1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
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 #include <lustre/ll_fiemap.h>
45 #if defined(__linux__)
46 #include <linux/lustre_user.h>
47 #elif defined(__APPLE__)
48 #include <darwin/lustre_user.h>
49 #elif defined(__WINNT__)
50 #include <winnt/lustre_user.h>
52 #error Unsupported operating system.
56 #define LL_SUPER_MAGIC 0x0BD00BD0
58 #ifndef FSFILT_IOC_GETFLAGS
59 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
60 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
61 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
62 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
63 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
64 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
65 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
68 /* FIEMAP flags supported by Lustre */
69 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
71 enum obd_statfs_state {
72 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
73 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
74 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
75 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
76 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
90 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
104 * The ioctl naming rules:
105 * LL_* - works on the currently opened filehandle instead of parent dir
106 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
107 * *_MDC_* - gets/sets data related to MDC
108 * *_LOV_* - gets/sets data related to OSC/LOV
109 * *FILE* - called on parent dir and passes in a filename
110 * *STRIPE* - set/get lov_user_md
111 * *INFO - set/get lov_user_mds_data
113 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
114 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
115 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
116 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
117 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
118 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
119 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
120 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
121 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
122 #define LL_IOC_QUOTACHECK _IOW ('f', 160, int)
123 #define LL_IOC_POLL_QUOTACHECK _IOR ('f', 161, struct if_quotacheck *)
124 #define LL_IOC_QUOTACTL _IOWR('f', 162, struct if_quotactl *)
125 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
126 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
127 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
128 #define LL_IOC_RMTACL _IOW ('f', 167, long)
130 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
132 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
133 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
134 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, long)
135 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
136 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
138 #define LL_STATFS_MDC 1
139 #define LL_STATFS_LOV 2
141 #define IOC_MDC_TYPE 'i'
142 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
143 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
144 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
145 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
147 /* Keep these for backward compartability. */
148 #define LL_IOC_OBD_STATFS IOC_OBD_STATFS
149 #define IOC_MDC_GETSTRIPE IOC_MDC_GETFILESTRIPE
151 #define O_LOV_DELAY_CREATE 0100000000 /* hopefully this does not conflict */
153 #define LL_FILE_IGNORE_LOCK 0x00000001
154 #define LL_FILE_GROUP_LOCKED 0x00000002
155 #define LL_FILE_READAHEAD 0x00000004
156 #define LL_FILE_RMTACL 0x00000008
158 #define LOV_USER_MAGIC_V1 0x0BD10BD0
159 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
160 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
161 #define LOV_USER_MAGIC_V3 0x0BD30BD0
163 #define LOV_PATTERN_RAID0 0x001
164 #define LOV_PATTERN_RAID1 0x002
165 #define LOV_PATTERN_FIRST 0x100
167 #define LOV_MAXPOOLNAME 16
168 #define LOV_POOLNAMEF "%.16s"
170 #define lov_user_ost_data lov_user_ost_data_v1
171 struct lov_user_ost_data_v1 { /* per-stripe data structure */
172 __u64 l_object_id; /* OST object ID */
173 __u64 l_object_gr; /* OST object group (creating MDS number) */
174 __u32 l_ost_gen; /* generation of this OST index */
175 __u32 l_ost_idx; /* OST index in LOV */
176 } __attribute__((packed));
178 #define lov_user_md lov_user_md_v1
179 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
180 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
181 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
182 __u64 lmm_object_id; /* LOV object ID */
183 __u64 lmm_object_gr; /* LOV object group */
184 __u32 lmm_stripe_size; /* size of stripe in bytes */
185 __u16 lmm_stripe_count; /* num stripes in use for this object */
186 __u16 lmm_stripe_offset; /* starting stripe offset in lmm_objects */
187 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
188 } __attribute__((packed));
190 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
191 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
192 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
193 __u64 lmm_object_id; /* LOV object ID */
194 __u64 lmm_object_gr; /* LOV object group */
195 __u32 lmm_stripe_size; /* size of stripe in bytes */
196 __u16 lmm_stripe_count; /* num stripes in use for this object */
197 __u16 lmm_stripe_offset; /* starting stripe offset in lmm_objects */
198 char lmm_pool_name[LOV_MAXPOOLNAME]; /* pool name */
199 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
200 } __attribute__((packed));
202 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
203 * use this. It is unsafe to #define those values in this header as it
204 * is possible the application has already #included <sys/stat.h>. */
205 #ifdef HAVE_LOV_USER_MDS_DATA
206 #define lov_user_mds_data lov_user_mds_data_v1
207 struct lov_user_mds_data_v1 {
208 lstat_t lmd_st; /* MDS stat struct */
209 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
210 } __attribute__((packed));
212 struct lov_user_mds_data_v3 {
213 lstat_t lmd_st; /* MDS stat struct */
214 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
215 } __attribute__((packed));
218 struct ll_recreate_obj {
225 __u64 id; /* holds object id */
226 __u32 generation; /* holds object generation */
227 __u32 f_type; /* holds object type or stripe idx when passing it to
228 * OST for saving into EA. */
232 struct ll_fid ff_fid; /* ff_fid.f_type == file stripe number */
241 static inline int obd_uuid_equals(struct obd_uuid *u1, struct obd_uuid *u2)
243 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
246 static inline int obd_uuid_empty(struct obd_uuid *uuid)
248 return uuid->uuid[0] == '\0';
251 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
253 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
254 uuid->uuid[sizeof(*uuid) - 1] = '\0';
257 /* For printf's only, make sure uuid is terminated */
258 static inline char *obd_uuid2str(struct obd_uuid *uuid)
260 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
261 /* Obviously not safe, but for printfs, no real harm done...
262 we're always null-terminated, even in a race. */
263 static char temp[sizeof(*uuid)];
264 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
265 temp[sizeof(*uuid) - 1] = '\0';
268 return (char *)(uuid->uuid);
271 /* Extract fsname from uuid (or target name) of a target
272 e.g. (myfs-OST0007_UUID -> myfs)
273 see also deuuidify. */
274 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
278 strncpy(buf, uuid, buflen - 1);
279 buf[buflen - 1] = '\0';
280 p = strrchr(buf, '-');
288 * FID is a cluster-wide unique identifier of a file or an object (stripe).
289 * FIDs are never reused.
293 * FID sequence. Sequence is a unit of migration: all files (objects)
294 * with FIDs from a given sequence are stored on the same server.
295 * Lustre should support 2^64 objects, so even if each sequence
296 * has only a single object we can still enumerate 2^64 objects.
299 /** FID number within sequence. */
302 * FID version, used to distinguish different versions (in the sense
303 * of snapshots, etc.) of the same file system object. Not currently
309 /* Userspace should treat lu_fid as opaque, and only use the following methods
310 to print or parse them. Other functions (e.g. compare, swab) could be moved
311 here from lustre_idl.h if needed. */
312 typedef struct lu_fid lustre_fid;
314 /* printf display format
315 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
316 #define DFID "["LPX64":0x%x:0x%x]"
322 /* scanf input parse format -- strip '[' first.
323 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
324 #define SFID "0x%llx:0x%x:0x%x"
331 /********* Quotas **********/
333 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
334 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
335 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
336 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
337 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
338 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
339 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
340 /* lustre-specific control commands */
341 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
342 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
344 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
345 #define IMMQUOTA 0x4 /* set immutable quota flag, cannot be turned on/off
346 * on-fly. temporary used by SOM */
348 struct if_quotacheck {
350 struct obd_uuid obd_uuid;
353 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd620
356 #define N_PERMS_MAX 64
358 struct perm_downcall_data {
363 struct identity_downcall_data {
369 struct perm_downcall_data idd_perms[N_PERMS_MAX];
374 /* for non-mapped uid/gid */
375 #define NOBODY_UID 99
376 #define NOBODY_GID 99
378 #define INVALID_ID (-1)
387 #ifdef NEED_QUOTA_DEFS
389 #define QIF_BLIMITS 1
391 #define QIF_ILIMITS 4
395 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
396 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
397 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
398 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
401 #endif /* !__KERNEL__ */
403 typedef enum lustre_quota_version {
405 } lustre_quota_version_t;
407 /* XXX: same as if_dqinfo struct in kernel */
415 /* XXX: same as if_dqblk struct in kernel, plus one padding */
417 __u64 dqb_bhardlimit;
418 __u64 dqb_bsoftlimit;
420 __u64 dqb_ihardlimit;
421 __u64 dqb_isoftlimit;
443 struct obd_dqinfo qc_dqinfo;
444 struct obd_dqblk qc_dqblk;
446 struct obd_uuid obd_uuid;
450 /********* Changelogs **********/
451 /** Changelog record types */
452 enum changelog_rec_type {
454 CL_CREATE = 1, /* namespace */
455 CL_MKDIR = 2, /* namespace */
456 CL_HARDLINK = 3, /* namespace */
457 CL_SOFTLINK = 4, /* namespace */
458 CL_MKNOD = 5, /* namespace */
459 CL_UNLINK = 6, /* namespace */
460 CL_RMDIR = 7, /* namespace */
461 CL_RENAME = 8, /* namespace */
462 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
463 CL_OPEN = 10, /* not currently used */
464 CL_CLOSE = 11, /* may be written to log only with mtime change */
469 CL_HSM = 16, /* HSM specific events, see flags */
473 static inline const char *changelog_type2str(int type) {
474 static const char *changelog_str[] = {
475 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
476 "RMDIR", "RNMFM", "RNMTO", "OPEN", "CLOSE", "IOCTL", "TRUNC",
477 "SATTR", "XATTR", "HSM" };
478 if (type >= 0 && type < CL_LAST)
479 return changelog_str[type];
483 /* per-record flags */
484 #define CLF_VERSION 0x1000
485 #define CLF_FLAGMASK 0x0FFF
486 /* Anything under the flagmask may be per-type (if desired) */
488 #define CR_MAXSIZE (PATH_MAX + sizeof(struct changelog_rec))
489 struct changelog_rec {
491 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
492 __u32 cr_type; /**< \a changelog_rec_type */
493 __u64 cr_index; /**< changelog record number */
494 __u64 cr_prev; /**< last index for this target fid */
497 lustre_fid cr_tfid; /**< target fid */
498 __u32 cr_markerflags; /**< CL_MARK flags */
500 lustre_fid cr_pfid; /**< parent fid */
501 char cr_name[0]; /**< last element */
502 } __attribute__((packed));
504 struct ioc_changelog_clear {
510 enum changelog_message_type {
511 CL_RECORD = 10, /* message is a changelog_rec */
512 CL_EOF = 11, /* at end of current changelog */
515 /********* Misc **********/
518 # define offsetof(typ,memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
521 #define dot_lustre_name ".lustre"
524 /********* HSM **********/
527 #define HSM_FLAGS_MASK 0
530 enum hsm_message_type {
531 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
534 /* User-generated (ioctl) request types */
536 HSMR_ARCHIVE = 10, /* copy to hsm */
537 HSMR_RESTORE = 11, /* prestage */
538 HSMR_RELEASE = 12, /* drop ost objects */
539 HSMR_REMOVE = 13, /* remove from archive */
543 /* Copytool commands */
545 HSMA_ARCHIVE = 20, /* arbitrary offset */
551 /* Copytool item action description */
552 struct hsm_action_item {
553 __u32 hai_len; /* valid size of this struct */
554 __u32 hai_action; /* enum actually, but use known size */
555 lustre_fid hai_fid; /* Lustre FID to operated on */
556 __u64 hai_cookie; /* action cookie from coordinator */
557 __u64 hai_extent_start; /* byte range to operate on */
558 __u64 hai_extent_end;
559 __u64 hai_gid; /* grouplock id */
560 char hai_data[0]; /* variable length */
561 } __attribute__((packed));
563 /* Copytool action list */
564 #define HAL_VERSION 1
565 #define HAL_MAXSIZE 4096 /* bytes, used in userspace only */
566 struct hsm_action_list {
568 __u32 hal_count; /* number of hai's to follow */
569 __u32 hal_archive_num; /* which archive backend */
571 char hal_fsname[0]; /* null-terminated */
572 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
573 boundaries. See hai_zero */
574 } __attribute__((packed));
576 /* Return pointer to first hai in action list */
577 static __inline__ struct hsm_action_item * hai_zero(struct hsm_action_list *hal)
579 return (struct hsm_action_item *)(hal->hal_fsname +
580 size_round(strlen(hal->hal_fsname)));
582 /* Return pointer to next hai */
583 static __inline__ struct hsm_action_item * hai_next(struct hsm_action_item *hai)
585 return (struct hsm_action_item *)((char *)hai +
586 size_round(hai->hai_len));
590 #endif /* _LUSTRE_USER_H */