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.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2010, 2014, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_user.h
38 * Lustre public user-space interface definitions.
41 #ifndef _LUSTRE_USER_H
42 #define _LUSTRE_USER_H
44 /** \defgroup lustreuser lustreuser
49 #include <libcfs/types.h>
52 # include <linux/string.h> /* snprintf() */
54 # include <stdio.h> /* snprintf() */
56 #include <lustre/ll_fiemap.h>
57 #include <linux/lustre_user.h>
59 #define LUSTRE_EOF 0xffffffffffffffffULL
62 #define LL_SUPER_MAGIC 0x0BD00BD0
64 #ifndef FSFILT_IOC_GETFLAGS
65 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
66 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
67 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
68 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
69 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
70 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
71 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
74 /* FIEMAP flags supported by Lustre */
75 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
77 enum obd_statfs_state {
78 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
79 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
80 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
81 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
82 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
96 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
97 __u32 os_fprecreated; /* objs available now to the caller */
98 /* used in QoS code to find preferred
113 * FID is a cluster-wide unique identifier of a file or an object (stripe).
114 * FIDs are never reused.
118 * FID sequence. Sequence is a unit of migration: all files (objects)
119 * with FIDs from a given sequence are stored on the same server.
120 * Lustre should support 2^64 objects, so even if each sequence
121 * has only a single object we can still enumerate 2^64 objects.
124 /* FID number within sequence. */
127 * FID version, used to distinguish different versions (in the sense
128 * of snapshots, etc.) of the same file system object. Not currently
134 static inline bool fid_is_zero(const struct lu_fid *fid)
136 return fid->f_seq == 0 && fid->f_oid == 0;
139 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
140 * MDT-object's FID::f_ver, instead it is the OST-object index in its
141 * parent MDT-object's layout EA. */
142 #define f_stripe_idx f_ver
145 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
148 /* keep this one for compatibility */
149 struct filter_fid_old {
150 struct lu_fid ff_parent;
155 /* Userspace should treat lu_fid as opaque, and only use the following methods
156 * to print or parse them. Other functions (e.g. compare, swab) could be moved
157 * here from lustre_idl.h if needed. */
158 typedef struct lu_fid lustre_fid;
161 * Following struct for object attributes, that will be kept inode's EA.
162 * Introduced in 2.0 release (please see b15993, for details)
163 * Added to all objects since Lustre 2.4 as contains self FID
165 struct lustre_mdt_attrs {
167 * Bitfield for supported data in this structure. From enum lma_compat.
168 * lma_self_fid and lma_flags are always available.
172 * Per-file incompat feature list. Lustre version should support all
173 * flags set in this field. The supported feature mask is available in
177 /** FID of this inode */
178 struct lu_fid lma_self_fid;
182 * Prior to 2.4, the LMA structure also included SOM attributes which has since
183 * been moved to a dedicated xattr
184 * lma_flags was also removed because of lma_compat/incompat fields.
186 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
189 * OST object IDentifier.
197 struct lu_fid oi_fid;
201 #define DOSTID LPX64":"LPU64
202 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
205 * The ioctl naming rules:
206 * LL_* - works on the currently opened filehandle instead of parent dir
207 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
208 * *_MDC_* - gets/sets data related to MDC
209 * *_LOV_* - gets/sets data related to OSC/LOV
210 * *FILE* - called on parent dir and passes in a filename
211 * *STRIPE* - set/get lov_user_md
212 * *INFO - set/get lov_user_mds_data
214 /* lustre_ioctl.h 101-150 */
215 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
216 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
217 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
218 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
219 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
220 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
221 /* LL_IOC_RECREATE_OBJ 157 obsolete */
222 /* LL_IOC_RECREATE_FID 157 obsolete */
223 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
224 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
225 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
226 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
227 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
228 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
229 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
230 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
231 #define LL_IOC_RMTACL _IOW ('f', 167, long)
232 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
233 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
234 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
235 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
236 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
237 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
238 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
239 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
240 /* lustre_ioctl.h 177-210 */
241 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
242 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
243 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
244 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
245 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
246 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
247 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
248 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
249 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
250 struct lustre_swap_layouts)
251 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
252 struct hsm_current_action)
253 /* lustre_ioctl.h 221-232 */
254 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
255 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
256 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
257 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
258 #define LL_IOC_GET_LEASE _IO('f', 244)
259 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
260 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
261 #define LL_IOC_MIGRATE _IOR('f', 247, int)
262 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
263 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
265 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
267 LL_LEASE_RDLCK = 0x1,
268 LL_LEASE_WRLCK = 0x2,
269 LL_LEASE_UNLCK = 0x4,
272 #define LL_STATFS_LMV 1
273 #define LL_STATFS_LOV 2
274 #define LL_STATFS_NODELAY 4
276 #define IOC_MDC_TYPE 'i'
277 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
278 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
279 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
280 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
282 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
284 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
285 * files, but are unlikely to be used in practice and are not harmful if
286 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
287 * devices and are safe for use on new files. See LU-4209. */
288 /* To be compatible with old statically linked binary we keep the check for
289 * the older 0100000000 flag. This is already removed upstream. LU-812. */
290 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
291 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
292 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
293 O_LOV_DELAY_CREATE_MASK)
295 #define LL_FILE_IGNORE_LOCK 0x00000001
296 #define LL_FILE_GROUP_LOCKED 0x00000002
297 #define LL_FILE_READAHEA 0x00000004
298 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
299 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
300 #define LL_FILE_RMTACL 0x00000020
302 #define LOV_USER_MAGIC_V1 0x0BD10BD0
303 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
304 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
305 #define LOV_USER_MAGIC_V3 0x0BD30BD0
306 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
307 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
309 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
311 #define LOV_PATTERN_RAID0 0x001
312 #define LOV_PATTERN_RAID1 0x002
313 #define LOV_PATTERN_FIRST 0x100
314 #define LOV_PATTERN_CMOBD 0x200
316 #define LOV_PATTERN_F_MASK 0xffff0000
317 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
318 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
320 #define LOV_MAXPOOLNAME 15
321 #define LOV_POOLNAMEF "%.15s"
323 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
324 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
325 #define LOV_MAX_STRIPE_COUNT_OLD 160
326 /* This calculation is crafted so that input of 4096 will result in 160
327 * which in turn is equal to old maximal stripe count.
328 * XXX: In fact this is too simpified for now, what it also need is to get
329 * ea_type argument to clearly know how much space each stripe consumes.
331 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
332 * allocation that is sufficient for the current generation of systems.
334 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
335 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
336 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
337 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
339 #define XATTR_LUSTRE_PREFIX "lustre."
340 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
342 #define lov_user_ost_data lov_user_ost_data_v1
343 struct lov_user_ost_data_v1 { /* per-stripe data structure */
344 struct ost_id l_ost_oi; /* OST object ID */
345 __u32 l_ost_gen; /* generation of this OST index */
346 __u32 l_ost_idx; /* OST index in LOV */
347 } __attribute__((packed));
349 #define lov_user_md lov_user_md_v1
350 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
351 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
352 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
353 struct ost_id lmm_oi; /* LOV object ID */
354 __u32 lmm_stripe_size; /* size of stripe in bytes */
355 __u16 lmm_stripe_count; /* num stripes in use for this object */
357 __u16 lmm_stripe_offset; /* starting stripe offset in
358 * lmm_objects, use when writing */
359 __u16 lmm_layout_gen; /* layout generation number
360 * used when reading */
362 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
363 } __attribute__((packed, __may_alias__));
365 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
366 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
367 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
368 struct ost_id lmm_oi; /* LOV object ID */
369 __u32 lmm_stripe_size; /* size of stripe in bytes */
370 __u16 lmm_stripe_count; /* num stripes in use for this object */
372 __u16 lmm_stripe_offset; /* starting stripe offset in
373 * lmm_objects, use when writing */
374 __u16 lmm_layout_gen; /* layout generation number
375 * used when reading */
377 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
378 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
379 } __attribute__((packed));
381 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
383 if (lmm_magic == LOV_USER_MAGIC_V1)
384 return sizeof(struct lov_user_md_v1) +
385 stripes * sizeof(struct lov_user_ost_data_v1);
386 return sizeof(struct lov_user_md_v3) +
387 stripes * sizeof(struct lov_user_ost_data_v1);
390 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
391 * use this. It is unsafe to #define those values in this header as it
392 * is possible the application has already #included <sys/stat.h>. */
393 #ifdef HAVE_LOV_USER_MDS_DATA
394 #define lov_user_mds_data lov_user_mds_data_v1
395 struct lov_user_mds_data_v1 {
396 lstat_t lmd_st; /* MDS stat struct */
397 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
398 } __attribute__((packed));
400 struct lov_user_mds_data_v3 {
401 lstat_t lmd_st; /* MDS stat struct */
402 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
403 } __attribute__((packed));
406 struct lmv_user_mds_data {
407 struct lu_fid lum_fid;
413 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
414 LMV_HASH_TYPE_ALL_CHARS = 1,
415 LMV_HASH_TYPE_FNV_1A_64 = 2,
418 #define LMV_HASH_NAME_ALL_CHARS "all_char"
419 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
421 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
422 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
423 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
424 #define lmv_user_md lmv_user_md_v1
425 struct lmv_user_md_v1 {
426 __u32 lum_magic; /* must be the first field */
427 __u32 lum_stripe_count; /* dirstripe count */
428 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
429 __u32 lum_hash_type; /* Dir stripe policy */
430 __u32 lum_type; /* LMV type: default or normal */
434 char lum_pool_name[LOV_MAXPOOLNAME + 1];
435 struct lmv_user_mds_data lum_objects[0];
436 } __attribute__((packed));
438 static inline int lmv_user_md_size(int stripes, int lmm_magic)
440 return sizeof(struct lmv_user_md) +
441 stripes * sizeof(struct lmv_user_mds_data);
444 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
446 struct ll_recreate_obj {
452 __u64 id; /* holds object id */
453 __u32 generation; /* holds object generation */
454 __u32 f_type; /* holds object type or stripe idx when passing it to
455 * OST for saving into EA. */
463 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
464 const struct obd_uuid *u2)
466 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
469 static inline int obd_uuid_empty(struct obd_uuid *uuid)
471 return uuid->uuid[0] == '\0';
474 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
476 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
477 uuid->uuid[sizeof(*uuid) - 1] = '\0';
480 /* For printf's only, make sure uuid is terminated */
481 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
486 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
487 /* Obviously not safe, but for printfs, no real harm done...
488 we're always null-terminated, even in a race. */
489 static char temp[sizeof(*uuid)];
490 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
491 temp[sizeof(*uuid) - 1] = '\0';
494 return (char *)(uuid->uuid);
497 /* Extract fsname from uuid (or target name) of a target
498 e.g. (myfs-OST0007_UUID -> myfs)
499 see also deuuidify. */
500 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
504 strncpy(buf, uuid, buflen - 1);
505 buf[buflen - 1] = '\0';
506 p = strrchr(buf, '-');
511 /* printf display format
512 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
513 #define FID_NOBRACE_LEN 40
514 #define FID_LEN (FID_NOBRACE_LEN + 2)
515 #define DFID_NOBRACE LPX64":0x%x:0x%x"
516 #define DFID "["DFID_NOBRACE"]"
522 /* scanf input parse format -- strip '[' first.
523 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
524 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
525 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
526 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
528 #define SFID "0x"LPX64i":0x%x:0x%x"
535 /********* Quotas **********/
537 #define LUSTRE_QUOTABLOCK_BITS 10
538 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
540 static inline __u64 lustre_stoqb(size_t space)
542 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
545 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
546 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
547 #define Q_GETOINFO 0x800102 /* get obd quota info */
548 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
549 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
551 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
552 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
553 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
554 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
555 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
556 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
557 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
558 /* lustre-specific control commands */
559 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
560 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
562 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
564 struct if_quotacheck {
566 struct obd_uuid obd_uuid;
569 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
572 #define N_PERMS_MAX 64
574 struct perm_downcall_data {
580 struct identity_downcall_data {
587 struct perm_downcall_data idd_perms[N_PERMS_MAX];
591 /* for non-mapped uid/gid */
592 #define NOBODY_UID 99
593 #define NOBODY_GID 99
595 #define INVALID_ID (-1)
604 #ifdef NEED_QUOTA_DEFS
606 #define QIF_BLIMITS 1
608 #define QIF_ILIMITS 4
612 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
613 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
614 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
615 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
618 #endif /* !__KERNEL__ */
620 /* lustre volatile file support
621 * file name header: .^L^S^T^R:volatile"
623 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
624 #define LUSTRE_VOLATILE_HDR_LEN 14
626 typedef enum lustre_quota_version {
628 } lustre_quota_version_t;
630 /* XXX: same as if_dqinfo struct in kernel */
638 /* XXX: same as if_dqblk struct in kernel, plus one padding */
640 __u64 dqb_bhardlimit;
641 __u64 dqb_bsoftlimit;
643 __u64 dqb_ihardlimit;
644 __u64 dqb_isoftlimit;
666 struct obd_dqinfo qc_dqinfo;
667 struct obd_dqblk qc_dqblk;
669 struct obd_uuid obd_uuid;
672 /* swap layout flags */
673 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
674 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
675 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
676 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
678 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
679 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
680 struct lustre_swap_layouts {
689 /********* Changelogs **********/
690 /** Changelog record types */
691 enum changelog_rec_type {
693 CL_CREATE = 1, /* namespace */
694 CL_MKDIR = 2, /* namespace */
695 CL_HARDLINK = 3, /* namespace */
696 CL_SOFTLINK = 4, /* namespace */
697 CL_MKNOD = 5, /* namespace */
698 CL_UNLINK = 6, /* namespace */
699 CL_RMDIR = 7, /* namespace */
700 CL_RENAME = 8, /* namespace */
701 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
702 CL_OPEN = 10, /* not currently used */
703 CL_CLOSE = 11, /* may be written to log only with mtime change */
704 CL_LAYOUT = 12, /* file layout/striping modified */
708 CL_HSM = 16, /* HSM specific events, see flags */
709 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
715 static inline const char *changelog_type2str(int type) {
716 static const char *changelog_str[] = {
717 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
718 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
719 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
722 if (type >= 0 && type < CL_LAST)
723 return changelog_str[type];
727 /* per-record flags */
728 #define CLF_FLAGSHIFT 12
729 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
730 #define CLF_VERMASK (~CLF_FLAGMASK)
731 enum changelog_rec_flags {
732 CLF_VERSION = 0x1000,
735 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
739 /* Anything under the flagmask may be per-type (if desired) */
740 /* Flags for unlink */
741 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
742 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
743 /* HSM cleaning needed */
744 /* Flags for rename */
745 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
747 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
748 * has an archive in backend */
751 /* 12b used (from high weight to low weight):
756 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
757 #define CLF_HSM_ERR_H 6
758 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
759 #define CLF_HSM_EVENT_H 9
760 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
761 #define CLF_HSM_FLAG_H 11
762 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
763 #define CLF_HSM_SPARE_H 15
764 #define CLF_HSM_LAST 15
766 /* Remove bits higher than _h, then extract the value
767 * between _h and _l by shifting lower weigth to bit 0. */
768 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
769 >> (CLF_HSM_LAST - _h + _l))
771 #define CLF_HSM_SUCCESS 0x00
772 #define CLF_HSM_MAXERROR 0x7E
773 #define CLF_HSM_ERROVERFLOW 0x7F
775 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
777 /* 3 bits field => 8 values allowed */
789 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
791 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
795 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
797 *flags |= (he << CLF_HSM_EVENT_L);
800 static inline __u16 hsm_get_cl_flags(int flags)
802 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
805 static inline void hsm_set_cl_flags(int *flags, int bits)
807 *flags |= (bits << CLF_HSM_FLAG_L);
810 static inline int hsm_get_cl_error(int flags)
812 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
815 static inline void hsm_set_cl_error(int *flags, int error)
817 *flags |= (error << CLF_HSM_ERR_L);
820 enum changelog_send_flag {
821 /* Not yet implemented */
822 CHANGELOG_FLAG_FOLLOW = 0x01,
823 /* Blocking IO makes sense in case of slow user parsing of the records,
824 * but it also prevents us from cleaning up if the records are not
826 CHANGELOG_FLAG_BLOCK = 0x02,
827 /* Pack jobid into the changelog records if available. */
828 CHANGELOG_FLAG_JOBID = 0x04,
831 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
832 changelog_rec_offset(CLF_SUPPORTED))
834 /* 31 usable bytes string + null terminator. */
835 #define LUSTRE_JOBID_SIZE 32
837 /* This is the minimal changelog record. It can contain extensions
838 * such as rename fields or process jobid. Its exact content is described
841 * Extensions are packed in the same order as their corresponding flags.
843 struct changelog_rec {
845 __u16 cr_flags; /**< \a changelog_rec_flags */
846 __u32 cr_type; /**< \a changelog_rec_type */
847 __u64 cr_index; /**< changelog record number */
848 __u64 cr_prev; /**< last index for this target fid */
851 lustre_fid cr_tfid; /**< target fid */
852 __u32 cr_markerflags; /**< CL_MARK flags */
854 lustre_fid cr_pfid; /**< parent fid */
857 /* Changelog extension for RENAME. */
858 struct changelog_ext_rename {
859 lustre_fid cr_sfid; /**< source fid, or zero */
860 lustre_fid cr_spfid; /**< source parent fid, or zero */
863 /* Changelog extension to include JOBID. */
864 struct changelog_ext_jobid {
865 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
869 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
871 size_t size = sizeof(struct changelog_rec);
873 if (crf & CLF_RENAME)
874 size += sizeof(struct changelog_ext_rename);
877 size += sizeof(struct changelog_ext_jobid);
882 static inline size_t changelog_rec_size(struct changelog_rec *rec)
884 return changelog_rec_offset(rec->cr_flags);
887 static inline size_t changelog_rec_varsize(struct changelog_rec *rec)
889 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
893 struct changelog_ext_rename *changelog_rec_rename(struct changelog_rec *rec)
895 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
897 return (struct changelog_ext_rename *)((char *)rec +
898 changelog_rec_offset(crf));
901 /* The jobid follows the rename extension, if present */
903 struct changelog_ext_jobid *changelog_rec_jobid(struct changelog_rec *rec)
905 enum changelog_rec_flags crf = rec->cr_flags &
906 (CLF_VERSION | CLF_RENAME);
908 return (struct changelog_ext_jobid *)((char *)rec +
909 changelog_rec_offset(crf));
912 /* The name follows the rename and jobid extensions, if present */
913 static inline char *changelog_rec_name(struct changelog_rec *rec)
915 return (char *)rec + changelog_rec_offset(rec->cr_flags &
919 static inline size_t changelog_rec_snamelen(struct changelog_rec *rec)
921 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
924 static inline char *changelog_rec_sname(struct changelog_rec *rec)
926 char *cr_name = changelog_rec_name(rec);
928 return cr_name + strlen(cr_name) + 1;
932 * Remap a record to the desired format as specified by the crf flags.
933 * The record must be big enough to contain the final remapped version.
934 * Superfluous extension fields are removed and missing ones are added
935 * and zeroed. The flags of the record are updated accordingly.
937 * The jobid and rename extensions can be added to a record, to match the
938 * format an application expects, typically. In this case, the newly added
939 * fields will be zeroed.
940 * The Jobid field can be removed, to guarantee compatibility with older
941 * clients that don't expect this field in the records they process.
943 * The following assumptions are being made:
944 * - CLF_RENAME will not be removed
945 * - CLF_JOBID will not be added without CLF_RENAME being added too
947 * @param[in,out] rec The record to remap.
948 * @param[in] crf_wanted Flags describing the desired extensions.
950 static inline void changelog_remap_rec(struct changelog_rec *rec,
951 enum changelog_rec_flags crf_wanted)
956 crf_wanted &= CLF_SUPPORTED;
958 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
961 /* First move the variable-length name field */
962 memmove((char *)rec + changelog_rec_offset(crf_wanted),
963 changelog_rec_name(rec), rec->cr_namelen);
965 /* Locations of jobid and rename extensions in the remapped record */
966 jid_mov = (char *)rec +
967 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
968 rnm_mov = (char *)rec +
969 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
971 /* Move the extension fields to the desired positions */
972 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
973 memmove(jid_mov, changelog_rec_jobid(rec),
974 sizeof(struct changelog_ext_jobid));
976 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
977 memmove(rnm_mov, changelog_rec_rename(rec),
978 sizeof(struct changelog_ext_rename));
980 /* Clear newly added fields */
981 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
982 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
984 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
985 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
987 /* Update the record's flags accordingly */
988 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
991 struct ioc_changelog {
998 enum changelog_message_type {
999 CL_RECORD = 10, /* message is a changelog_rec */
1000 CL_EOF = 11, /* at end of current changelog */
1003 /********* Misc **********/
1005 struct ioc_data_version {
1007 __u64 idv_flags; /* See LL_DV_xxx */
1009 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
1010 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
1013 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1016 #define dot_lustre_name ".lustre"
1019 /********* HSM **********/
1021 /** HSM per-file state
1022 * See HSM_FLAGS below.
1025 HS_EXISTS = 0x00000001,
1026 HS_DIRTY = 0x00000002,
1027 HS_RELEASED = 0x00000004,
1028 HS_ARCHIVED = 0x00000008,
1029 HS_NORELEASE = 0x00000010,
1030 HS_NOARCHIVE = 0x00000020,
1031 HS_LOST = 0x00000040,
1034 /* HSM user-setable flags. */
1035 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1037 /* Other HSM flags. */
1038 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1041 * All HSM-related possible flags that could be applied to a file.
1042 * This should be kept in sync with hsm_states.
1044 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1047 * HSMÂ request progress state
1049 enum hsm_progress_states {
1056 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1059 case HPS_WAITING: return "waiting";
1060 case HPS_RUNNING: return "running";
1061 case HPS_DONE: return "done";
1062 default: return "unknown";
1069 } __attribute__((packed));
1072 * Current HSM states of a Lustre file.
1074 * This structure purpose is to be sent to user-space mainly. It describes the
1075 * current HSM flags and in-progress action.
1077 struct hsm_user_state {
1078 /** Current HSM states, from enum hsm_states. */
1080 __u32 hus_archive_id;
1081 /** The current undergoing action, if there is one */
1082 __u32 hus_in_progress_state;
1083 __u32 hus_in_progress_action;
1084 struct hsm_extent hus_in_progress_location;
1085 char hus_extended_info[];
1088 struct hsm_state_set_ioc {
1089 struct lu_fid hssi_fid;
1091 __u64 hssi_clearmask;
1095 * This structure describes the current in-progress action for a file.
1096 * it is retuned to user space and send over the wire
1098 struct hsm_current_action {
1099 /** The current undergoing action, if there is one */
1100 /* state is one of hsm_progress_states */
1102 /* action is one of hsm_user_action */
1104 struct hsm_extent hca_location;
1107 /***** HSM user requests ******/
1108 /* User-generated (lfs/ioctl) request types */
1109 enum hsm_user_action {
1110 HUA_NONE = 1, /* no action (noop) */
1111 HUA_ARCHIVE = 10, /* copy to hsm */
1112 HUA_RESTORE = 11, /* prestage */
1113 HUA_RELEASE = 12, /* drop ost objects */
1114 HUA_REMOVE = 13, /* remove from archive */
1115 HUA_CANCEL = 14 /* cancel a request */
1118 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1121 case HUA_NONE: return "NOOP";
1122 case HUA_ARCHIVE: return "ARCHIVE";
1123 case HUA_RESTORE: return "RESTORE";
1124 case HUA_RELEASE: return "RELEASE";
1125 case HUA_REMOVE: return "REMOVE";
1126 case HUA_CANCEL: return "CANCEL";
1127 default: return "UNKNOWN";
1132 * List of hr_flags (bit field)
1134 #define HSM_FORCE_ACTION 0x0001
1135 /* used by CT, connot be set by user */
1136 #define HSM_GHOST_COPY 0x0002
1139 * Contains all the fixed part of struct hsm_user_request.
1142 struct hsm_request {
1143 __u32 hr_action; /* enum hsm_user_action */
1144 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1145 __u64 hr_flags; /* request flags */
1146 __u32 hr_itemcount; /* item count in hur_user_item vector */
1150 struct hsm_user_item {
1152 struct hsm_extent hui_extent;
1153 } __attribute__((packed));
1155 struct hsm_user_request {
1156 struct hsm_request hur_request;
1157 struct hsm_user_item hur_user_item[0];
1158 /* extra data blob at end of struct (after all
1159 * hur_user_items), only use helpers to access it
1161 } __attribute__((packed));
1163 /** Return pointer to data field in a hsm user request */
1164 static inline void *hur_data(struct hsm_user_request *hur)
1166 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1170 * Compute the current length of the provided hsm_user_request. This returns -1
1171 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1173 * return -1 on bounds check error.
1175 static inline ssize_t hur_len(struct hsm_user_request *hur)
1179 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1180 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1181 (__u64)hur->hur_request.hr_itemcount *
1182 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1184 if (size > ~((size_t)0)>>1)
1190 /****** HSM RPCs to copytool *****/
1191 /* Message types the copytool may receive */
1192 enum hsm_message_type {
1193 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1196 /* Actions the copytool may be instructed to take for a given action_item */
1197 enum hsm_copytool_action {
1198 HSMA_NONE = 10, /* no action */
1199 HSMA_ARCHIVE = 20, /* arbitrary offset */
1205 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1208 case HSMA_NONE: return "NOOP";
1209 case HSMA_ARCHIVE: return "ARCHIVE";
1210 case HSMA_RESTORE: return "RESTORE";
1211 case HSMA_REMOVE: return "REMOVE";
1212 case HSMA_CANCEL: return "CANCEL";
1213 default: return "UNKNOWN";
1217 /* Copytool item action description */
1218 struct hsm_action_item {
1219 __u32 hai_len; /* valid size of this struct */
1220 __u32 hai_action; /* hsm_copytool_action, but use known size */
1221 lustre_fid hai_fid; /* Lustre FID to operated on */
1222 lustre_fid hai_dfid; /* fid used for data access */
1223 struct hsm_extent hai_extent; /* byte range to operate on */
1224 __u64 hai_cookie; /* action cookie from coordinator */
1225 __u64 hai_gid; /* grouplock id */
1226 char hai_data[0]; /* variable length */
1227 } __attribute__((packed));
1230 * helper function which print in hexa the first bytes of
1232 * \param hai [IN] record to print
1233 * \param buffer [OUT] output buffer
1234 * \param len [IN] max buffer len
1237 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1238 char *buffer, int len)
1240 int i, sz, data_len;
1245 data_len = hai->hai_len - sizeof(*hai);
1246 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1250 cnt = snprintf(ptr, sz, "%.2X",
1251 (unsigned char)hai->hai_data[i]);
1259 /* Copytool action list */
1260 #define HAL_VERSION 1
1261 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1262 struct hsm_action_list {
1264 __u32 hal_count; /* number of hai's to follow */
1265 __u64 hal_compound_id; /* returned by coordinator */
1267 __u32 hal_archive_id; /* which archive backend */
1269 char hal_fsname[0]; /* null-terminated */
1270 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1271 boundaries. See hai_zero */
1272 } __attribute__((packed));
1274 #ifndef HAVE_CFS_SIZE_ROUND
1275 static inline int cfs_size_round (int val)
1277 return (val + 7) & (~0x7);
1279 #define HAVE_CFS_SIZE_ROUND
1282 /* Return pointer to first hai in action list */
1283 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1285 return (struct hsm_action_item *)(hal->hal_fsname +
1286 cfs_size_round(strlen(hal-> \
1290 /* Return pointer to next hai */
1291 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1293 return (struct hsm_action_item *)((char *)hai +
1294 cfs_size_round(hai->hai_len));
1297 /* Return size of an hsm_action_list */
1298 static inline size_t hal_size(struct hsm_action_list *hal)
1302 struct hsm_action_item *hai;
1304 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1305 hai = hai_first(hal);
1306 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1307 sz += cfs_size_round(hai->hai_len);
1313 * describe the attributes to be set on imported file
1315 struct hsm_user_import {
1324 __u32 hui_archive_id;
1327 /* Copytool progress reporting */
1328 #define HP_FLAG_COMPLETED 0x01
1329 #define HP_FLAG_RETRY 0x02
1331 struct hsm_progress {
1334 struct hsm_extent hp_extent;
1336 __u16 hp_errval; /* positive val */
1341 __u64 hc_data_version;
1343 __u16 hc_errval; /* positive val */
1345 struct hsm_action_item hc_hai;
1349 enum llapi_json_types {
1350 LLAPI_JSON_INTEGER = 1,
1356 struct llapi_json_item {
1365 struct llapi_json_item *lji_next;
1368 struct llapi_json_item_list {
1369 int ljil_item_count;
1370 struct llapi_json_item *ljil_items;
1373 /** @} lustreuser */
1375 #endif /* _LUSTRE_USER_H */