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/quota.h>
53 # include <linux/string.h> /* snprintf() */
54 # include <linux/version.h>
55 #else /* !__KERNEL__ */
56 # define NEED_QUOTA_DEFS
57 # include <stdio.h> /* snprintf() */
59 # include <sys/quota.h>
60 # include <sys/stat.h>
61 #endif /* __KERNEL__ */
62 #include <lustre/ll_fiemap.h>
64 #if defined(__x86_64__) || defined(__ia64__) || defined(__ppc64__) || \
65 defined(__craynv) || defined(__mips64__) || defined(__powerpc64__)
66 typedef struct stat lstat_t;
67 # define lstat_f lstat
68 # define fstat_f fstat
69 # define fstatat_f fstatat
70 # define HAVE_LOV_USER_MDS_DATA
71 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
72 typedef struct stat64 lstat_t;
73 # define lstat_f lstat64
74 # define fstat_f fstat64
75 # define fstatat_f fstatat64
76 # 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)
91 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
94 /* FIEMAP flags supported by Lustre */
95 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
97 enum obd_statfs_state {
98 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
99 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
100 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
101 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
102 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
116 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
117 __u32 os_fprecreated; /* objs available now to the caller */
118 /* used in QoS code to find preferred
133 * FID is a cluster-wide unique identifier of a file or an object (stripe).
134 * FIDs are never reused.
138 * FID sequence. Sequence is a unit of migration: all files (objects)
139 * with FIDs from a given sequence are stored on the same server.
140 * Lustre should support 2^64 objects, so even if each sequence
141 * has only a single object we can still enumerate 2^64 objects.
144 /* FID number within sequence. */
147 * FID version, used to distinguish different versions (in the sense
148 * of snapshots, etc.) of the same file system object. Not currently
154 static inline bool fid_is_zero(const struct lu_fid *fid)
156 return fid->f_seq == 0 && fid->f_oid == 0;
159 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
160 * MDT-object's FID::f_ver, instead it is the OST-object index in its
161 * parent MDT-object's layout EA. */
162 #define f_stripe_idx f_ver
165 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
168 /* keep this one for compatibility */
169 struct filter_fid_old {
170 struct lu_fid ff_parent;
175 /* Userspace should treat lu_fid as opaque, and only use the following methods
176 * to print or parse them. Other functions (e.g. compare, swab) could be moved
177 * here from lustre_idl.h if needed. */
178 typedef struct lu_fid lustre_fid;
181 * Following struct for object attributes, that will be kept inode's EA.
182 * Introduced in 2.0 release (please see b15993, for details)
183 * Added to all objects since Lustre 2.4 as contains self FID
185 struct lustre_mdt_attrs {
187 * Bitfield for supported data in this structure. From enum lma_compat.
188 * lma_self_fid and lma_flags are always available.
192 * Per-file incompat feature list. Lustre version should support all
193 * flags set in this field. The supported feature mask is available in
197 /** FID of this inode */
198 struct lu_fid lma_self_fid;
202 * Prior to 2.4, the LMA structure also included SOM attributes which has since
203 * been moved to a dedicated xattr
204 * lma_flags was also removed because of lma_compat/incompat fields.
206 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
209 * OST object IDentifier.
217 struct lu_fid oi_fid;
221 #define DOSTID LPX64":"LPU64
222 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
225 * The ioctl naming rules:
226 * LL_* - works on the currently opened filehandle instead of parent dir
227 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
228 * *_MDC_* - gets/sets data related to MDC
229 * *_LOV_* - gets/sets data related to OSC/LOV
230 * *FILE* - called on parent dir and passes in a filename
231 * *STRIPE* - set/get lov_user_md
232 * *INFO - set/get lov_user_mds_data
234 /* lustre_ioctl.h 101-150 */
235 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
236 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
237 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
238 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
239 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
240 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
241 /* LL_IOC_RECREATE_OBJ 157 obsolete */
242 /* LL_IOC_RECREATE_FID 157 obsolete */
243 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
244 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
245 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
246 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
247 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
248 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
249 /* IOC_LOV_GETINFO 165 obsolete */
250 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
251 #define LL_IOC_RMTACL _IOW ('f', 167, long)
252 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
253 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
254 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
255 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
256 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
257 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
258 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
259 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
260 /* lustre_ioctl.h 177-210 */
261 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
262 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
263 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
264 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
265 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
266 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
267 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
268 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
269 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
270 struct lustre_swap_layouts)
271 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
272 struct hsm_current_action)
273 /* lustre_ioctl.h 221-232 */
274 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
275 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
276 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
277 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
278 #define LL_IOC_GET_LEASE _IO('f', 244)
279 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
280 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
281 #define LL_IOC_MIGRATE _IOR('f', 247, int)
282 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
283 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
285 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
287 LL_LEASE_RDLCK = 0x1,
288 LL_LEASE_WRLCK = 0x2,
289 LL_LEASE_UNLCK = 0x4,
292 #define LL_STATFS_LMV 1
293 #define LL_STATFS_LOV 2
294 #define LL_STATFS_NODELAY 4
296 #define IOC_MDC_TYPE 'i'
297 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
298 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
299 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
300 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
302 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
304 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
305 * files, but are unlikely to be used in practice and are not harmful if
306 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
307 * devices and are safe for use on new files. See LU-4209. */
308 /* To be compatible with old statically linked binary we keep the check for
309 * the older 0100000000 flag. This is already removed upstream. LU-812. */
310 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
311 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
312 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
313 O_LOV_DELAY_CREATE_MASK)
315 #define LL_FILE_IGNORE_LOCK 0x00000001
316 #define LL_FILE_GROUP_LOCKED 0x00000002
317 #define LL_FILE_READAHEA 0x00000004
318 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
319 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
320 #define LL_FILE_RMTACL 0x00000020
322 #define LOV_USER_MAGIC_V1 0x0BD10BD0
323 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
324 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
325 #define LOV_USER_MAGIC_V3 0x0BD30BD0
326 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
327 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
329 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
331 #define LOV_PATTERN_RAID0 0x001
332 #define LOV_PATTERN_RAID1 0x002
333 #define LOV_PATTERN_FIRST 0x100
334 #define LOV_PATTERN_CMOBD 0x200
336 #define LOV_PATTERN_F_MASK 0xffff0000
337 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
338 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
340 #define LOV_MAXPOOLNAME 15
341 #define LOV_POOLNAMEF "%.15s"
343 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
344 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
345 #define LOV_MAX_STRIPE_COUNT_OLD 160
346 /* This calculation is crafted so that input of 4096 will result in 160
347 * which in turn is equal to old maximal stripe count.
348 * XXX: In fact this is too simpified for now, what it also need is to get
349 * ea_type argument to clearly know how much space each stripe consumes.
351 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
352 * allocation that is sufficient for the current generation of systems.
354 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
355 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
356 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
357 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
359 #define XATTR_LUSTRE_PREFIX "lustre."
360 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
362 #define lov_user_ost_data lov_user_ost_data_v1
363 struct lov_user_ost_data_v1 { /* per-stripe data structure */
364 struct ost_id l_ost_oi; /* OST object ID */
365 __u32 l_ost_gen; /* generation of this OST index */
366 __u32 l_ost_idx; /* OST index in LOV */
367 } __attribute__((packed));
369 #define lov_user_md lov_user_md_v1
370 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
371 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
372 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
373 struct ost_id lmm_oi; /* LOV object ID */
374 __u32 lmm_stripe_size; /* size of stripe in bytes */
375 __u16 lmm_stripe_count; /* num stripes in use for this object */
377 __u16 lmm_stripe_offset; /* starting stripe offset in
378 * lmm_objects, use when writing */
379 __u16 lmm_layout_gen; /* layout generation number
380 * used when reading */
382 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
383 } __attribute__((packed, __may_alias__));
385 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
386 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
387 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
388 struct ost_id lmm_oi; /* LOV object ID */
389 __u32 lmm_stripe_size; /* size of stripe in bytes */
390 __u16 lmm_stripe_count; /* num stripes in use for this object */
392 __u16 lmm_stripe_offset; /* starting stripe offset in
393 * lmm_objects, use when writing */
394 __u16 lmm_layout_gen; /* layout generation number
395 * used when reading */
397 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
398 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
399 } __attribute__((packed));
401 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
403 if (lmm_magic == LOV_USER_MAGIC_V1)
404 return sizeof(struct lov_user_md_v1) +
405 stripes * sizeof(struct lov_user_ost_data_v1);
406 return sizeof(struct lov_user_md_v3) +
407 stripes * sizeof(struct lov_user_ost_data_v1);
410 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
411 * use this. It is unsafe to #define those values in this header as it
412 * is possible the application has already #included <sys/stat.h>. */
413 #ifdef HAVE_LOV_USER_MDS_DATA
414 #define lov_user_mds_data lov_user_mds_data_v1
415 struct lov_user_mds_data_v1 {
416 lstat_t lmd_st; /* MDS stat struct */
417 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
418 } __attribute__((packed));
420 struct lov_user_mds_data_v3 {
421 lstat_t lmd_st; /* MDS stat struct */
422 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
423 } __attribute__((packed));
426 struct lmv_user_mds_data {
427 struct lu_fid lum_fid;
433 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
434 LMV_HASH_TYPE_ALL_CHARS = 1,
435 LMV_HASH_TYPE_FNV_1A_64 = 2,
438 #define LMV_HASH_NAME_ALL_CHARS "all_char"
439 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
441 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
442 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
443 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
444 #define lmv_user_md lmv_user_md_v1
445 struct lmv_user_md_v1 {
446 __u32 lum_magic; /* must be the first field */
447 __u32 lum_stripe_count; /* dirstripe count */
448 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
449 __u32 lum_hash_type; /* Dir stripe policy */
450 __u32 lum_type; /* LMV type: default or normal */
454 char lum_pool_name[LOV_MAXPOOLNAME + 1];
455 struct lmv_user_mds_data lum_objects[0];
456 } __attribute__((packed));
458 static inline int lmv_user_md_size(int stripes, int lmm_magic)
460 return sizeof(struct lmv_user_md) +
461 stripes * sizeof(struct lmv_user_mds_data);
464 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
466 struct ll_recreate_obj {
472 __u64 id; /* holds object id */
473 __u32 generation; /* holds object generation */
474 __u32 f_type; /* holds object type or stripe idx when passing it to
475 * OST for saving into EA. */
483 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
484 const struct obd_uuid *u2)
486 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
489 static inline int obd_uuid_empty(struct obd_uuid *uuid)
491 return uuid->uuid[0] == '\0';
494 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
496 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
497 uuid->uuid[sizeof(*uuid) - 1] = '\0';
500 /* For printf's only, make sure uuid is terminated */
501 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
506 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
507 /* Obviously not safe, but for printfs, no real harm done...
508 we're always null-terminated, even in a race. */
509 static char temp[sizeof(*uuid)];
510 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
511 temp[sizeof(*uuid) - 1] = '\0';
514 return (char *)(uuid->uuid);
517 /* Extract fsname from uuid (or target name) of a target
518 e.g. (myfs-OST0007_UUID -> myfs)
519 see also deuuidify. */
520 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
524 strncpy(buf, uuid, buflen - 1);
525 buf[buflen - 1] = '\0';
526 p = strrchr(buf, '-');
531 /* printf display format
532 e.g. 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 LPX64":0x%x:0x%x"
536 #define DFID "["DFID_NOBRACE"]"
542 /* scanf input parse format -- strip '[' first.
543 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
544 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
545 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
546 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
548 #define SFID "0x"LPX64i":0x%x:0x%x"
555 /********* Quotas **********/
557 #define LUSTRE_QUOTABLOCK_BITS 10
558 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
560 static inline __u64 lustre_stoqb(size_t space)
562 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
565 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
566 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
567 #define Q_GETOINFO 0x800102 /* get obd quota info */
568 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
569 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
571 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
572 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
573 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
574 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
575 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
576 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
577 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
578 /* lustre-specific control commands */
579 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
580 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
582 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
584 struct if_quotacheck {
586 struct obd_uuid obd_uuid;
589 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
592 #define N_PERMS_MAX 64
594 struct perm_downcall_data {
600 struct identity_downcall_data {
607 struct perm_downcall_data idd_perms[N_PERMS_MAX];
611 /* for non-mapped uid/gid */
612 #define NOBODY_UID 99
613 #define NOBODY_GID 99
615 #define INVALID_ID (-1)
624 #ifdef NEED_QUOTA_DEFS
626 #define QIF_BLIMITS 1
628 #define QIF_ILIMITS 4
632 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
633 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
634 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
635 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
638 #endif /* !__KERNEL__ */
640 /* lustre volatile file support
641 * file name header: .^L^S^T^R:volatile"
643 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
644 #define LUSTRE_VOLATILE_HDR_LEN 14
646 typedef enum lustre_quota_version {
648 } lustre_quota_version_t;
650 /* XXX: same as if_dqinfo struct in kernel */
658 /* XXX: same as if_dqblk struct in kernel, plus one padding */
660 __u64 dqb_bhardlimit;
661 __u64 dqb_bsoftlimit;
663 __u64 dqb_ihardlimit;
664 __u64 dqb_isoftlimit;
686 struct obd_dqinfo qc_dqinfo;
687 struct obd_dqblk qc_dqblk;
689 struct obd_uuid obd_uuid;
692 /* swap layout flags */
693 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
694 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
695 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
696 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
698 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
699 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
700 struct lustre_swap_layouts {
709 /********* Changelogs **********/
710 /** Changelog record types */
711 enum changelog_rec_type {
713 CL_CREATE = 1, /* namespace */
714 CL_MKDIR = 2, /* namespace */
715 CL_HARDLINK = 3, /* namespace */
716 CL_SOFTLINK = 4, /* namespace */
717 CL_MKNOD = 5, /* namespace */
718 CL_UNLINK = 6, /* namespace */
719 CL_RMDIR = 7, /* namespace */
720 CL_RENAME = 8, /* namespace */
721 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
722 CL_OPEN = 10, /* not currently used */
723 CL_CLOSE = 11, /* may be written to log only with mtime change */
724 CL_LAYOUT = 12, /* file layout/striping modified */
728 CL_HSM = 16, /* HSM specific events, see flags */
729 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
735 static inline const char *changelog_type2str(int type) {
736 static const char *changelog_str[] = {
737 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
738 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
739 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
742 if (type >= 0 && type < CL_LAST)
743 return changelog_str[type];
747 /* per-record flags */
748 #define CLF_FLAGSHIFT 12
749 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
750 #define CLF_VERMASK (~CLF_FLAGMASK)
751 enum changelog_rec_flags {
752 CLF_VERSION = 0x1000,
755 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
759 /* Anything under the flagmask may be per-type (if desired) */
760 /* Flags for unlink */
761 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
762 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
763 /* HSM cleaning needed */
764 /* Flags for rename */
765 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
767 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
768 * has an archive in backend */
771 /* 12b used (from high weight to low weight):
776 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
777 #define CLF_HSM_ERR_H 6
778 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
779 #define CLF_HSM_EVENT_H 9
780 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
781 #define CLF_HSM_FLAG_H 11
782 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
783 #define CLF_HSM_SPARE_H 15
784 #define CLF_HSM_LAST 15
786 /* Remove bits higher than _h, then extract the value
787 * between _h and _l by shifting lower weigth to bit 0. */
788 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
789 >> (CLF_HSM_LAST - _h + _l))
791 #define CLF_HSM_SUCCESS 0x00
792 #define CLF_HSM_MAXERROR 0x7E
793 #define CLF_HSM_ERROVERFLOW 0x7F
795 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
797 /* 3 bits field => 8 values allowed */
809 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
811 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
815 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
817 *flags |= (he << CLF_HSM_EVENT_L);
820 static inline __u16 hsm_get_cl_flags(int flags)
822 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
825 static inline void hsm_set_cl_flags(int *flags, int bits)
827 *flags |= (bits << CLF_HSM_FLAG_L);
830 static inline int hsm_get_cl_error(int flags)
832 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
835 static inline void hsm_set_cl_error(int *flags, int error)
837 *flags |= (error << CLF_HSM_ERR_L);
840 enum changelog_send_flag {
841 /* Not yet implemented */
842 CHANGELOG_FLAG_FOLLOW = 0x01,
843 /* Blocking IO makes sense in case of slow user parsing of the records,
844 * but it also prevents us from cleaning up if the records are not
846 CHANGELOG_FLAG_BLOCK = 0x02,
847 /* Pack jobid into the changelog records if available. */
848 CHANGELOG_FLAG_JOBID = 0x04,
851 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
852 changelog_rec_offset(CLF_SUPPORTED))
854 /* 31 usable bytes string + null terminator. */
855 #define LUSTRE_JOBID_SIZE 32
857 /* This is the minimal changelog record. It can contain extensions
858 * such as rename fields or process jobid. Its exact content is described
861 * Extensions are packed in the same order as their corresponding flags.
863 struct changelog_rec {
865 __u16 cr_flags; /**< \a changelog_rec_flags */
866 __u32 cr_type; /**< \a changelog_rec_type */
867 __u64 cr_index; /**< changelog record number */
868 __u64 cr_prev; /**< last index for this target fid */
871 lustre_fid cr_tfid; /**< target fid */
872 __u32 cr_markerflags; /**< CL_MARK flags */
874 lustre_fid cr_pfid; /**< parent fid */
877 /* Changelog extension for RENAME. */
878 struct changelog_ext_rename {
879 lustre_fid cr_sfid; /**< source fid, or zero */
880 lustre_fid cr_spfid; /**< source parent fid, or zero */
883 /* Changelog extension to include JOBID. */
884 struct changelog_ext_jobid {
885 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
889 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
891 size_t size = sizeof(struct changelog_rec);
893 if (crf & CLF_RENAME)
894 size += sizeof(struct changelog_ext_rename);
897 size += sizeof(struct changelog_ext_jobid);
902 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
904 return changelog_rec_offset(rec->cr_flags);
907 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
909 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
913 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
915 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
917 return (struct changelog_ext_rename *)((char *)rec +
918 changelog_rec_offset(crf));
921 /* The jobid follows the rename extension, if present */
923 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
925 enum changelog_rec_flags crf = rec->cr_flags &
926 (CLF_VERSION | CLF_RENAME);
928 return (struct changelog_ext_jobid *)((char *)rec +
929 changelog_rec_offset(crf));
932 /* The name follows the rename and jobid extensions, if present */
933 static inline char *changelog_rec_name(const struct changelog_rec *rec)
935 return (char *)rec + changelog_rec_offset(rec->cr_flags &
939 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
941 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
944 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
946 char *cr_name = changelog_rec_name(rec);
948 return cr_name + strlen(cr_name) + 1;
952 * Remap a record to the desired format as specified by the crf flags.
953 * The record must be big enough to contain the final remapped version.
954 * Superfluous extension fields are removed and missing ones are added
955 * and zeroed. The flags of the record are updated accordingly.
957 * The jobid and rename extensions can be added to a record, to match the
958 * format an application expects, typically. In this case, the newly added
959 * fields will be zeroed.
960 * The Jobid field can be removed, to guarantee compatibility with older
961 * clients that don't expect this field in the records they process.
963 * The following assumptions are being made:
964 * - CLF_RENAME will not be removed
965 * - CLF_JOBID will not be added without CLF_RENAME being added too
967 * @param[in,out] rec The record to remap.
968 * @param[in] crf_wanted Flags describing the desired extensions.
970 static inline void changelog_remap_rec(struct changelog_rec *rec,
971 enum changelog_rec_flags crf_wanted)
976 crf_wanted &= CLF_SUPPORTED;
978 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
981 /* First move the variable-length name field */
982 memmove((char *)rec + changelog_rec_offset(crf_wanted),
983 changelog_rec_name(rec), rec->cr_namelen);
985 /* Locations of jobid and rename extensions in the remapped record */
986 jid_mov = (char *)rec +
987 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
988 rnm_mov = (char *)rec +
989 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
991 /* Move the extension fields to the desired positions */
992 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
993 memmove(jid_mov, changelog_rec_jobid(rec),
994 sizeof(struct changelog_ext_jobid));
996 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
997 memmove(rnm_mov, changelog_rec_rename(rec),
998 sizeof(struct changelog_ext_rename));
1000 /* Clear newly added fields */
1001 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1002 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1004 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1005 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1007 /* Update the record's flags accordingly */
1008 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1011 struct ioc_changelog {
1018 enum changelog_message_type {
1019 CL_RECORD = 10, /* message is a changelog_rec */
1020 CL_EOF = 11, /* at end of current changelog */
1023 /********* Misc **********/
1025 struct ioc_data_version {
1027 __u64 idv_flags; /* See LL_DV_xxx */
1029 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
1030 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
1033 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1036 #define dot_lustre_name ".lustre"
1039 /********* HSM **********/
1041 /** HSM per-file state
1042 * See HSM_FLAGS below.
1045 HS_EXISTS = 0x00000001,
1046 HS_DIRTY = 0x00000002,
1047 HS_RELEASED = 0x00000004,
1048 HS_ARCHIVED = 0x00000008,
1049 HS_NORELEASE = 0x00000010,
1050 HS_NOARCHIVE = 0x00000020,
1051 HS_LOST = 0x00000040,
1054 /* HSM user-setable flags. */
1055 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1057 /* Other HSM flags. */
1058 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1061 * All HSM-related possible flags that could be applied to a file.
1062 * This should be kept in sync with hsm_states.
1064 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1067 * HSMÂ request progress state
1069 enum hsm_progress_states {
1076 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1079 case HPS_WAITING: return "waiting";
1080 case HPS_RUNNING: return "running";
1081 case HPS_DONE: return "done";
1082 default: return "unknown";
1089 } __attribute__((packed));
1092 * Current HSM states of a Lustre file.
1094 * This structure purpose is to be sent to user-space mainly. It describes the
1095 * current HSM flags and in-progress action.
1097 struct hsm_user_state {
1098 /** Current HSM states, from enum hsm_states. */
1100 __u32 hus_archive_id;
1101 /** The current undergoing action, if there is one */
1102 __u32 hus_in_progress_state;
1103 __u32 hus_in_progress_action;
1104 struct hsm_extent hus_in_progress_location;
1105 char hus_extended_info[];
1108 struct hsm_state_set_ioc {
1109 struct lu_fid hssi_fid;
1111 __u64 hssi_clearmask;
1115 * This structure describes the current in-progress action for a file.
1116 * it is retuned to user space and send over the wire
1118 struct hsm_current_action {
1119 /** The current undergoing action, if there is one */
1120 /* state is one of hsm_progress_states */
1122 /* action is one of hsm_user_action */
1124 struct hsm_extent hca_location;
1127 /***** HSM user requests ******/
1128 /* User-generated (lfs/ioctl) request types */
1129 enum hsm_user_action {
1130 HUA_NONE = 1, /* no action (noop) */
1131 HUA_ARCHIVE = 10, /* copy to hsm */
1132 HUA_RESTORE = 11, /* prestage */
1133 HUA_RELEASE = 12, /* drop ost objects */
1134 HUA_REMOVE = 13, /* remove from archive */
1135 HUA_CANCEL = 14 /* cancel a request */
1138 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1141 case HUA_NONE: return "NOOP";
1142 case HUA_ARCHIVE: return "ARCHIVE";
1143 case HUA_RESTORE: return "RESTORE";
1144 case HUA_RELEASE: return "RELEASE";
1145 case HUA_REMOVE: return "REMOVE";
1146 case HUA_CANCEL: return "CANCEL";
1147 default: return "UNKNOWN";
1152 * List of hr_flags (bit field)
1154 #define HSM_FORCE_ACTION 0x0001
1155 /* used by CT, connot be set by user */
1156 #define HSM_GHOST_COPY 0x0002
1159 * Contains all the fixed part of struct hsm_user_request.
1162 struct hsm_request {
1163 __u32 hr_action; /* enum hsm_user_action */
1164 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1165 __u64 hr_flags; /* request flags */
1166 __u32 hr_itemcount; /* item count in hur_user_item vector */
1170 struct hsm_user_item {
1172 struct hsm_extent hui_extent;
1173 } __attribute__((packed));
1175 struct hsm_user_request {
1176 struct hsm_request hur_request;
1177 struct hsm_user_item hur_user_item[0];
1178 /* extra data blob at end of struct (after all
1179 * hur_user_items), only use helpers to access it
1181 } __attribute__((packed));
1183 /** Return pointer to data field in a hsm user request */
1184 static inline void *hur_data(struct hsm_user_request *hur)
1186 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1190 * Compute the current length of the provided hsm_user_request. This returns -1
1191 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1193 * return -1 on bounds check error.
1195 static inline ssize_t hur_len(struct hsm_user_request *hur)
1199 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1200 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1201 (__u64)hur->hur_request.hr_itemcount *
1202 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1204 if (size != (ssize_t)size)
1210 /****** HSM RPCs to copytool *****/
1211 /* Message types the copytool may receive */
1212 enum hsm_message_type {
1213 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1216 /* Actions the copytool may be instructed to take for a given action_item */
1217 enum hsm_copytool_action {
1218 HSMA_NONE = 10, /* no action */
1219 HSMA_ARCHIVE = 20, /* arbitrary offset */
1225 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1228 case HSMA_NONE: return "NOOP";
1229 case HSMA_ARCHIVE: return "ARCHIVE";
1230 case HSMA_RESTORE: return "RESTORE";
1231 case HSMA_REMOVE: return "REMOVE";
1232 case HSMA_CANCEL: return "CANCEL";
1233 default: return "UNKNOWN";
1237 /* Copytool item action description */
1238 struct hsm_action_item {
1239 __u32 hai_len; /* valid size of this struct */
1240 __u32 hai_action; /* hsm_copytool_action, but use known size */
1241 lustre_fid hai_fid; /* Lustre FID to operated on */
1242 lustre_fid hai_dfid; /* fid used for data access */
1243 struct hsm_extent hai_extent; /* byte range to operate on */
1244 __u64 hai_cookie; /* action cookie from coordinator */
1245 __u64 hai_gid; /* grouplock id */
1246 char hai_data[0]; /* variable length */
1247 } __attribute__((packed));
1250 * helper function which print in hexa the first bytes of
1252 * \param hai [IN] record to print
1253 * \param buffer [OUT] output buffer
1254 * \param len [IN] max buffer len
1257 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1258 char *buffer, int len)
1260 int i, sz, data_len;
1265 data_len = hai->hai_len - sizeof(*hai);
1266 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1270 cnt = snprintf(ptr, sz, "%.2X",
1271 (unsigned char)hai->hai_data[i]);
1279 /* Copytool action list */
1280 #define HAL_VERSION 1
1281 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1282 struct hsm_action_list {
1284 __u32 hal_count; /* number of hai's to follow */
1285 __u64 hal_compound_id; /* returned by coordinator */
1287 __u32 hal_archive_id; /* which archive backend */
1289 char hal_fsname[0]; /* null-terminated */
1290 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1291 boundaries. See hai_zero */
1292 } __attribute__((packed));
1294 #ifndef HAVE_CFS_SIZE_ROUND
1295 static inline int cfs_size_round (int val)
1297 return (val + 7) & (~0x7);
1299 #define HAVE_CFS_SIZE_ROUND
1302 /* Return pointer to first hai in action list */
1303 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1305 return (struct hsm_action_item *)(hal->hal_fsname +
1306 cfs_size_round(strlen(hal-> \
1310 /* Return pointer to next hai */
1311 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1313 return (struct hsm_action_item *)((char *)hai +
1314 cfs_size_round(hai->hai_len));
1317 /* Return size of an hsm_action_list */
1318 static inline size_t hal_size(struct hsm_action_list *hal)
1322 struct hsm_action_item *hai;
1324 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1325 hai = hai_first(hal);
1326 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1327 sz += cfs_size_round(hai->hai_len);
1333 * describe the attributes to be set on imported file
1335 struct hsm_user_import {
1344 __u32 hui_archive_id;
1347 /* Copytool progress reporting */
1348 #define HP_FLAG_COMPLETED 0x01
1349 #define HP_FLAG_RETRY 0x02
1351 struct hsm_progress {
1354 struct hsm_extent hp_extent;
1356 __u16 hp_errval; /* positive val */
1361 __u64 hc_data_version;
1363 __u16 hc_errval; /* positive val */
1365 struct hsm_action_item hc_hai;
1369 enum llapi_json_types {
1370 LLAPI_JSON_INTEGER = 1,
1376 struct llapi_json_item {
1385 struct llapi_json_item *lji_next;
1388 struct llapi_json_item_list {
1389 int ljil_item_count;
1390 struct llapi_json_item *ljil_items;
1393 /** @} lustreuser */
1395 #endif /* _LUSTRE_USER_H */