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 HAVE_LOV_USER_MDS_DATA
69 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
70 typedef struct stat64 lstat_t;
71 # define lstat_f lstat64
72 # define HAVE_LOV_USER_MDS_DATA
75 #define LUSTRE_EOF 0xffffffffffffffffULL
78 #define LL_SUPER_MAGIC 0x0BD00BD0
80 #ifndef FSFILT_IOC_GETFLAGS
81 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
82 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
83 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
84 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
85 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
86 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
87 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
90 /* FIEMAP flags supported by Lustre */
91 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
93 enum obd_statfs_state {
94 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
95 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
96 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
97 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
98 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
112 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
113 __u32 os_fprecreated; /* objs available now to the caller */
114 /* used in QoS code to find preferred
129 * FID is a cluster-wide unique identifier of a file or an object (stripe).
130 * FIDs are never reused.
134 * FID sequence. Sequence is a unit of migration: all files (objects)
135 * with FIDs from a given sequence are stored on the same server.
136 * Lustre should support 2^64 objects, so even if each sequence
137 * has only a single object we can still enumerate 2^64 objects.
140 /* FID number within sequence. */
143 * FID version, used to distinguish different versions (in the sense
144 * of snapshots, etc.) of the same file system object. Not currently
150 static inline bool fid_is_zero(const struct lu_fid *fid)
152 return fid->f_seq == 0 && fid->f_oid == 0;
155 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
156 * MDT-object's FID::f_ver, instead it is the OST-object index in its
157 * parent MDT-object's layout EA. */
158 #define f_stripe_idx f_ver
161 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
164 /* keep this one for compatibility */
165 struct filter_fid_old {
166 struct lu_fid ff_parent;
171 /* Userspace should treat lu_fid as opaque, and only use the following methods
172 * to print or parse them. Other functions (e.g. compare, swab) could be moved
173 * here from lustre_idl.h if needed. */
174 typedef struct lu_fid lustre_fid;
177 * Following struct for object attributes, that will be kept inode's EA.
178 * Introduced in 2.0 release (please see b15993, for details)
179 * Added to all objects since Lustre 2.4 as contains self FID
181 struct lustre_mdt_attrs {
183 * Bitfield for supported data in this structure. From enum lma_compat.
184 * lma_self_fid and lma_flags are always available.
188 * Per-file incompat feature list. Lustre version should support all
189 * flags set in this field. The supported feature mask is available in
193 /** FID of this inode */
194 struct lu_fid lma_self_fid;
198 * Prior to 2.4, the LMA structure also included SOM attributes which has since
199 * been moved to a dedicated xattr
200 * lma_flags was also removed because of lma_compat/incompat fields.
202 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
205 * OST object IDentifier.
213 struct lu_fid oi_fid;
217 #define DOSTID LPX64":"LPU64
218 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
221 * The ioctl naming rules:
222 * LL_* - works on the currently opened filehandle instead of parent dir
223 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
224 * *_MDC_* - gets/sets data related to MDC
225 * *_LOV_* - gets/sets data related to OSC/LOV
226 * *FILE* - called on parent dir and passes in a filename
227 * *STRIPE* - set/get lov_user_md
228 * *INFO - set/get lov_user_mds_data
230 /* lustre_ioctl.h 101-150 */
231 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
232 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
233 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
234 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
235 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
236 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
237 /* LL_IOC_RECREATE_OBJ 157 obsolete */
238 /* LL_IOC_RECREATE_FID 157 obsolete */
239 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
240 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
241 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
242 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
243 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
244 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
245 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
246 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
247 #define LL_IOC_RMTACL _IOW ('f', 167, long)
248 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
249 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
250 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
251 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
252 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
253 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
254 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
255 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
256 /* lustre_ioctl.h 177-210 */
257 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
258 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
259 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
260 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
261 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
262 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
263 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
264 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
265 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
266 struct lustre_swap_layouts)
267 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
268 struct hsm_current_action)
269 /* lustre_ioctl.h 221-232 */
270 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
271 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
272 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
273 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
274 #define LL_IOC_GET_LEASE _IO('f', 244)
275 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
276 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
277 #define LL_IOC_MIGRATE _IOR('f', 247, int)
278 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
279 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
281 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
283 LL_LEASE_RDLCK = 0x1,
284 LL_LEASE_WRLCK = 0x2,
285 LL_LEASE_UNLCK = 0x4,
288 #define LL_STATFS_LMV 1
289 #define LL_STATFS_LOV 2
290 #define LL_STATFS_NODELAY 4
292 #define IOC_MDC_TYPE 'i'
293 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
294 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
295 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
296 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
298 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
300 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
301 * files, but are unlikely to be used in practice and are not harmful if
302 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
303 * devices and are safe for use on new files. See LU-4209. */
304 /* To be compatible with old statically linked binary we keep the check for
305 * the older 0100000000 flag. This is already removed upstream. LU-812. */
306 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
307 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
308 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
309 O_LOV_DELAY_CREATE_MASK)
311 #define LL_FILE_IGNORE_LOCK 0x00000001
312 #define LL_FILE_GROUP_LOCKED 0x00000002
313 #define LL_FILE_READAHEA 0x00000004
314 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
315 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
316 #define LL_FILE_RMTACL 0x00000020
318 #define LOV_USER_MAGIC_V1 0x0BD10BD0
319 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
320 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
321 #define LOV_USER_MAGIC_V3 0x0BD30BD0
322 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
323 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
325 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
327 #define LOV_PATTERN_RAID0 0x001
328 #define LOV_PATTERN_RAID1 0x002
329 #define LOV_PATTERN_FIRST 0x100
330 #define LOV_PATTERN_CMOBD 0x200
332 #define LOV_PATTERN_F_MASK 0xffff0000
333 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
334 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
336 #define LOV_MAXPOOLNAME 15
337 #define LOV_POOLNAMEF "%.15s"
339 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
340 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
341 #define LOV_MAX_STRIPE_COUNT_OLD 160
342 /* This calculation is crafted so that input of 4096 will result in 160
343 * which in turn is equal to old maximal stripe count.
344 * XXX: In fact this is too simpified for now, what it also need is to get
345 * ea_type argument to clearly know how much space each stripe consumes.
347 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
348 * allocation that is sufficient for the current generation of systems.
350 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
351 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
352 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
353 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
355 #define XATTR_LUSTRE_PREFIX "lustre."
356 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
358 #define lov_user_ost_data lov_user_ost_data_v1
359 struct lov_user_ost_data_v1 { /* per-stripe data structure */
360 struct ost_id l_ost_oi; /* OST object ID */
361 __u32 l_ost_gen; /* generation of this OST index */
362 __u32 l_ost_idx; /* OST index in LOV */
363 } __attribute__((packed));
365 #define lov_user_md lov_user_md_v1
366 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
367 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
368 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
369 struct ost_id lmm_oi; /* LOV object ID */
370 __u32 lmm_stripe_size; /* size of stripe in bytes */
371 __u16 lmm_stripe_count; /* num stripes in use for this object */
373 __u16 lmm_stripe_offset; /* starting stripe offset in
374 * lmm_objects, use when writing */
375 __u16 lmm_layout_gen; /* layout generation number
376 * used when reading */
378 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
379 } __attribute__((packed, __may_alias__));
381 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
382 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
383 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
384 struct ost_id lmm_oi; /* LOV object ID */
385 __u32 lmm_stripe_size; /* size of stripe in bytes */
386 __u16 lmm_stripe_count; /* num stripes in use for this object */
388 __u16 lmm_stripe_offset; /* starting stripe offset in
389 * lmm_objects, use when writing */
390 __u16 lmm_layout_gen; /* layout generation number
391 * used when reading */
393 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
394 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
395 } __attribute__((packed));
397 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
399 if (lmm_magic == LOV_USER_MAGIC_V1)
400 return sizeof(struct lov_user_md_v1) +
401 stripes * sizeof(struct lov_user_ost_data_v1);
402 return sizeof(struct lov_user_md_v3) +
403 stripes * sizeof(struct lov_user_ost_data_v1);
406 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
407 * use this. It is unsafe to #define those values in this header as it
408 * is possible the application has already #included <sys/stat.h>. */
409 #ifdef HAVE_LOV_USER_MDS_DATA
410 #define lov_user_mds_data lov_user_mds_data_v1
411 struct lov_user_mds_data_v1 {
412 lstat_t lmd_st; /* MDS stat struct */
413 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
414 } __attribute__((packed));
416 struct lov_user_mds_data_v3 {
417 lstat_t lmd_st; /* MDS stat struct */
418 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
419 } __attribute__((packed));
422 struct lmv_user_mds_data {
423 struct lu_fid lum_fid;
429 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
430 LMV_HASH_TYPE_ALL_CHARS = 1,
431 LMV_HASH_TYPE_FNV_1A_64 = 2,
434 #define LMV_HASH_NAME_ALL_CHARS "all_char"
435 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
437 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
438 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
439 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
440 #define lmv_user_md lmv_user_md_v1
441 struct lmv_user_md_v1 {
442 __u32 lum_magic; /* must be the first field */
443 __u32 lum_stripe_count; /* dirstripe count */
444 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
445 __u32 lum_hash_type; /* Dir stripe policy */
446 __u32 lum_type; /* LMV type: default or normal */
450 char lum_pool_name[LOV_MAXPOOLNAME + 1];
451 struct lmv_user_mds_data lum_objects[0];
452 } __attribute__((packed));
454 static inline int lmv_user_md_size(int stripes, int lmm_magic)
456 return sizeof(struct lmv_user_md) +
457 stripes * sizeof(struct lmv_user_mds_data);
460 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
462 struct ll_recreate_obj {
468 __u64 id; /* holds object id */
469 __u32 generation; /* holds object generation */
470 __u32 f_type; /* holds object type or stripe idx when passing it to
471 * OST for saving into EA. */
479 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
480 const struct obd_uuid *u2)
482 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
485 static inline int obd_uuid_empty(struct obd_uuid *uuid)
487 return uuid->uuid[0] == '\0';
490 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
492 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
493 uuid->uuid[sizeof(*uuid) - 1] = '\0';
496 /* For printf's only, make sure uuid is terminated */
497 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
502 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
503 /* Obviously not safe, but for printfs, no real harm done...
504 we're always null-terminated, even in a race. */
505 static char temp[sizeof(*uuid)];
506 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
507 temp[sizeof(*uuid) - 1] = '\0';
510 return (char *)(uuid->uuid);
513 /* Extract fsname from uuid (or target name) of a target
514 e.g. (myfs-OST0007_UUID -> myfs)
515 see also deuuidify. */
516 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
520 strncpy(buf, uuid, buflen - 1);
521 buf[buflen - 1] = '\0';
522 p = strrchr(buf, '-');
527 /* printf display format
528 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
529 #define FID_NOBRACE_LEN 40
530 #define FID_LEN (FID_NOBRACE_LEN + 2)
531 #define DFID_NOBRACE LPX64":0x%x:0x%x"
532 #define DFID "["DFID_NOBRACE"]"
538 /* scanf input parse format -- strip '[' first.
539 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
540 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
541 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
542 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
544 #define SFID "0x"LPX64i":0x%x:0x%x"
551 /********* Quotas **********/
553 #define LUSTRE_QUOTABLOCK_BITS 10
554 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
556 static inline __u64 lustre_stoqb(size_t space)
558 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
561 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
562 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
563 #define Q_GETOINFO 0x800102 /* get obd quota info */
564 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
565 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
567 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
568 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
569 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
570 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
571 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
572 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
573 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
574 /* lustre-specific control commands */
575 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
576 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
578 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
580 struct if_quotacheck {
582 struct obd_uuid obd_uuid;
585 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
588 #define N_PERMS_MAX 64
590 struct perm_downcall_data {
596 struct identity_downcall_data {
603 struct perm_downcall_data idd_perms[N_PERMS_MAX];
607 /* for non-mapped uid/gid */
608 #define NOBODY_UID 99
609 #define NOBODY_GID 99
611 #define INVALID_ID (-1)
620 #ifdef NEED_QUOTA_DEFS
622 #define QIF_BLIMITS 1
624 #define QIF_ILIMITS 4
628 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
629 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
630 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
631 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
634 #endif /* !__KERNEL__ */
636 /* lustre volatile file support
637 * file name header: .^L^S^T^R:volatile"
639 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
640 #define LUSTRE_VOLATILE_HDR_LEN 14
642 typedef enum lustre_quota_version {
644 } lustre_quota_version_t;
646 /* XXX: same as if_dqinfo struct in kernel */
654 /* XXX: same as if_dqblk struct in kernel, plus one padding */
656 __u64 dqb_bhardlimit;
657 __u64 dqb_bsoftlimit;
659 __u64 dqb_ihardlimit;
660 __u64 dqb_isoftlimit;
682 struct obd_dqinfo qc_dqinfo;
683 struct obd_dqblk qc_dqblk;
685 struct obd_uuid obd_uuid;
688 /* swap layout flags */
689 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
690 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
691 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
692 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
694 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
695 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
696 struct lustre_swap_layouts {
705 /********* Changelogs **********/
706 /** Changelog record types */
707 enum changelog_rec_type {
709 CL_CREATE = 1, /* namespace */
710 CL_MKDIR = 2, /* namespace */
711 CL_HARDLINK = 3, /* namespace */
712 CL_SOFTLINK = 4, /* namespace */
713 CL_MKNOD = 5, /* namespace */
714 CL_UNLINK = 6, /* namespace */
715 CL_RMDIR = 7, /* namespace */
716 CL_RENAME = 8, /* namespace */
717 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
718 CL_OPEN = 10, /* not currently used */
719 CL_CLOSE = 11, /* may be written to log only with mtime change */
720 CL_LAYOUT = 12, /* file layout/striping modified */
724 CL_HSM = 16, /* HSM specific events, see flags */
725 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
731 static inline const char *changelog_type2str(int type) {
732 static const char *changelog_str[] = {
733 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
734 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
735 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
738 if (type >= 0 && type < CL_LAST)
739 return changelog_str[type];
743 /* per-record flags */
744 #define CLF_FLAGSHIFT 12
745 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
746 #define CLF_VERMASK (~CLF_FLAGMASK)
747 enum changelog_rec_flags {
748 CLF_VERSION = 0x1000,
751 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID
755 /* Anything under the flagmask may be per-type (if desired) */
756 /* Flags for unlink */
757 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
758 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
759 /* HSM cleaning needed */
760 /* Flags for rename */
761 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
763 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
764 * has an archive in backend */
767 /* 12b used (from high weight to low weight):
772 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
773 #define CLF_HSM_ERR_H 6
774 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
775 #define CLF_HSM_EVENT_H 9
776 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
777 #define CLF_HSM_FLAG_H 11
778 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
779 #define CLF_HSM_SPARE_H 15
780 #define CLF_HSM_LAST 15
782 /* Remove bits higher than _h, then extract the value
783 * between _h and _l by shifting lower weigth to bit 0. */
784 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
785 >> (CLF_HSM_LAST - _h + _l))
787 #define CLF_HSM_SUCCESS 0x00
788 #define CLF_HSM_MAXERROR 0x7E
789 #define CLF_HSM_ERROVERFLOW 0x7F
791 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
793 /* 3 bits field => 8 values allowed */
805 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
807 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
811 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
813 *flags |= (he << CLF_HSM_EVENT_L);
816 static inline __u16 hsm_get_cl_flags(int flags)
818 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
821 static inline void hsm_set_cl_flags(int *flags, int bits)
823 *flags |= (bits << CLF_HSM_FLAG_L);
826 static inline int hsm_get_cl_error(int flags)
828 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
831 static inline void hsm_set_cl_error(int *flags, int error)
833 *flags |= (error << CLF_HSM_ERR_L);
836 enum changelog_send_flag {
837 /* Not yet implemented */
838 CHANGELOG_FLAG_FOLLOW = 0x01,
839 /* Blocking IO makes sense in case of slow user parsing of the records,
840 * but it also prevents us from cleaning up if the records are not
842 CHANGELOG_FLAG_BLOCK = 0x02,
843 /* Pack jobid into the changelog records if available. */
844 CHANGELOG_FLAG_JOBID = 0x04,
847 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
848 changelog_rec_offset(CLF_SUPPORTED))
850 /* 31 usable bytes string + null terminator. */
851 #define LUSTRE_JOBID_SIZE 32
853 /* This is the minimal changelog record. It can contain extensions
854 * such as rename fields or process jobid. Its exact content is described
857 * Extensions are packed in the same order as their corresponding flags.
859 struct changelog_rec {
861 __u16 cr_flags; /**< \a changelog_rec_flags */
862 __u32 cr_type; /**< \a changelog_rec_type */
863 __u64 cr_index; /**< changelog record number */
864 __u64 cr_prev; /**< last index for this target fid */
867 lustre_fid cr_tfid; /**< target fid */
868 __u32 cr_markerflags; /**< CL_MARK flags */
870 lustre_fid cr_pfid; /**< parent fid */
873 /* Changelog extension for RENAME. */
874 struct changelog_ext_rename {
875 lustre_fid cr_sfid; /**< source fid, or zero */
876 lustre_fid cr_spfid; /**< source parent fid, or zero */
879 /* Changelog extension to include JOBID. */
880 struct changelog_ext_jobid {
881 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
885 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
887 size_t size = sizeof(struct changelog_rec);
889 if (crf & CLF_RENAME)
890 size += sizeof(struct changelog_ext_rename);
893 size += sizeof(struct changelog_ext_jobid);
898 static inline size_t changelog_rec_size(struct changelog_rec *rec)
900 return changelog_rec_offset(rec->cr_flags);
903 static inline size_t changelog_rec_varsize(struct changelog_rec *rec)
905 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
909 struct changelog_ext_rename *changelog_rec_rename(struct changelog_rec *rec)
911 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
913 return (struct changelog_ext_rename *)((char *)rec +
914 changelog_rec_offset(crf));
917 /* The jobid follows the rename extension, if present */
919 struct changelog_ext_jobid *changelog_rec_jobid(struct changelog_rec *rec)
921 enum changelog_rec_flags crf = rec->cr_flags &
922 (CLF_VERSION | CLF_RENAME);
924 return (struct changelog_ext_jobid *)((char *)rec +
925 changelog_rec_offset(crf));
928 /* The name follows the rename and jobid extensions, if present */
929 static inline char *changelog_rec_name(struct changelog_rec *rec)
931 return (char *)rec + changelog_rec_offset(rec->cr_flags &
935 static inline size_t changelog_rec_snamelen(struct changelog_rec *rec)
937 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
940 static inline char *changelog_rec_sname(struct changelog_rec *rec)
942 char *cr_name = changelog_rec_name(rec);
944 return cr_name + strlen(cr_name) + 1;
948 * Remap a record to the desired format as specified by the crf flags.
949 * The record must be big enough to contain the final remapped version.
950 * Superfluous extension fields are removed and missing ones are added
951 * and zeroed. The flags of the record are updated accordingly.
953 * The jobid and rename extensions can be added to a record, to match the
954 * format an application expects, typically. In this case, the newly added
955 * fields will be zeroed.
956 * The Jobid field can be removed, to guarantee compatibility with older
957 * clients that don't expect this field in the records they process.
959 * The following assumptions are being made:
960 * - CLF_RENAME will not be removed
961 * - CLF_JOBID will not be added without CLF_RENAME being added too
963 * @param[in,out] rec The record to remap.
964 * @param[in] crf_wanted Flags describing the desired extensions.
966 static inline void changelog_remap_rec(struct changelog_rec *rec,
967 enum changelog_rec_flags crf_wanted)
972 crf_wanted &= CLF_SUPPORTED;
974 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
977 /* First move the variable-length name field */
978 memmove((char *)rec + changelog_rec_offset(crf_wanted),
979 changelog_rec_name(rec), rec->cr_namelen);
981 /* Locations of jobid and rename extensions in the remapped record */
982 jid_mov = (char *)rec +
983 changelog_rec_offset(crf_wanted & ~CLF_JOBID);
984 rnm_mov = (char *)rec +
985 changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
987 /* Move the extension fields to the desired positions */
988 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
989 memmove(jid_mov, changelog_rec_jobid(rec),
990 sizeof(struct changelog_ext_jobid));
992 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
993 memmove(rnm_mov, changelog_rec_rename(rec),
994 sizeof(struct changelog_ext_rename));
996 /* Clear newly added fields */
997 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
998 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1000 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1001 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1003 /* Update the record's flags accordingly */
1004 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1007 struct ioc_changelog {
1014 enum changelog_message_type {
1015 CL_RECORD = 10, /* message is a changelog_rec */
1016 CL_EOF = 11, /* at end of current changelog */
1019 /********* Misc **********/
1021 struct ioc_data_version {
1023 __u64 idv_flags; /* See LL_DV_xxx */
1025 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
1026 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
1029 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1032 #define dot_lustre_name ".lustre"
1035 /********* HSM **********/
1037 /** HSM per-file state
1038 * See HSM_FLAGS below.
1041 HS_EXISTS = 0x00000001,
1042 HS_DIRTY = 0x00000002,
1043 HS_RELEASED = 0x00000004,
1044 HS_ARCHIVED = 0x00000008,
1045 HS_NORELEASE = 0x00000010,
1046 HS_NOARCHIVE = 0x00000020,
1047 HS_LOST = 0x00000040,
1050 /* HSM user-setable flags. */
1051 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1053 /* Other HSM flags. */
1054 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1057 * All HSM-related possible flags that could be applied to a file.
1058 * This should be kept in sync with hsm_states.
1060 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1063 * HSMÂ request progress state
1065 enum hsm_progress_states {
1072 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1075 case HPS_WAITING: return "waiting";
1076 case HPS_RUNNING: return "running";
1077 case HPS_DONE: return "done";
1078 default: return "unknown";
1085 } __attribute__((packed));
1088 * Current HSM states of a Lustre file.
1090 * This structure purpose is to be sent to user-space mainly. It describes the
1091 * current HSM flags and in-progress action.
1093 struct hsm_user_state {
1094 /** Current HSM states, from enum hsm_states. */
1096 __u32 hus_archive_id;
1097 /** The current undergoing action, if there is one */
1098 __u32 hus_in_progress_state;
1099 __u32 hus_in_progress_action;
1100 struct hsm_extent hus_in_progress_location;
1101 char hus_extended_info[];
1104 struct hsm_state_set_ioc {
1105 struct lu_fid hssi_fid;
1107 __u64 hssi_clearmask;
1111 * This structure describes the current in-progress action for a file.
1112 * it is retuned to user space and send over the wire
1114 struct hsm_current_action {
1115 /** The current undergoing action, if there is one */
1116 /* state is one of hsm_progress_states */
1118 /* action is one of hsm_user_action */
1120 struct hsm_extent hca_location;
1123 /***** HSM user requests ******/
1124 /* User-generated (lfs/ioctl) request types */
1125 enum hsm_user_action {
1126 HUA_NONE = 1, /* no action (noop) */
1127 HUA_ARCHIVE = 10, /* copy to hsm */
1128 HUA_RESTORE = 11, /* prestage */
1129 HUA_RELEASE = 12, /* drop ost objects */
1130 HUA_REMOVE = 13, /* remove from archive */
1131 HUA_CANCEL = 14 /* cancel a request */
1134 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1137 case HUA_NONE: return "NOOP";
1138 case HUA_ARCHIVE: return "ARCHIVE";
1139 case HUA_RESTORE: return "RESTORE";
1140 case HUA_RELEASE: return "RELEASE";
1141 case HUA_REMOVE: return "REMOVE";
1142 case HUA_CANCEL: return "CANCEL";
1143 default: return "UNKNOWN";
1148 * List of hr_flags (bit field)
1150 #define HSM_FORCE_ACTION 0x0001
1151 /* used by CT, connot be set by user */
1152 #define HSM_GHOST_COPY 0x0002
1155 * Contains all the fixed part of struct hsm_user_request.
1158 struct hsm_request {
1159 __u32 hr_action; /* enum hsm_user_action */
1160 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1161 __u64 hr_flags; /* request flags */
1162 __u32 hr_itemcount; /* item count in hur_user_item vector */
1166 struct hsm_user_item {
1168 struct hsm_extent hui_extent;
1169 } __attribute__((packed));
1171 struct hsm_user_request {
1172 struct hsm_request hur_request;
1173 struct hsm_user_item hur_user_item[0];
1174 /* extra data blob at end of struct (after all
1175 * hur_user_items), only use helpers to access it
1177 } __attribute__((packed));
1179 /** Return pointer to data field in a hsm user request */
1180 static inline void *hur_data(struct hsm_user_request *hur)
1182 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1186 * Compute the current length of the provided hsm_user_request. This returns -1
1187 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1189 * return -1 on bounds check error.
1191 static inline ssize_t hur_len(struct hsm_user_request *hur)
1195 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1196 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1197 (__u64)hur->hur_request.hr_itemcount *
1198 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1200 if (size > ~((size_t)0)>>1)
1206 /****** HSM RPCs to copytool *****/
1207 /* Message types the copytool may receive */
1208 enum hsm_message_type {
1209 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1212 /* Actions the copytool may be instructed to take for a given action_item */
1213 enum hsm_copytool_action {
1214 HSMA_NONE = 10, /* no action */
1215 HSMA_ARCHIVE = 20, /* arbitrary offset */
1221 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1224 case HSMA_NONE: return "NOOP";
1225 case HSMA_ARCHIVE: return "ARCHIVE";
1226 case HSMA_RESTORE: return "RESTORE";
1227 case HSMA_REMOVE: return "REMOVE";
1228 case HSMA_CANCEL: return "CANCEL";
1229 default: return "UNKNOWN";
1233 /* Copytool item action description */
1234 struct hsm_action_item {
1235 __u32 hai_len; /* valid size of this struct */
1236 __u32 hai_action; /* hsm_copytool_action, but use known size */
1237 lustre_fid hai_fid; /* Lustre FID to operated on */
1238 lustre_fid hai_dfid; /* fid used for data access */
1239 struct hsm_extent hai_extent; /* byte range to operate on */
1240 __u64 hai_cookie; /* action cookie from coordinator */
1241 __u64 hai_gid; /* grouplock id */
1242 char hai_data[0]; /* variable length */
1243 } __attribute__((packed));
1246 * helper function which print in hexa the first bytes of
1248 * \param hai [IN] record to print
1249 * \param buffer [OUT] output buffer
1250 * \param len [IN] max buffer len
1253 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1254 char *buffer, int len)
1256 int i, sz, data_len;
1261 data_len = hai->hai_len - sizeof(*hai);
1262 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1266 cnt = snprintf(ptr, sz, "%.2X",
1267 (unsigned char)hai->hai_data[i]);
1275 /* Copytool action list */
1276 #define HAL_VERSION 1
1277 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1278 struct hsm_action_list {
1280 __u32 hal_count; /* number of hai's to follow */
1281 __u64 hal_compound_id; /* returned by coordinator */
1283 __u32 hal_archive_id; /* which archive backend */
1285 char hal_fsname[0]; /* null-terminated */
1286 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1287 boundaries. See hai_zero */
1288 } __attribute__((packed));
1290 #ifndef HAVE_CFS_SIZE_ROUND
1291 static inline int cfs_size_round (int val)
1293 return (val + 7) & (~0x7);
1295 #define HAVE_CFS_SIZE_ROUND
1298 /* Return pointer to first hai in action list */
1299 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1301 return (struct hsm_action_item *)(hal->hal_fsname +
1302 cfs_size_round(strlen(hal-> \
1306 /* Return pointer to next hai */
1307 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1309 return (struct hsm_action_item *)((char *)hai +
1310 cfs_size_round(hai->hai_len));
1313 /* Return size of an hsm_action_list */
1314 static inline size_t hal_size(struct hsm_action_list *hal)
1318 struct hsm_action_item *hai;
1320 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1321 hai = hai_first(hal);
1322 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1323 sz += cfs_size_round(hai->hai_len);
1329 * describe the attributes to be set on imported file
1331 struct hsm_user_import {
1340 __u32 hui_archive_id;
1343 /* Copytool progress reporting */
1344 #define HP_FLAG_COMPLETED 0x01
1345 #define HP_FLAG_RETRY 0x02
1347 struct hsm_progress {
1350 struct hsm_extent hp_extent;
1352 __u16 hp_errval; /* positive val */
1357 __u64 hc_data_version;
1359 __u16 hc_errval; /* positive val */
1361 struct hsm_action_item hc_hai;
1365 enum llapi_json_types {
1366 LLAPI_JSON_INTEGER = 1,
1372 struct llapi_json_item {
1381 struct llapi_json_item *lji_next;
1384 struct llapi_json_item_list {
1385 int ljil_item_count;
1386 struct llapi_json_item *ljil_items;
1389 /** @} lustreuser */
1391 #endif /* _LUSTRE_USER_H */