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__) || \
67 typedef struct stat lstat_t;
68 # define lstat_f lstat
69 # define fstat_f fstat
70 # define fstatat_f fstatat
71 # define HAVE_LOV_USER_MDS_DATA
72 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
73 typedef struct stat64 lstat_t;
74 # define lstat_f lstat64
75 # define fstat_f fstat64
76 # define fstatat_f fstatat64
77 # define HAVE_LOV_USER_MDS_DATA
80 #define LUSTRE_EOF 0xffffffffffffffffULL
83 #define LL_SUPER_MAGIC 0x0BD00BD0
85 #ifndef FSFILT_IOC_GETFLAGS
86 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
87 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
88 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
89 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
90 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
91 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
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)
224 struct ll_futimes_3 {
226 __u64 lfu_atime_nsec;
228 __u64 lfu_mtime_nsec;
230 __u64 lfu_ctime_nsec;
234 * The ioctl naming rules:
235 * LL_* - works on the currently opened filehandle instead of parent dir
236 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
237 * *_MDC_* - gets/sets data related to MDC
238 * *_LOV_* - gets/sets data related to OSC/LOV
239 * *FILE* - called on parent dir and passes in a filename
240 * *STRIPE* - set/get lov_user_md
241 * *INFO - set/get lov_user_mds_data
243 /* lustre_ioctl.h 101-150 */
244 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
245 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
246 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
247 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
248 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
249 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
250 /* LL_IOC_RECREATE_OBJ 157 obsolete */
251 /* LL_IOC_RECREATE_FID 157 obsolete */
252 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
253 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
254 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
255 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
256 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
257 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
258 /* IOC_LOV_GETINFO 165 obsolete */
259 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
260 #define LL_IOC_RMTACL _IOW ('f', 167, long)
261 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
262 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
263 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
264 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
265 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
266 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
267 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
268 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
269 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
270 /* lustre_ioctl.h 177-210 */
271 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
272 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
273 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
274 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
275 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
276 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
277 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
278 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
279 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
280 struct lustre_swap_layouts)
281 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
282 struct hsm_current_action)
283 /* lustre_ioctl.h 221-232 */
284 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
285 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
286 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
287 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
288 #define LL_IOC_GET_LEASE _IO('f', 244)
289 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
290 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
291 #define LL_IOC_MIGRATE _IOR('f', 247, int)
292 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
293 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
295 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
297 LL_LEASE_RDLCK = 0x1,
298 LL_LEASE_WRLCK = 0x2,
299 LL_LEASE_UNLCK = 0x4,
302 #define LL_STATFS_LMV 1
303 #define LL_STATFS_LOV 2
304 #define LL_STATFS_NODELAY 4
306 #define IOC_MDC_TYPE 'i'
307 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
308 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
309 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
310 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
312 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
314 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
315 * files, but are unlikely to be used in practice and are not harmful if
316 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
317 * devices and are safe for use on new files. See LU-4209. */
318 /* To be compatible with old statically linked binary we keep the check for
319 * the older 0100000000 flag. This is already removed upstream. LU-812. */
320 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
321 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
322 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
323 O_LOV_DELAY_CREATE_MASK)
325 #define LL_FILE_IGNORE_LOCK 0x00000001
326 #define LL_FILE_GROUP_LOCKED 0x00000002
327 #define LL_FILE_READAHEA 0x00000004
328 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
329 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
330 #define LL_FILE_RMTACL 0x00000020
332 #define LOV_USER_MAGIC_V1 0x0BD10BD0
333 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
334 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
335 #define LOV_USER_MAGIC_V3 0x0BD30BD0
336 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
337 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
339 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
341 #define LOV_PATTERN_RAID0 0x001
342 #define LOV_PATTERN_RAID1 0x002
343 #define LOV_PATTERN_FIRST 0x100
344 #define LOV_PATTERN_CMOBD 0x200
346 #define LOV_PATTERN_F_MASK 0xffff0000
347 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
348 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
350 #define LOV_MAXPOOLNAME 15
351 #define LOV_POOLNAMEF "%.15s"
353 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
354 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
355 #define LOV_MAX_STRIPE_COUNT_OLD 160
356 /* This calculation is crafted so that input of 4096 will result in 160
357 * which in turn is equal to old maximal stripe count.
358 * XXX: In fact this is too simpified for now, what it also need is to get
359 * ea_type argument to clearly know how much space each stripe consumes.
361 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
362 * allocation that is sufficient for the current generation of systems.
364 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
365 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
366 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
367 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
369 #define XATTR_LUSTRE_PREFIX "lustre."
370 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
372 #define lov_user_ost_data lov_user_ost_data_v1
373 struct lov_user_ost_data_v1 { /* per-stripe data structure */
374 struct ost_id l_ost_oi; /* OST object ID */
375 __u32 l_ost_gen; /* generation of this OST index */
376 __u32 l_ost_idx; /* OST index in LOV */
377 } __attribute__((packed));
379 #define lov_user_md lov_user_md_v1
380 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
381 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
382 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
383 struct ost_id lmm_oi; /* LOV object ID */
384 __u32 lmm_stripe_size; /* size of stripe in bytes */
385 __u16 lmm_stripe_count; /* num stripes in use for this object */
387 __u16 lmm_stripe_offset; /* starting stripe offset in
388 * lmm_objects, use when writing */
389 __u16 lmm_layout_gen; /* layout generation number
390 * used when reading */
392 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
393 } __attribute__((packed, __may_alias__));
395 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
396 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
397 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
398 struct ost_id lmm_oi; /* LOV object ID */
399 __u32 lmm_stripe_size; /* size of stripe in bytes */
400 __u16 lmm_stripe_count; /* num stripes in use for this object */
402 __u16 lmm_stripe_offset; /* starting stripe offset in
403 * lmm_objects, use when writing */
404 __u16 lmm_layout_gen; /* layout generation number
405 * used when reading */
407 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
408 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
409 } __attribute__((packed));
411 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
413 if (lmm_magic == LOV_USER_MAGIC_V1)
414 return sizeof(struct lov_user_md_v1) +
415 stripes * sizeof(struct lov_user_ost_data_v1);
416 return sizeof(struct lov_user_md_v3) +
417 stripes * sizeof(struct lov_user_ost_data_v1);
420 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
421 * use this. It is unsafe to #define those values in this header as it
422 * is possible the application has already #included <sys/stat.h>. */
423 #ifdef HAVE_LOV_USER_MDS_DATA
424 #define lov_user_mds_data lov_user_mds_data_v1
425 struct lov_user_mds_data_v1 {
426 lstat_t lmd_st; /* MDS stat struct */
427 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
428 } __attribute__((packed));
430 struct lov_user_mds_data_v3 {
431 lstat_t lmd_st; /* MDS stat struct */
432 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
433 } __attribute__((packed));
436 struct lmv_user_mds_data {
437 struct lu_fid lum_fid;
443 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
444 LMV_HASH_TYPE_ALL_CHARS = 1,
445 LMV_HASH_TYPE_FNV_1A_64 = 2,
448 #define LMV_HASH_NAME_ALL_CHARS "all_char"
449 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
451 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
452 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
453 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
454 #define lmv_user_md lmv_user_md_v1
455 struct lmv_user_md_v1 {
456 __u32 lum_magic; /* must be the first field */
457 __u32 lum_stripe_count; /* dirstripe count */
458 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
459 __u32 lum_hash_type; /* Dir stripe policy */
460 __u32 lum_type; /* LMV type: default or normal */
464 char lum_pool_name[LOV_MAXPOOLNAME + 1];
465 struct lmv_user_mds_data lum_objects[0];
466 } __attribute__((packed));
468 static inline int lmv_user_md_size(int stripes, int lmm_magic)
470 return sizeof(struct lmv_user_md) +
471 stripes * sizeof(struct lmv_user_mds_data);
474 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
476 struct ll_recreate_obj {
482 __u64 id; /* holds object id */
483 __u32 generation; /* holds object generation */
484 __u32 f_type; /* holds object type or stripe idx when passing it to
485 * OST for saving into EA. */
493 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
494 const struct obd_uuid *u2)
496 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
499 static inline int obd_uuid_empty(struct obd_uuid *uuid)
501 return uuid->uuid[0] == '\0';
504 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
506 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
507 uuid->uuid[sizeof(*uuid) - 1] = '\0';
510 /* For printf's only, make sure uuid is terminated */
511 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
516 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
517 /* Obviously not safe, but for printfs, no real harm done...
518 we're always null-terminated, even in a race. */
519 static char temp[sizeof(*uuid)];
520 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
521 temp[sizeof(*uuid) - 1] = '\0';
524 return (char *)(uuid->uuid);
527 /* Extract fsname from uuid (or target name) of a target
528 e.g. (myfs-OST0007_UUID -> myfs)
529 see also deuuidify. */
530 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
534 strncpy(buf, uuid, buflen - 1);
535 buf[buflen - 1] = '\0';
536 p = strrchr(buf, '-');
541 /* printf display format
542 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
543 #define FID_NOBRACE_LEN 40
544 #define FID_LEN (FID_NOBRACE_LEN + 2)
545 #define DFID_NOBRACE LPX64":0x%x:0x%x"
546 #define DFID "["DFID_NOBRACE"]"
552 /* scanf input parse format -- strip '[' first.
553 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
554 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
555 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
556 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
558 #define SFID "0x"LPX64i":0x%x:0x%x"
565 /********* Quotas **********/
567 #define LUSTRE_QUOTABLOCK_BITS 10
568 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
570 static inline __u64 lustre_stoqb(size_t space)
572 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
575 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
576 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
577 #define Q_GETOINFO 0x800102 /* get obd quota info */
578 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
579 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
581 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
582 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
583 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
584 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
585 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
586 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
587 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
588 /* lustre-specific control commands */
589 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
590 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
592 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
594 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
597 #define N_PERMS_MAX 64
599 struct perm_downcall_data {
605 struct identity_downcall_data {
612 struct perm_downcall_data idd_perms[N_PERMS_MAX];
616 /* for non-mapped uid/gid */
617 #define NOBODY_UID 99
618 #define NOBODY_GID 99
620 #define INVALID_ID (-1)
629 #ifdef NEED_QUOTA_DEFS
631 #define QIF_BLIMITS 1
633 #define QIF_ILIMITS 4
637 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
638 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
639 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
640 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
643 #endif /* !__KERNEL__ */
645 /* lustre volatile file support
646 * file name header: .^L^S^T^R:volatile"
648 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
649 #define LUSTRE_VOLATILE_HDR_LEN 14
651 typedef enum lustre_quota_version {
653 } lustre_quota_version_t;
655 /* XXX: same as if_dqinfo struct in kernel */
663 /* XXX: same as if_dqblk struct in kernel, plus one padding */
665 __u64 dqb_bhardlimit;
666 __u64 dqb_bsoftlimit;
668 __u64 dqb_ihardlimit;
669 __u64 dqb_isoftlimit;
691 struct obd_dqinfo qc_dqinfo;
692 struct obd_dqblk qc_dqblk;
694 struct obd_uuid obd_uuid;
697 /* swap layout flags */
698 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
699 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
700 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
701 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
702 #define SWAP_LAYOUTS_CLOSE (1 << 4)
704 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
705 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
706 struct lustre_swap_layouts {
715 /********* Changelogs **********/
716 /** Changelog record types */
717 enum changelog_rec_type {
719 CL_CREATE = 1, /* namespace */
720 CL_MKDIR = 2, /* namespace */
721 CL_HARDLINK = 3, /* namespace */
722 CL_SOFTLINK = 4, /* namespace */
723 CL_MKNOD = 5, /* namespace */
724 CL_UNLINK = 6, /* namespace */
725 CL_RMDIR = 7, /* namespace */
726 CL_RENAME = 8, /* namespace */
727 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
728 CL_OPEN = 10, /* not currently used */
729 CL_CLOSE = 11, /* may be written to log only with mtime change */
730 CL_LAYOUT = 12, /* file layout/striping modified */
734 CL_HSM = 16, /* HSM specific events, see flags */
735 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
741 static inline const char *changelog_type2str(int type) {
742 static const char *changelog_str[] = {
743 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
744 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
745 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
748 if (type >= 0 && type < CL_LAST)
749 return changelog_str[type];
753 /* per-record flags */
754 #define CLF_FLAGSHIFT 12
755 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
756 #define CLF_VERMASK (~CLF_FLAGMASK)
757 enum changelog_rec_flags {
758 CLF_VERSION = 0x1000,
761 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
765 /* Anything under the flagmask may be per-type (if desired) */
766 /* Flags for unlink */
767 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
768 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
769 /* HSM cleaning needed */
770 /* Flags for rename */
771 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
773 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
774 * has an archive in backend */
777 /* 12b used (from high weight to low weight):
782 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
783 #define CLF_HSM_ERR_H 6
784 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
785 #define CLF_HSM_EVENT_H 9
786 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
787 #define CLF_HSM_FLAG_H 11
788 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
789 #define CLF_HSM_SPARE_H 15
790 #define CLF_HSM_LAST 15
792 /* Remove bits higher than _h, then extract the value
793 * between _h and _l by shifting lower weigth to bit 0. */
794 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
795 >> (CLF_HSM_LAST - _h + _l))
797 #define CLF_HSM_SUCCESS 0x00
798 #define CLF_HSM_MAXERROR 0x7E
799 #define CLF_HSM_ERROVERFLOW 0x7F
801 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
803 /* 3 bits field => 8 values allowed */
815 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
817 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
821 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
823 *flags |= (he << CLF_HSM_EVENT_L);
826 static inline __u16 hsm_get_cl_flags(int flags)
828 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
831 static inline void hsm_set_cl_flags(int *flags, int bits)
833 *flags |= (bits << CLF_HSM_FLAG_L);
836 static inline int hsm_get_cl_error(int flags)
838 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
841 static inline void hsm_set_cl_error(int *flags, int error)
843 *flags |= (error << CLF_HSM_ERR_L);
846 enum changelog_send_flag {
847 /* Not yet implemented */
848 CHANGELOG_FLAG_FOLLOW = 0x01,
849 /* Blocking IO makes sense in case of slow user parsing of the records,
850 * but it also prevents us from cleaning up if the records are not
852 CHANGELOG_FLAG_BLOCK = 0x02,
853 /* Pack jobid into the changelog records if available. */
854 CHANGELOG_FLAG_JOBID = 0x04,
857 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
858 changelog_rec_offset(CLF_SUPPORTED))
860 /* 31 usable bytes string + null terminator. */
861 #define LUSTRE_JOBID_SIZE 32
863 /* This is the minimal changelog record. It can contain extensions
864 * such as rename fields or process jobid. Its exact content is described
867 * Extensions are packed in the same order as their corresponding flags.
869 struct changelog_rec {
871 __u16 cr_flags; /**< \a changelog_rec_flags */
872 __u32 cr_type; /**< \a changelog_rec_type */
873 __u64 cr_index; /**< changelog record number */
874 __u64 cr_prev; /**< last index for this target fid */
877 lustre_fid cr_tfid; /**< target fid */
878 __u32 cr_markerflags; /**< CL_MARK flags */
880 lustre_fid cr_pfid; /**< parent fid */
883 /* Changelog extension for RENAME. */
884 struct changelog_ext_rename {
885 lustre_fid cr_sfid; /**< source fid, or zero */
886 lustre_fid cr_spfid; /**< source parent fid, or zero */
889 /* Changelog extension to include JOBID. */
890 struct changelog_ext_jobid {
891 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
895 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
897 size_t size = sizeof(struct changelog_rec);
899 if (crf & CLF_RENAME)
900 size += sizeof(struct changelog_ext_rename);
903 size += sizeof(struct changelog_ext_jobid);
908 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
910 return changelog_rec_offset(rec->cr_flags);
913 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
915 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
919 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
921 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
923 return (struct changelog_ext_rename *)((char *)rec +
924 changelog_rec_offset(crf));
927 /* The jobid follows the rename extension, if present */
929 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
931 enum changelog_rec_flags crf = rec->cr_flags &
932 (CLF_VERSION | CLF_RENAME);
934 return (struct changelog_ext_jobid *)((char *)rec +
935 changelog_rec_offset(crf));
938 /* The name follows the rename and jobid extensions, if present */
939 static inline char *changelog_rec_name(const struct changelog_rec *rec)
941 return (char *)rec + changelog_rec_offset(rec->cr_flags &
945 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
947 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
950 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
952 char *cr_name = changelog_rec_name(rec);
954 return cr_name + strlen(cr_name) + 1;
958 * Remap a record to the desired format as specified by the crf flags.
959 * The record must be big enough to contain the final remapped version.
960 * Superfluous extension fields are removed and missing ones are added
961 * and zeroed. The flags of the record are updated accordingly.
963 * The jobid and rename extensions can be added to a record, to match the
964 * format an application expects, typically. In this case, the newly added
965 * fields will be zeroed.
966 * The Jobid field can be removed, to guarantee compatibility with older
967 * clients that don't expect this field in the records they process.
969 * The following assumptions are being made:
970 * - CLF_RENAME will not be removed
971 * - CLF_JOBID will not be added without CLF_RENAME being added too
973 * @param[in,out] rec The record to remap.
974 * @param[in] crf_wanted Flags describing the desired extensions.
976 static inline void changelog_remap_rec(struct changelog_rec *rec,
977 enum changelog_rec_flags crf_wanted)
982 crf_wanted &= CLF_SUPPORTED;
984 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
987 /* First move the variable-length name field */
988 memmove((char *)rec + changelog_rec_offset(crf_wanted),
989 changelog_rec_name(rec), rec->cr_namelen);
991 /* Locations of jobid and rename extensions in the remapped record */
992 jid_mov = (char *)rec +
993 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
994 rnm_mov = (char *)rec +
995 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
997 /* Move the extension fields to the desired positions */
998 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
999 memmove(jid_mov, changelog_rec_jobid(rec),
1000 sizeof(struct changelog_ext_jobid));
1002 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1003 memmove(rnm_mov, changelog_rec_rename(rec),
1004 sizeof(struct changelog_ext_rename));
1006 /* Clear newly added fields */
1007 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1008 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1010 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1011 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1013 /* Update the record's flags accordingly */
1014 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1017 struct ioc_changelog {
1024 enum changelog_message_type {
1025 CL_RECORD = 10, /* message is a changelog_rec */
1026 CL_EOF = 11, /* at end of current changelog */
1029 /********* Misc **********/
1031 struct ioc_data_version {
1033 __u64 idv_flags; /* See LL_DV_xxx */
1035 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
1036 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
1039 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1042 #define dot_lustre_name ".lustre"
1045 /********* HSM **********/
1047 /** HSM per-file state
1048 * See HSM_FLAGS below.
1051 HS_EXISTS = 0x00000001,
1052 HS_DIRTY = 0x00000002,
1053 HS_RELEASED = 0x00000004,
1054 HS_ARCHIVED = 0x00000008,
1055 HS_NORELEASE = 0x00000010,
1056 HS_NOARCHIVE = 0x00000020,
1057 HS_LOST = 0x00000040,
1060 /* HSM user-setable flags. */
1061 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1063 /* Other HSM flags. */
1064 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1067 * All HSM-related possible flags that could be applied to a file.
1068 * This should be kept in sync with hsm_states.
1070 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1073 * HSMÂ request progress state
1075 enum hsm_progress_states {
1082 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1085 case HPS_WAITING: return "waiting";
1086 case HPS_RUNNING: return "running";
1087 case HPS_DONE: return "done";
1088 default: return "unknown";
1095 } __attribute__((packed));
1098 * Current HSM states of a Lustre file.
1100 * This structure purpose is to be sent to user-space mainly. It describes the
1101 * current HSM flags and in-progress action.
1103 struct hsm_user_state {
1104 /** Current HSM states, from enum hsm_states. */
1106 __u32 hus_archive_id;
1107 /** The current undergoing action, if there is one */
1108 __u32 hus_in_progress_state;
1109 __u32 hus_in_progress_action;
1110 struct hsm_extent hus_in_progress_location;
1111 char hus_extended_info[];
1114 struct hsm_state_set_ioc {
1115 struct lu_fid hssi_fid;
1117 __u64 hssi_clearmask;
1121 * This structure describes the current in-progress action for a file.
1122 * it is retuned to user space and send over the wire
1124 struct hsm_current_action {
1125 /** The current undergoing action, if there is one */
1126 /* state is one of hsm_progress_states */
1128 /* action is one of hsm_user_action */
1130 struct hsm_extent hca_location;
1133 /***** HSM user requests ******/
1134 /* User-generated (lfs/ioctl) request types */
1135 enum hsm_user_action {
1136 HUA_NONE = 1, /* no action (noop) */
1137 HUA_ARCHIVE = 10, /* copy to hsm */
1138 HUA_RESTORE = 11, /* prestage */
1139 HUA_RELEASE = 12, /* drop ost objects */
1140 HUA_REMOVE = 13, /* remove from archive */
1141 HUA_CANCEL = 14 /* cancel a request */
1144 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1147 case HUA_NONE: return "NOOP";
1148 case HUA_ARCHIVE: return "ARCHIVE";
1149 case HUA_RESTORE: return "RESTORE";
1150 case HUA_RELEASE: return "RELEASE";
1151 case HUA_REMOVE: return "REMOVE";
1152 case HUA_CANCEL: return "CANCEL";
1153 default: return "UNKNOWN";
1158 * List of hr_flags (bit field)
1160 #define HSM_FORCE_ACTION 0x0001
1161 /* used by CT, cannot be set by user */
1162 #define HSM_GHOST_COPY 0x0002
1165 * Contains all the fixed part of struct hsm_user_request.
1168 struct hsm_request {
1169 __u32 hr_action; /* enum hsm_user_action */
1170 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1171 __u64 hr_flags; /* request flags */
1172 __u32 hr_itemcount; /* item count in hur_user_item vector */
1176 struct hsm_user_item {
1178 struct hsm_extent hui_extent;
1179 } __attribute__((packed));
1181 struct hsm_user_request {
1182 struct hsm_request hur_request;
1183 struct hsm_user_item hur_user_item[0];
1184 /* extra data blob at end of struct (after all
1185 * hur_user_items), only use helpers to access it
1187 } __attribute__((packed));
1189 /** Return pointer to data field in a hsm user request */
1190 static inline void *hur_data(struct hsm_user_request *hur)
1192 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1196 * Compute the current length of the provided hsm_user_request. This returns -1
1197 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1199 * return -1 on bounds check error.
1201 static inline ssize_t hur_len(struct hsm_user_request *hur)
1205 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1206 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1207 (__u64)hur->hur_request.hr_itemcount *
1208 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1210 if (size != (ssize_t)size)
1216 /****** HSM RPCs to copytool *****/
1217 /* Message types the copytool may receive */
1218 enum hsm_message_type {
1219 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1222 /* Actions the copytool may be instructed to take for a given action_item */
1223 enum hsm_copytool_action {
1224 HSMA_NONE = 10, /* no action */
1225 HSMA_ARCHIVE = 20, /* arbitrary offset */
1231 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1234 case HSMA_NONE: return "NOOP";
1235 case HSMA_ARCHIVE: return "ARCHIVE";
1236 case HSMA_RESTORE: return "RESTORE";
1237 case HSMA_REMOVE: return "REMOVE";
1238 case HSMA_CANCEL: return "CANCEL";
1239 default: return "UNKNOWN";
1243 /* Copytool item action description */
1244 struct hsm_action_item {
1245 __u32 hai_len; /* valid size of this struct */
1246 __u32 hai_action; /* hsm_copytool_action, but use known size */
1247 lustre_fid hai_fid; /* Lustre FID to operate on */
1248 lustre_fid hai_dfid; /* fid used for data access */
1249 struct hsm_extent hai_extent; /* byte range to operate on */
1250 __u64 hai_cookie; /* action cookie from coordinator */
1251 __u64 hai_gid; /* grouplock id */
1252 char hai_data[0]; /* variable length */
1253 } __attribute__((packed));
1256 * helper function which print in hexa the first bytes of
1258 * \param hai [IN] record to print
1259 * \param buffer [OUT] output buffer
1260 * \param len [IN] max buffer len
1263 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1264 char *buffer, int len)
1266 int i, sz, data_len;
1271 data_len = hai->hai_len - sizeof(*hai);
1272 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1276 cnt = snprintf(ptr, sz, "%.2X",
1277 (unsigned char)hai->hai_data[i]);
1285 /* Copytool action list */
1286 #define HAL_VERSION 1
1287 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1288 struct hsm_action_list {
1290 __u32 hal_count; /* number of hai's to follow */
1291 __u64 hal_compound_id; /* returned by coordinator */
1293 __u32 hal_archive_id; /* which archive backend */
1295 char hal_fsname[0]; /* null-terminated */
1296 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1297 boundaries. See hai_zero */
1298 } __attribute__((packed));
1300 #ifndef HAVE_CFS_SIZE_ROUND
1301 static inline int cfs_size_round (int val)
1303 return (val + 7) & (~0x7);
1305 #define HAVE_CFS_SIZE_ROUND
1308 /* Return pointer to first hai in action list */
1309 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1311 return (struct hsm_action_item *)(hal->hal_fsname +
1312 cfs_size_round(strlen(hal-> \
1316 /* Return pointer to next hai */
1317 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1319 return (struct hsm_action_item *)((char *)hai +
1320 cfs_size_round(hai->hai_len));
1323 /* Return size of an hsm_action_list */
1324 static inline size_t hal_size(struct hsm_action_list *hal)
1328 struct hsm_action_item *hai;
1330 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1331 hai = hai_first(hal);
1332 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1333 sz += cfs_size_round(hai->hai_len);
1339 * describe the attributes to be set on imported file
1341 struct hsm_user_import {
1350 __u32 hui_archive_id;
1353 /* Copytool progress reporting */
1354 #define HP_FLAG_COMPLETED 0x01
1355 #define HP_FLAG_RETRY 0x02
1357 struct hsm_progress {
1360 struct hsm_extent hp_extent;
1362 __u16 hp_errval; /* positive val */
1367 __u64 hc_data_version;
1369 __u16 hc_errval; /* positive val */
1371 struct hsm_action_item hc_hai;
1375 enum llapi_json_types {
1376 LLAPI_JSON_INTEGER = 1,
1382 struct llapi_json_item {
1391 struct llapi_json_item *lji_next;
1394 struct llapi_json_item_list {
1395 int ljil_item_count;
1396 struct llapi_json_item *ljil_items;
1399 /** @} lustreuser */
1401 #endif /* _LUSTRE_USER_H */