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, 2013, 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 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
222 #define LL_IOC_RECREATE_FID _IOW ('f', 157, struct lu_fid)
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
307 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
309 #define LOV_PATTERN_RAID0 0x001
310 #define LOV_PATTERN_RAID1 0x002
311 #define LOV_PATTERN_FIRST 0x100
312 #define LOV_PATTERN_CMOBD 0x200
314 #define LOV_PATTERN_F_MASK 0xffff0000
315 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
316 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
318 #define LOV_MAXPOOLNAME 15
319 #define LOV_POOLNAMEF "%.15s"
321 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
322 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
323 #define LOV_MAX_STRIPE_COUNT_OLD 160
324 /* This calculation is crafted so that input of 4096 will result in 160
325 * which in turn is equal to old maximal stripe count.
326 * XXX: In fact this is too simpified for now, what it also need is to get
327 * ea_type argument to clearly know how much space each stripe consumes.
329 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
330 * allocation that is sufficient for the current generation of systems.
332 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
333 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
334 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
335 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
337 #define XATTR_LUSTRE_PREFIX "lustre."
338 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
340 #define lov_user_ost_data lov_user_ost_data_v1
341 struct lov_user_ost_data_v1 { /* per-stripe data structure */
342 struct ost_id l_ost_oi; /* OST object ID */
343 __u32 l_ost_gen; /* generation of this OST index */
344 __u32 l_ost_idx; /* OST index in LOV */
345 } __attribute__((packed));
347 #define lov_user_md lov_user_md_v1
348 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
349 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
350 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
351 struct ost_id lmm_oi; /* LOV object ID */
352 __u32 lmm_stripe_size; /* size of stripe in bytes */
353 __u16 lmm_stripe_count; /* num stripes in use for this object */
355 __u16 lmm_stripe_offset; /* starting stripe offset in
356 * lmm_objects, use when writing */
357 __u16 lmm_layout_gen; /* layout generation number
358 * used when reading */
360 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
361 } __attribute__((packed, __may_alias__));
363 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
364 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
365 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
366 struct ost_id lmm_oi; /* LOV object ID */
367 __u32 lmm_stripe_size; /* size of stripe in bytes */
368 __u16 lmm_stripe_count; /* num stripes in use for this object */
370 __u16 lmm_stripe_offset; /* starting stripe offset in
371 * lmm_objects, use when writing */
372 __u16 lmm_layout_gen; /* layout generation number
373 * used when reading */
375 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
376 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
377 } __attribute__((packed));
379 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
381 if (lmm_magic == LOV_USER_MAGIC_V3)
382 return sizeof(struct lov_user_md_v3) +
383 stripes * sizeof(struct lov_user_ost_data_v1);
385 return sizeof(struct lov_user_md_v1) +
386 stripes * sizeof(struct lov_user_ost_data_v1);
389 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
390 * use this. It is unsafe to #define those values in this header as it
391 * is possible the application has already #included <sys/stat.h>. */
392 #ifdef HAVE_LOV_USER_MDS_DATA
393 #define lov_user_mds_data lov_user_mds_data_v1
394 struct lov_user_mds_data_v1 {
395 lstat_t lmd_st; /* MDS stat struct */
396 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
397 } __attribute__((packed));
399 struct lov_user_mds_data_v3 {
400 lstat_t lmd_st; /* MDS stat struct */
401 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
402 } __attribute__((packed));
405 struct lmv_user_mds_data {
406 struct lu_fid lum_fid;
412 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
413 LMV_HASH_TYPE_ALL_CHARS = 1,
414 LMV_HASH_TYPE_FNV_1A_64 = 2,
417 #define LMV_HASH_NAME_ALL_CHARS "all_char"
418 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
420 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
421 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
422 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
423 #define lmv_user_md lmv_user_md_v1
424 struct lmv_user_md_v1 {
425 __u32 lum_magic; /* must be the first field */
426 __u32 lum_stripe_count; /* dirstripe count */
427 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
428 __u32 lum_hash_type; /* Dir stripe policy */
429 __u32 lum_type; /* LMV type: default or normal */
433 char lum_pool_name[LOV_MAXPOOLNAME + 1];
434 struct lmv_user_mds_data lum_objects[0];
435 } __attribute__((packed));
437 static inline int lmv_user_md_size(int stripes, int lmm_magic)
439 return sizeof(struct lmv_user_md) +
440 stripes * sizeof(struct lmv_user_mds_data);
443 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
445 struct ll_recreate_obj {
451 __u64 id; /* holds object id */
452 __u32 generation; /* holds object generation */
453 __u32 f_type; /* holds object type or stripe idx when passing it to
454 * OST for saving into EA. */
462 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
463 const struct obd_uuid *u2)
465 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
468 static inline int obd_uuid_empty(struct obd_uuid *uuid)
470 return uuid->uuid[0] == '\0';
473 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
475 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
476 uuid->uuid[sizeof(*uuid) - 1] = '\0';
479 /* For printf's only, make sure uuid is terminated */
480 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
485 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
486 /* Obviously not safe, but for printfs, no real harm done...
487 we're always null-terminated, even in a race. */
488 static char temp[sizeof(*uuid)];
489 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
490 temp[sizeof(*uuid) - 1] = '\0';
493 return (char *)(uuid->uuid);
496 /* Extract fsname from uuid (or target name) of a target
497 e.g. (myfs-OST0007_UUID -> myfs)
498 see also deuuidify. */
499 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
503 strncpy(buf, uuid, buflen - 1);
504 buf[buflen - 1] = '\0';
505 p = strrchr(buf, '-');
510 /* printf display format
511 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
512 #define FID_NOBRACE_LEN 40
513 #define FID_LEN (FID_NOBRACE_LEN + 2)
514 #define DFID_NOBRACE LPX64":0x%x:0x%x"
515 #define DFID "["DFID_NOBRACE"]"
521 /* scanf input parse format -- strip '[' first.
522 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
523 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
524 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
525 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
527 #define SFID "0x"LPX64i":0x%x:0x%x"
534 /********* Quotas **********/
536 #define LUSTRE_QUOTABLOCK_BITS 10
537 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
539 static inline __u64 lustre_stoqb(size_t space)
541 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
544 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
545 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
546 #define Q_GETOINFO 0x800102 /* get obd quota info */
547 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
548 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
550 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
551 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
552 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
553 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
554 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
555 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
556 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
557 /* lustre-specific control commands */
558 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
559 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
561 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
563 struct if_quotacheck {
565 struct obd_uuid obd_uuid;
568 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
571 #define N_PERMS_MAX 64
573 struct perm_downcall_data {
579 struct identity_downcall_data {
586 struct perm_downcall_data idd_perms[N_PERMS_MAX];
590 /* for non-mapped uid/gid */
591 #define NOBODY_UID 99
592 #define NOBODY_GID 99
594 #define INVALID_ID (-1)
603 #ifdef NEED_QUOTA_DEFS
605 #define QIF_BLIMITS 1
607 #define QIF_ILIMITS 4
611 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
612 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
613 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
614 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
617 #endif /* !__KERNEL__ */
619 /* lustre volatile file support
620 * file name header: .^L^S^T^R:volatile"
622 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
623 #define LUSTRE_VOLATILE_HDR_LEN 14
625 typedef enum lustre_quota_version {
627 } lustre_quota_version_t;
629 /* XXX: same as if_dqinfo struct in kernel */
637 /* XXX: same as if_dqblk struct in kernel, plus one padding */
639 __u64 dqb_bhardlimit;
640 __u64 dqb_bsoftlimit;
642 __u64 dqb_ihardlimit;
643 __u64 dqb_isoftlimit;
665 struct obd_dqinfo qc_dqinfo;
666 struct obd_dqblk qc_dqblk;
668 struct obd_uuid obd_uuid;
671 /* swap layout flags */
672 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
673 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
674 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
675 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
677 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
678 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
679 struct lustre_swap_layouts {
688 /********* Changelogs **********/
689 /** Changelog record types */
690 enum changelog_rec_type {
692 CL_CREATE = 1, /* namespace */
693 CL_MKDIR = 2, /* namespace */
694 CL_HARDLINK = 3, /* namespace */
695 CL_SOFTLINK = 4, /* namespace */
696 CL_MKNOD = 5, /* namespace */
697 CL_UNLINK = 6, /* namespace */
698 CL_RMDIR = 7, /* namespace */
699 CL_RENAME = 8, /* namespace */
700 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
701 CL_OPEN = 10, /* not currently used */
702 CL_CLOSE = 11, /* may be written to log only with mtime change */
703 CL_LAYOUT = 12, /* file layout/striping modified */
707 CL_HSM = 16, /* HSM specific events, see flags */
708 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
714 static inline const char *changelog_type2str(int type) {
715 static const char *changelog_str[] = {
716 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
717 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
718 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
721 if (type >= 0 && type < CL_LAST)
722 return changelog_str[type];
726 /* per-record flags */
727 #define CLF_FLAGSHIFT 12
728 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
729 #define CLF_VERMASK (~CLF_FLAGMASK)
730 enum changelog_rec_flags {
731 CLF_VERSION = 0x1000,
734 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
738 /* Anything under the flagmask may be per-type (if desired) */
739 /* Flags for unlink */
740 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
741 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
742 /* HSM cleaning needed */
743 /* Flags for rename */
744 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
746 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
747 * has an archive in backend */
750 /* 12b used (from high weight to low weight):
755 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
756 #define CLF_HSM_ERR_H 6
757 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
758 #define CLF_HSM_EVENT_H 9
759 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
760 #define CLF_HSM_FLAG_H 11
761 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
762 #define CLF_HSM_SPARE_H 15
763 #define CLF_HSM_LAST 15
765 /* Remove bits higher than _h, then extract the value
766 * between _h and _l by shifting lower weigth to bit 0. */
767 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
768 >> (CLF_HSM_LAST - _h + _l))
770 #define CLF_HSM_SUCCESS 0x00
771 #define CLF_HSM_MAXERROR 0x7E
772 #define CLF_HSM_ERROVERFLOW 0x7F
774 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
776 /* 3 bits field => 8 values allowed */
788 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
790 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
794 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
796 *flags |= (he << CLF_HSM_EVENT_L);
799 static inline __u16 hsm_get_cl_flags(int flags)
801 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
804 static inline void hsm_set_cl_flags(int *flags, int bits)
806 *flags |= (bits << CLF_HSM_FLAG_L);
809 static inline int hsm_get_cl_error(int flags)
811 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
814 static inline void hsm_set_cl_error(int *flags, int error)
816 *flags |= (error << CLF_HSM_ERR_L);
819 enum changelog_send_flag {
820 /* Not yet implemented */
821 CHANGELOG_FLAG_FOLLOW = 0x01,
822 /* Blocking IO makes sense in case of slow user parsing of the records,
823 * but it also prevents us from cleaning up if the records are not
825 CHANGELOG_FLAG_BLOCK = 0x02,
826 /* Pack jobid into the changelog records if available. */
827 CHANGELOG_FLAG_JOBID = 0x04,
830 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
831 changelog_rec_offset(CLF_SUPPORTED))
833 /* 31 usable bytes string + null terminator. */
834 #define LUSTRE_JOBID_SIZE 32
836 /* This is the minimal changelog record. It can contain extensions
837 * such as rename fields or process jobid. Its exact content is described
840 * Extensions are packed in the same order as their corresponding flags.
842 struct changelog_rec {
844 __u16 cr_flags; /**< \a changelog_rec_flags */
845 __u32 cr_type; /**< \a changelog_rec_type */
846 __u64 cr_index; /**< changelog record number */
847 __u64 cr_prev; /**< last index for this target fid */
850 lustre_fid cr_tfid; /**< target fid */
851 __u32 cr_markerflags; /**< CL_MARK flags */
853 lustre_fid cr_pfid; /**< parent fid */
856 /* Changelog extension for RENAME. */
857 struct changelog_ext_rename {
858 lustre_fid cr_sfid; /**< source fid, or zero */
859 lustre_fid cr_spfid; /**< source parent fid, or zero */
862 /* Changelog extension to include JOBID. */
863 struct changelog_ext_jobid {
864 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
868 static inline unsigned int changelog_rec_offset(enum changelog_rec_flags crf)
870 unsigned int size = sizeof(struct changelog_rec);
872 if (crf & CLF_RENAME)
873 size += sizeof(struct changelog_ext_rename);
876 size += sizeof(struct changelog_ext_jobid);
881 static inline unsigned int changelog_rec_size(struct changelog_rec *rec)
883 return changelog_rec_offset(rec->cr_flags);
886 static inline unsigned int changelog_rec_varsize(struct changelog_rec *rec)
888 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
892 struct changelog_ext_rename *changelog_rec_rename(struct changelog_rec *rec)
894 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
896 return (struct changelog_ext_rename *)((char *)rec +
897 changelog_rec_offset(crf));
900 /* The jobid follows the rename extension, if present */
902 struct changelog_ext_jobid *changelog_rec_jobid(struct changelog_rec *rec)
904 enum changelog_rec_flags crf = rec->cr_flags &
905 (CLF_VERSION | CLF_RENAME);
907 return (struct changelog_ext_jobid *)((char *)rec +
908 changelog_rec_offset(crf));
911 /* The name follows the rename and jobid extensions, if present */
912 static inline char *changelog_rec_name(struct changelog_rec *rec)
914 return (char *)rec + changelog_rec_offset(rec->cr_flags &
918 static inline int changelog_rec_snamelen(struct changelog_rec *rec)
920 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
923 static inline char *changelog_rec_sname(struct changelog_rec *rec)
925 char *cr_name = changelog_rec_name(rec);
927 return cr_name + strlen(cr_name) + 1;
930 /* Remap a record to the desired format as specified by the crf flags.
931 * The record must be big enough to contain the final remapped version. */
932 static inline void changelog_remap_rec(struct changelog_rec *rec,
933 enum changelog_rec_flags crf)
938 crf &= CLF_SUPPORTED;
940 if ((rec->cr_flags & CLF_SUPPORTED) == crf)
943 if ((crf & CLF_JOBID) && rec->cr_flags & CLF_JOBID) {
944 var_part = (char *)changelog_rec_jobid(rec);
945 var_size = rec->cr_namelen + sizeof(struct changelog_ext_jobid);
947 var_part = changelog_rec_name(rec);
948 var_size = rec->cr_namelen;
951 memmove((char *)rec + changelog_rec_offset(crf & ~CLF_JOBID), var_part,
954 if ((crf & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
955 memset(changelog_rec_rename(rec), 0,
956 sizeof(struct changelog_ext_rename));
958 if ((crf & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
959 memset(changelog_rec_jobid(rec), 0,
960 sizeof(struct changelog_ext_jobid));
962 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf;
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 */
983 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
984 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
987 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
990 #define dot_lustre_name ".lustre"
993 /********* HSM **********/
995 /** HSM per-file state
996 * See HSM_FLAGS below.
999 HS_EXISTS = 0x00000001,
1000 HS_DIRTY = 0x00000002,
1001 HS_RELEASED = 0x00000004,
1002 HS_ARCHIVED = 0x00000008,
1003 HS_NORELEASE = 0x00000010,
1004 HS_NOARCHIVE = 0x00000020,
1005 HS_LOST = 0x00000040,
1008 /* HSM user-setable flags. */
1009 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1011 /* Other HSM flags. */
1012 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1015 * All HSM-related possible flags that could be applied to a file.
1016 * This should be kept in sync with hsm_states.
1018 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1021 * HSMÂ request progress state
1023 enum hsm_progress_states {
1030 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1033 case HPS_WAITING: return "waiting";
1034 case HPS_RUNNING: return "running";
1035 case HPS_DONE: return "done";
1036 default: return "unknown";
1043 } __attribute__((packed));
1046 * Current HSM states of a Lustre file.
1048 * This structure purpose is to be sent to user-space mainly. It describes the
1049 * current HSM flags and in-progress action.
1051 struct hsm_user_state {
1052 /** Current HSM states, from enum hsm_states. */
1054 __u32 hus_archive_id;
1055 /** The current undergoing action, if there is one */
1056 __u32 hus_in_progress_state;
1057 __u32 hus_in_progress_action;
1058 struct hsm_extent hus_in_progress_location;
1059 char hus_extended_info[];
1062 struct hsm_state_set_ioc {
1063 struct lu_fid hssi_fid;
1065 __u64 hssi_clearmask;
1069 * This structure describes the current in-progress action for a file.
1070 * it is retuned to user space and send over the wire
1072 struct hsm_current_action {
1073 /** The current undergoing action, if there is one */
1074 /* state is one of hsm_progress_states */
1076 /* action is one of hsm_user_action */
1078 struct hsm_extent hca_location;
1081 /***** HSM user requests ******/
1082 /* User-generated (lfs/ioctl) request types */
1083 enum hsm_user_action {
1084 HUA_NONE = 1, /* no action (noop) */
1085 HUA_ARCHIVE = 10, /* copy to hsm */
1086 HUA_RESTORE = 11, /* prestage */
1087 HUA_RELEASE = 12, /* drop ost objects */
1088 HUA_REMOVE = 13, /* remove from archive */
1089 HUA_CANCEL = 14 /* cancel a request */
1092 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1095 case HUA_NONE: return "NOOP";
1096 case HUA_ARCHIVE: return "ARCHIVE";
1097 case HUA_RESTORE: return "RESTORE";
1098 case HUA_RELEASE: return "RELEASE";
1099 case HUA_REMOVE: return "REMOVE";
1100 case HUA_CANCEL: return "CANCEL";
1101 default: return "UNKNOWN";
1106 * List of hr_flags (bit field)
1108 #define HSM_FORCE_ACTION 0x0001
1109 /* used by CT, connot be set by user */
1110 #define HSM_GHOST_COPY 0x0002
1113 * Contains all the fixed part of struct hsm_user_request.
1116 struct hsm_request {
1117 __u32 hr_action; /* enum hsm_user_action */
1118 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1119 __u64 hr_flags; /* request flags */
1120 __u32 hr_itemcount; /* item count in hur_user_item vector */
1124 struct hsm_user_item {
1126 struct hsm_extent hui_extent;
1127 } __attribute__((packed));
1129 struct hsm_user_request {
1130 struct hsm_request hur_request;
1131 struct hsm_user_item hur_user_item[0];
1132 /* extra data blob at end of struct (after all
1133 * hur_user_items), only use helpers to access it
1135 } __attribute__((packed));
1137 /** Return pointer to data field in a hsm user request */
1138 static inline void *hur_data(struct hsm_user_request *hur)
1140 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1144 * Compute the current length of the provided hsm_user_request. This returns -1
1145 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1147 * return -1 on bounds check error.
1149 static inline ssize_t hur_len(struct hsm_user_request *hur)
1153 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1154 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1155 (__u64)hur->hur_request.hr_itemcount *
1156 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1158 if (size > ~((size_t)0)>>1)
1164 /****** HSM RPCs to copytool *****/
1165 /* Message types the copytool may receive */
1166 enum hsm_message_type {
1167 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1170 /* Actions the copytool may be instructed to take for a given action_item */
1171 enum hsm_copytool_action {
1172 HSMA_NONE = 10, /* no action */
1173 HSMA_ARCHIVE = 20, /* arbitrary offset */
1179 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1182 case HSMA_NONE: return "NOOP";
1183 case HSMA_ARCHIVE: return "ARCHIVE";
1184 case HSMA_RESTORE: return "RESTORE";
1185 case HSMA_REMOVE: return "REMOVE";
1186 case HSMA_CANCEL: return "CANCEL";
1187 default: return "UNKNOWN";
1191 /* Copytool item action description */
1192 struct hsm_action_item {
1193 __u32 hai_len; /* valid size of this struct */
1194 __u32 hai_action; /* hsm_copytool_action, but use known size */
1195 lustre_fid hai_fid; /* Lustre FID to operated on */
1196 lustre_fid hai_dfid; /* fid used for data access */
1197 struct hsm_extent hai_extent; /* byte range to operate on */
1198 __u64 hai_cookie; /* action cookie from coordinator */
1199 __u64 hai_gid; /* grouplock id */
1200 char hai_data[0]; /* variable length */
1201 } __attribute__((packed));
1204 * helper function which print in hexa the first bytes of
1206 * \param hai [IN] record to print
1207 * \param buffer [OUT] output buffer
1208 * \param len [IN] max buffer len
1211 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1212 char *buffer, int len)
1214 int i, sz, data_len;
1219 data_len = hai->hai_len - sizeof(*hai);
1220 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1224 cnt = snprintf(ptr, sz, "%.2X",
1225 (unsigned char)hai->hai_data[i]);
1233 /* Copytool action list */
1234 #define HAL_VERSION 1
1235 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1236 struct hsm_action_list {
1238 __u32 hal_count; /* number of hai's to follow */
1239 __u64 hal_compound_id; /* returned by coordinator */
1241 __u32 hal_archive_id; /* which archive backend */
1243 char hal_fsname[0]; /* null-terminated */
1244 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1245 boundaries. See hai_zero */
1246 } __attribute__((packed));
1248 #ifndef HAVE_CFS_SIZE_ROUND
1249 static inline int cfs_size_round (int val)
1251 return (val + 7) & (~0x7);
1253 #define HAVE_CFS_SIZE_ROUND
1256 /* Return pointer to first hai in action list */
1257 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1259 return (struct hsm_action_item *)(hal->hal_fsname +
1260 cfs_size_round(strlen(hal-> \
1264 /* Return pointer to next hai */
1265 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1267 return (struct hsm_action_item *)((char *)hai +
1268 cfs_size_round(hai->hai_len));
1271 /* Return size of an hsm_action_list */
1272 static inline size_t hal_size(struct hsm_action_list *hal)
1276 struct hsm_action_item *hai;
1278 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1279 hai = hai_first(hal);
1280 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1281 sz += cfs_size_round(hai->hai_len);
1287 * describe the attributes to be set on imported file
1289 struct hsm_user_import {
1298 __u32 hui_archive_id;
1301 /* Copytool progress reporting */
1302 #define HP_FLAG_COMPLETED 0x01
1303 #define HP_FLAG_RETRY 0x02
1305 struct hsm_progress {
1308 struct hsm_extent hp_extent;
1310 __u16 hp_errval; /* positive val */
1315 __u64 hc_data_version;
1317 __u16 hc_errval; /* positive val */
1319 struct hsm_action_item hc_hai;
1323 enum llapi_json_types {
1324 LLAPI_JSON_INTEGER = 1,
1330 struct llapi_json_item {
1339 struct llapi_json_item *lji_next;
1342 struct llapi_json_item_list {
1343 int ljil_item_count;
1344 struct llapi_json_item *ljil_items;
1347 /** @} lustreuser */
1349 #endif /* _LUSTRE_USER_H */