid->uuid);
515 /* Extract fsname from uuid (or target name) of a target
516 * e.g. (myfs-OST0007_UUID -> myfs)
517 * see also deuuidify.
519 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
523 strncpy(buf, uuid, buflen - 1);
524 buf[buflen - 1] = '\0';
525 p = strrchr(buf, '-');
530 /* printf display format
531 * * usage: printf("file FID is "DFID"\n", PFID(fid));
533 #define FID_NOBRACE_LEN 40
534 #define FID_LEN (FID_NOBRACE_LEN + 2)
535 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
536 #define DFID "[" DFID_NOBRACE "]"
537 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
539 /* scanf input parse format for fids in DFID_NOBRACE format
540 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
541 * usage: sscanf(fidstr, SFID, RFID(&fid));
543 #define SFID "0x%llx:0x%x:0x%x"
544 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
546 /********* Quotas **********/
548 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
549 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
550 #define Q_GETOINFO 0x800102 /* get obd quota info */
551 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
552 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
554 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
555 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
556 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
557 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
558 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
559 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
560 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
561 /* lustre-specific control commands */
562 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
563 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
565 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
567 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
570 #define N_PERMS_MAX 64
572 struct perm_downcall_data {
578 struct identity_downcall_data {
585 struct perm_downcall_data idd_perms[N_PERMS_MAX];
589 /* lustre volatile file support
590 * file name header: .^L^S^T^R:volatile"
592 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
593 #define LUSTRE_VOLATILE_HDR_LEN 14
594 /* hdr + MDT index */
595 #define LUSTRE_VOLATILE_IDX LUSTRE_VOLATILE_HDR":%.4X:"
597 enum lustre_quota_version {
601 /* XXX: same as if_dqinfo struct in kernel */
609 /* XXX: same as if_dqblk struct in kernel, plus one padding */
611 __u64 dqb_bhardlimit;
612 __u64 dqb_bsoftlimit;
614 __u64 dqb_ihardlimit;
615 __u64 dqb_isoftlimit;
637 struct obd_dqinfo qc_dqinfo;
638 struct obd_dqblk qc_dqblk;
640 struct obd_uuid obd_uuid;
643 /* swap layout flags */
644 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
645 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
646 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
647 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
648 #define SWAP_LAYOUTS_CLOSE (1 << 4)
650 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
651 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
652 struct lustre_swap_layouts {
660 /********* Changelogs **********/
661 /** Changelog record types */
662 enum changelog_rec_type {
664 CL_CREATE = 1, /* namespace */
665 CL_MKDIR = 2, /* namespace */
666 CL_HARDLINK = 3, /* namespace */
667 CL_SOFTLINK = 4, /* namespace */
668 CL_MKNOD = 5, /* namespace */
669 CL_UNLINK = 6, /* namespace */
670 CL_RMDIR = 7, /* namespace */
671 CL_RENAME = 8, /* namespace */
672 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
673 CL_OPEN = 10, /* not currently used */
674 CL_CLOSE = 11, /* may be written to log only with mtime change */
675 CL_LAYOUT = 12, /* file layout/striping modified */
679 CL_HSM = 16, /* HSM specific events, see flags */
680 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
686 static inline const char *changelog_type2str(int type)
688 static const char *changelog_str[] = {
689 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
690 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
691 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
694 if (type >= 0 && type < CL_LAST)
695 return changelog_str[type];
699 /* per-record flags */
700 #define CLF_FLAGSHIFT 12
701 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
702 #define CLF_VERMASK (~CLF_FLAGMASK)
703 enum changelog_rec_flags {
704 CLF_VERSION = 0x1000,
707 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
710 /* Anything under the flagmask may be per-type (if desired) */
711 /* Flags for unlink */
712 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
713 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
714 /* HSM cleaning needed */
715 /* Flags for rename */
716 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink of
719 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
720 * has an archive in backend
724 /* 12b used (from high weight to low weight):
729 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
730 #define CLF_HSM_ERR_H 6
731 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
732 #define CLF_HSM_EVENT_H 9
733 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
734 #define CLF_HSM_FLAG_H 11
735 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
736 #define CLF_HSM_SPARE_H 15
737 #define CLF_HSM_LAST 15
739 /* Remove bits higher than _h, then extract the value
740 * between _h and _l by shifting lower weigth to bit 0.
742 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
743 >> (CLF_HSM_LAST - _h + _l))
745 #define CLF_HSM_SUCCESS 0x00
746 #define CLF_HSM_MAXERROR 0x7E
747 #define CLF_HSM_ERROVERFLOW 0x7F
749 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
751 /* 3 bits field => 8 values allowed */
763 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
765 return CLF_GET_BITS(flags, CLF_HSM_EVENT_H, CLF_HSM_EVENT_L);
768 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
770 *flags |= (he << CLF_HSM_EVENT_L);
773 static inline __u16 hsm_get_cl_flags(int flags)
775 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
778 static inline void hsm_set_cl_flags(int *flags, int bits)
780 *flags |= (bits << CLF_HSM_FLAG_L);
783 static inline int hsm_get_cl_error(int flags)
785 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
788 static inline void hsm_set_cl_error(int *flags, int error)
790 *flags |= (error << CLF_HSM_ERR_L);
793 enum changelog_send_flag {
794 /* Not yet implemented */
795 CHANGELOG_FLAG_FOLLOW = 0x01,
797 * Blocking IO makes sense in case of slow user parsing of the records,
798 * but it also prevents us from cleaning up if the records are not
801 CHANGELOG_FLAG_BLOCK = 0x02,
802 /* Pack jobid into the changelog records if available. */
803 CHANGELOG_FLAG_JOBID = 0x04,
806 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
807 changelog_rec_offset(CLF_SUPPORTED))
809 /* 31 usable bytes string + null terminator. */
810 #define LUSTRE_JOBID_SIZE 32
813 * This is the minimal changelog record. It can contain extensions
814 * such as rename fields or process jobid. Its exact content is described
817 * Extensions are packed in the same order as their corresponding flags.
819 struct changelog_rec {
821 __u16 cr_flags; /**< \a changelog_rec_flags */
822 __u32 cr_type; /**< \a changelog_rec_type */
823 __u64 cr_index; /**< changelog record number */
824 __u64 cr_prev; /**< last index for this target fid */
827 struct lu_fid cr_tfid; /**< target fid */
828 __u32 cr_markerflags; /**< CL_MARK flags */
830 struct lu_fid cr_pfid; /**< parent fid */
833 /* Changelog extension for RENAME. */
834 struct changelog_ext_rename {
835 struct lu_fid cr_sfid; /**< source fid, or zero */
836 struct lu_fid cr_spfid; /**< source parent fid, or zero */
839 /* Changelog extension to include JOBID. */
840 struct changelog_ext_jobid {
841 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
844 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
846 size_t size = sizeof(struct changelog_rec);
848 if (crf & CLF_RENAME)
849 size += sizeof(struct changelog_ext_rename);
852 size += sizeof(struct changelog_ext_jobid);
857 static inline size_t changelog_rec_size(struct changelog_rec *rec)
859 return changelog_rec_offset(rec->cr_flags);
862 static inline size_t changelog_rec_varsize(struct changelog_rec *rec)
864 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
868 struct changelog_ext_rename *changelog_rec_rename(struct changelog_rec *rec)
870 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
872 return (struct changelog_ext_rename *)((char *)rec +
873 changelog_rec_offset(crf));
876 /* The jobid follows the rename extension, if present */
878 struct changelog_ext_jobid *changelog_rec_jobid(struct changelog_rec *rec)
880 enum changelog_rec_flags crf = rec->cr_flags &
881 (CLF_VERSION | CLF_RENAME);
883 return (struct changelog_ext_jobid *)((char *)rec +
884 changelog_rec_offset(crf));
887 /* The name follows the rename and jobid extensions, if present */
888 static inline char *changelog_rec_name(struct changelog_rec *rec)
890 return (char *)rec + changelog_rec_offset(rec->cr_flags &
894 static inline size_t changelog_rec_snamelen(struct changelog_rec *rec)
896 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
899 static inline char *changelog_rec_sname(struct changelog_rec *rec)
901 char *cr_name = changelog_rec_name(rec);
903 return cr_name + strlen(cr_name) + 1;
907 * Remap a record to the desired format as specified by the crf flags.
908 * The record must be big enough to contain the final remapped version.
909 * Superfluous extension fields are removed and missing ones are added
910 * and zeroed. The flags of the record are updated accordingly.
912 * The jobid and rename extensions can be added to a record, to match the
913 * format an application expects, typically. In this case, the newly added
914 * fields will be zeroed.
915 * The Jobid field can be removed, to guarantee compatibility with older
916 * clients that don't expect this field in the records they process.
918 * The following assumptions are being made:
919 * - CLF_RENAME will not be removed
920 * - CLF_JOBID will not be added without CLF_RENAME being added too
922 * @param[in,out] rec The record to remap.
923 * @param[in] crf_wanted Flags describing the desired extensions.
925 static inline void changelog_remap_rec(struct changelog_rec *rec,
926 enum changelog_rec_flags crf_wanted)
928 char *jid_mov, *rnm_mov;
930 crf_wanted &= CLF_SUPPORTED;
932 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
935 /* First move the variable-length name field */
936 memmove((char *)rec + changelog_rec_offset(crf_wanted),
937 changelog_rec_name(rec), rec->cr_namelen);
939 /* Locations of jobid and rename extensions in the remapped record */
940 jid_mov = (char *)rec +
941 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
942 rnm_mov = (char *)rec +
943 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
945 /* Move the extension fields to the desired positions */
946 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
947 memmove(jid_mov, changelog_rec_jobid(rec),
948 sizeof(struct changelog_ext_jobid));
950 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
951 memmove(rnm_mov, changelog_rec_rename(rec),
952 sizeof(struct changelog_ext_rename));
954 /* Clear newly added fields */
955 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
956 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
958 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
959 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
961 /* Update the record's flags accordingly */
962 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
965 struct ioc_changelog {
972 enum changelog_message_type {
973 CL_RECORD = 10, /* message is a changelog_rec */
974 CL_EOF = 11, /* at end of current changelog */
977 /********* Misc **********/
979 struct ioc_data_version {
981 __u64 idv_flags; /* See LL_DV_xxx */
984 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
985 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
988 # define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
991 #define dot_lustre_name ".lustre"
993 /********* HSM **********/
995 /** HSM per-file state
996 * See HSM_FLAGS below.
999 HS_NONE = 0x00000000,
1000 HS_EXISTS = 0x00000001,
1001 HS_DIRTY = 0x00000002,
1002 HS_RELEASED = 0x00000004,
1003 HS_ARCHIVED = 0x00000008,
1004 HS_NORELEASE = 0x00000010,
1005 HS_NOARCHIVE = 0x00000020,
1006 HS_LOST = 0x00000040,
1009 /* HSM user-setable flags. */
1010 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1012 /* Other HSM flags. */
1013 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1016 * All HSM-related possible flags that could be applied to a file.
1017 * This should be kept in sync with hsm_states.
1019 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1022 * HSMÂ request progress state
1024 enum hsm_progress_states {
1032 static inline char *hsm_progress_state2name(enum hsm_progress_states s)
1035 case HPS_WAITING: return "waiting";
1036 case HPS_RUNNING: return "running";
1037 case HPS_DONE: return "done";
1038 default: return "unknown";
1048 * Current HSM states of a Lustre file.
1050 * This structure purpose is to be sent to user-space mainly. It describes the
1051 * current HSM flags and in-progress action.
1053 struct hsm_user_state {
1054 /** Current HSM states, from enum hsm_states. */
1056 __u32 hus_archive_id;
1057 /** The current undergoing action, if there is one */
1058 __u32 hus_in_progress_state;
1059 __u32 hus_in_progress_action;
1060 struct hsm_extent hus_in_progress_location;
1061 char hus_extended_info[];
1064 struct hsm_state_set_ioc {
1065 struct lu_fid hssi_fid;
1067 __u64 hssi_clearmask;
1071 * This structure describes the current in-progress action for a file.
1072 * it is returned to user space and send over the wire
1074 struct hsm_current_action {
1075 /** The current undergoing action, if there is one */
1076 /* state is one of hsm_progress_states */
1078 /* action is one of hsm_user_action */
1080 struct hsm_extent hca_location;
1083 /***** HSM user requests ******/
1084 /* User-generated (lfs/ioctl) request types */
1085 enum hsm_user_action {
1086 HUA_NONE = 1, /* no action (noop) */
1087 HUA_ARCHIVE = 10, /* copy to hsm */
1088 HUA_RESTORE = 11, /* prestage */
1089 HUA_RELEASE = 12, /* drop ost objects */
1090 HUA_REMOVE = 13, /* remove from archive */
1091 HUA_CANCEL = 14 /* cancel a request */
1094 static inline char *hsm_user_action2name(enum hsm_user_action a)
1097 case HUA_NONE: return "NOOP";
1098 case HUA_ARCHIVE: return "ARCHIVE";
1099 case HUA_RESTORE: return "RESTORE";
1100 case HUA_RELEASE: return "RELEASE";
1101 case HUA_REMOVE: return "REMOVE";
1102 case HUA_CANCEL: return "CANCEL";
1103 default: return "UNKNOWN";
1108 * List of hr_flags (bit field)
1110 #define HSM_FORCE_ACTION 0x0001
1111 /* used by CT, connot be set by user */
1112 #define HSM_GHOST_COPY 0x0002
1115 * Contains all the fixed part of struct hsm_user_request.
1118 struct hsm_request {
1119 __u32 hr_action; /* enum hsm_user_action */
1120 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1121 __u64 hr_flags; /* request flags */
1122 __u32 hr_itemcount; /* item count in hur_user_item vector */
1126 struct hsm_user_item {
1127 struct lu_fid hui_fid;
1128 struct hsm_extent hui_extent;
1131 struct hsm_user_request {
1132 struct hsm_request hur_request;
1133 struct hsm_user_item hur_user_item[0];
1134 /* extra data blob at end of struct (after all
1135 * hur_user_items), only use helpers to access it
1139 /** Return pointer to data field in a hsm user request */
1140 static inline void *hur_data(struct hsm_user_request *hur)
1142 return &hur->hur_user_item[hur->hur_request.hr_itemcount];
1146 * Compute the current length of the provided hsm_user_request. This returns -1
1147 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1149 * return -1 on bounds check error.
1151 static inline ssize_t hur_len(struct hsm_user_request *hur)
1155 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1156 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1157 (__u64)hur->hur_request.hr_itemcount *
1158 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1160 if (size != (ssize_t)size)
1166 /****** HSM RPCs to copytool *****/
1167 /* Message types the copytool may receive */
1168 enum hsm_message_type {
1169 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1172 /* Actions the copytool may be instructed to take for a given action_item */
1173 enum hsm_copytool_action {
1174 HSMA_NONE = 10, /* no action */
1175 HSMA_ARCHIVE = 20, /* arbitrary offset */
1181 static inline char *hsm_copytool_action2name(enum hsm_copytool_action a)
1184 case HSMA_NONE: return "NOOP";
1185 case HSMA_ARCHIVE: return "ARCHIVE";
1186 case HSMA_RESTORE: return "RESTORE";
1187 case HSMA_REMOVE: return "REMOVE";
1188 case HSMA_CANCEL: return "CANCEL";
1189 default: return "UNKNOWN";
1193 /* Copytool item action description */
1194 struct hsm_action_item {
1195 __u32 hai_len; /* valid size of this struct */
1196 __u32 hai_action; /* hsm_copytool_action, but use known size */
1197 struct lu_fid hai_fid; /* Lustre FID to operated on */
1198 struct lu_fid hai_dfid; /* fid used for data access */
1199 struct hsm_extent hai_extent; /* byte range to operate on */
1200 __u64 hai_cookie; /* action cookie from coordinator */
1201 __u64 hai_gid; /* grouplock id */
1202 char hai_data[0]; /* variable length */
1206 * helper function which print in hexa the first bytes of
1208 * \param hai [IN] record to print
1209 * \param buffer [OUT] output buffer
1210 * \param len [IN] max buffer len
1213 static inline char *hai_dump_data_field(struct hsm_action_item *hai,
1214 char *buffer, size_t len)
1220 data_len = hai->hai_len - sizeof(*hai);
1221 for (i = 0; (i < data_len) && (len > 2); i++) {
1222 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1232 /* Copytool action list */
1233 #define HAL_VERSION 1
1234 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1235 struct hsm_action_list {
1237 __u32 hal_count; /* number of hai's to follow */
1238 __u64 hal_compound_id; /* returned by coordinator */
1240 __u32 hal_archive_id; /* which archive backend */
1242 char hal_fsname[0]; /* null-terminated */
1243 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1244 * boundaries. See hai_first
1248 #ifndef HAVE_CFS_SIZE_ROUND
1249 static inline int cfs_size_round(int val)
1251 return (val + 7) & (~0x7);
1254 #define HAVE_CFS_SIZE_ROUND
1257 /* Return pointer to first hai in action list */
1258 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1260 return (struct hsm_action_item *)(hal->hal_fsname +
1261 cfs_size_round(strlen(hal-> \
1266 /* Return pointer to next hai */
1267 static inline struct hsm_action_item *hai_next(struct hsm_action_item *hai)
1269 return (struct hsm_action_item *)((char *)hai +
1270 cfs_size_round(hai->hai_len));
1273 /* Return size of an hsm_action_list */
1274 static inline int hal_size(struct hsm_action_list *hal)
1277 struct hsm_action_item *hai;
1279 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1280 hai = hai_first(hal);
1281 for (i = 0; i < hal->hal_count; i++, hai = hai_next(hai))
1282 sz += cfs_size_round(hai->hai_len);
1288 * describe the attributes to be set on imported file
1290 struct hsm_user_import {
1299 __u32 hui_archive_id;
1302 /* Copytool progress reporting */
1303 #define HP_FLAG_COMPLETED 0x01
1304 #define HP_FLAG_RETRY 0x02
1306 struct hsm_progress {
1307 struct lu_fid hp_fid;
1309 struct hsm_extent hp_extent;
1311 __u16 hp_errval; /* positive val */
1316 __u64 hc_data_version;
1318 __u16 hc_errval; /* positive val */
1320 struct hsm_action_item hc_hai;
1323 /** @} lustreuser */
1325 #endif /* _LUSTRE_USER_H */