4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2010, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_user.h
38 * Lustre public user-space interface definitions.
41 #ifndef _LUSTRE_USER_H
42 #define _LUSTRE_USER_H
44 /** \defgroup lustreuser lustreuser
51 #include <libcfs/posix/posix-types.h>
53 #include <lustre/ll_fiemap.h>
54 #if defined(__linux__)
55 #include <linux/lustre_user.h>
56 #elif defined(__APPLE__)
57 #include <darwin/lustre_user.h>
58 #elif defined(__WINNT__)
59 #include <winnt/lustre_user.h>
61 #error Unsupported operating system.
64 #define LUSTRE_EOF 0xffffffffffffffffULL
67 #define LL_SUPER_MAGIC 0x0BD00BD0
69 #ifndef FSFILT_IOC_GETFLAGS
70 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
71 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
72 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
73 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
74 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
75 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
76 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
79 /* FIEMAP flags supported by Lustre */
80 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
82 enum obd_statfs_state {
83 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
84 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
85 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
86 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
87 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
101 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
102 __u32 os_fprecreated; /* objs available now to the caller */
103 /* used in QoS code to find preferred
118 * FID is a cluster-wide unique identifier of a file or an object (stripe).
119 * FIDs are never reused.
123 * FID sequence. Sequence is a unit of migration: all files (objects)
124 * with FIDs from a given sequence are stored on the same server.
125 * Lustre should support 2^64 objects, so even if each sequence
126 * has only a single object we can still enumerate 2^64 objects.
129 /* FID number within sequence. */
132 * FID version, used to distinguish different versions (in the sense
133 * of snapshots, etc.) of the same file system object. Not currently
139 static inline bool fid_is_zero(const struct lu_fid *fid)
141 return fid->f_seq == 0 && fid->f_oid == 0;
144 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
145 * MDT-object's FID::f_ver, instead it is the OST-object index in its
146 * parent MDT-object's layout EA. */
147 #define f_stripe_idx f_ver
150 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
153 /* keep this one for compatibility */
154 struct filter_fid_old {
155 struct lu_fid ff_parent;
160 /* Userspace should treat lu_fid as opaque, and only use the following methods
161 * to print or parse them. Other functions (e.g. compare, swab) could be moved
162 * here from lustre_idl.h if needed. */
163 typedef struct lu_fid lustre_fid;
166 * Following struct for object attributes, that will be kept inode's EA.
167 * Introduced in 2.0 release (please see b15993, for details)
168 * Added to all objects since Lustre 2.4 as contains self FID
170 struct lustre_mdt_attrs {
172 * Bitfield for supported data in this structure. From enum lma_compat.
173 * lma_self_fid and lma_flags are always available.
177 * Per-file incompat feature list. Lustre version should support all
178 * flags set in this field. The supported feature mask is available in
182 /** FID of this inode */
183 struct lu_fid lma_self_fid;
187 * Prior to 2.4, the LMA structure also included SOM attributes which has since
188 * been moved to a dedicated xattr
189 * lma_flags was also removed because of lma_compat/incompat fields.
191 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
194 * OST object IDentifier.
202 struct lu_fid oi_fid;
206 #define DOSTID LPX64":"LPU64
207 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
210 * The ioctl naming rules:
211 * LL_* - works on the currently opened filehandle instead of parent dir
212 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
213 * *_MDC_* - gets/sets data related to MDC
214 * *_LOV_* - gets/sets data related to OSC/LOV
215 * *FILE* - called on parent dir and passes in a filename
216 * *STRIPE* - set/get lov_user_md
217 * *INFO - set/get lov_user_mds_data
219 /* lustre_ioctl.h 101-150 */
220 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
221 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
222 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
223 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
224 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
225 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
226 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
227 #define LL_IOC_RECREATE_FID _IOW ('f', 157, struct lu_fid)
228 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
229 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
230 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
231 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
232 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
233 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
234 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
235 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
236 #define LL_IOC_RMTACL _IOW ('f', 167, long)
237 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
238 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
239 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
240 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
241 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
242 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
243 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
244 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
245 /* lustre_ioctl.h 177-210 */
246 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
247 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
248 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
249 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
250 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
251 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
252 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
253 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
254 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
255 struct lustre_swap_layouts)
256 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
257 struct hsm_current_action)
258 /* lustre_ioctl.h 221-232 */
259 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
260 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
261 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
262 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
263 #define LL_IOC_GET_LEASE _IO('f', 244)
264 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
265 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
266 #define LL_IOC_MIGRATE _IOR('f', 247, int)
267 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
269 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
271 LL_LEASE_RDLCK = 0x1,
272 LL_LEASE_WRLCK = 0x2,
273 LL_LEASE_UNLCK = 0x4,
276 #define LL_STATFS_LMV 1
277 #define LL_STATFS_LOV 2
278 #define LL_STATFS_NODELAY 4
280 #define IOC_MDC_TYPE 'i'
281 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
282 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
283 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
284 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
286 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
288 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
289 * files, but are unlikely to be used in practice and are not harmful if
290 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
291 * devices and are safe for use on new files. See LU-4209. */
292 /* To be compatible with old statically linked binary we keep the check for
293 * the older 0100000000 flag. This is already removed upstream. LU-812. */
294 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
295 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
296 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
297 O_LOV_DELAY_CREATE_MASK)
299 #define LL_FILE_IGNORE_LOCK 0x00000001
300 #define LL_FILE_GROUP_LOCKED 0x00000002
301 #define LL_FILE_READAHEA 0x00000004
302 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
303 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
304 #define LL_FILE_RMTACL 0x00000020
306 #define LOV_USER_MAGIC_V1 0x0BD10BD0
307 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
308 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
309 #define LOV_USER_MAGIC_V3 0x0BD30BD0
311 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
313 #define LOV_PATTERN_RAID0 0x001
314 #define LOV_PATTERN_RAID1 0x002
315 #define LOV_PATTERN_FIRST 0x100
316 #define LOV_PATTERN_CMOBD 0x200
318 #define LOV_PATTERN_F_MASK 0xffff0000
319 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
320 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
322 #define LOV_MAXPOOLNAME 16
323 #define LOV_POOLNAMEF "%.16s"
325 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
326 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
327 #define LOV_MAX_STRIPE_COUNT_OLD 160
328 /* This calculation is crafted so that input of 4096 will result in 160
329 * which in turn is equal to old maximal stripe count.
330 * XXX: In fact this is too simpified for now, what it also need is to get
331 * ea_type argument to clearly know how much space each stripe consumes.
333 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
334 * allocation that is sufficient for the current generation of systems.
336 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
337 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
338 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
339 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
341 #define XATTR_LUSTRE_PREFIX "lustre."
342 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
344 #define lov_user_ost_data lov_user_ost_data_v1
345 struct lov_user_ost_data_v1 { /* per-stripe data structure */
346 struct ost_id l_ost_oi; /* OST object ID */
347 __u32 l_ost_gen; /* generation of this OST index */
348 __u32 l_ost_idx; /* OST index in LOV */
349 } __attribute__((packed));
351 #define lov_user_md lov_user_md_v1
352 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
353 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
354 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
355 struct ost_id lmm_oi; /* LOV object ID */
356 __u32 lmm_stripe_size; /* size of stripe in bytes */
357 __u16 lmm_stripe_count; /* num stripes in use for this object */
359 __u16 lmm_stripe_offset; /* starting stripe offset in
360 * lmm_objects, use when writing */
361 __u16 lmm_layout_gen; /* layout generation number
362 * used when reading */
364 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
365 } __attribute__((packed, __may_alias__));
367 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
368 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
369 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
370 struct ost_id lmm_oi; /* LOV object ID */
371 __u32 lmm_stripe_size; /* size of stripe in bytes */
372 __u16 lmm_stripe_count; /* num stripes in use for this object */
374 __u16 lmm_stripe_offset; /* starting stripe offset in
375 * lmm_objects, use when writing */
376 __u16 lmm_layout_gen; /* layout generation number
377 * used when reading */
379 char lmm_pool_name[LOV_MAXPOOLNAME]; /* pool name */
380 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
381 } __attribute__((packed));
383 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
385 if (lmm_magic == LOV_USER_MAGIC_V3)
386 return sizeof(struct lov_user_md_v3) +
387 stripes * sizeof(struct lov_user_ost_data_v1);
389 return sizeof(struct lov_user_md_v1) +
390 stripes * sizeof(struct lov_user_ost_data_v1);
393 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
394 * use this. It is unsafe to #define those values in this header as it
395 * is possible the application has already #included <sys/stat.h>. */
396 #ifdef HAVE_LOV_USER_MDS_DATA
397 #define lov_user_mds_data lov_user_mds_data_v1
398 struct lov_user_mds_data_v1 {
399 lstat_t lmd_st; /* MDS stat struct */
400 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
401 } __attribute__((packed));
403 struct lov_user_mds_data_v3 {
404 lstat_t lmd_st; /* MDS stat struct */
405 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
406 } __attribute__((packed));
409 struct lmv_user_mds_data {
410 struct lu_fid lum_fid;
416 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
417 LMV_HASH_TYPE_ALL_CHARS = 1,
418 LMV_HASH_TYPE_FNV_1A_64 = 2,
421 #define LMV_HASH_NAME_ALL_CHARS "all_char"
422 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
424 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
425 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
426 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
427 #define lmv_user_md lmv_user_md_v1
428 struct lmv_user_md_v1 {
429 __u32 lum_magic; /* must be the first field */
430 __u32 lum_stripe_count; /* dirstripe count */
431 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
432 __u32 lum_hash_type; /* Dir stripe policy */
433 __u32 lum_type; /* LMV type: default or normal */
437 char lum_pool_name[LOV_MAXPOOLNAME];
438 struct lmv_user_mds_data lum_objects[0];
439 } __attribute__((packed));
441 static inline int lmv_user_md_size(int stripes, int lmm_magic)
443 return sizeof(struct lmv_user_md) +
444 stripes * sizeof(struct lmv_user_mds_data);
447 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
449 struct ll_recreate_obj {
455 __u64 id; /* holds object id */
456 __u32 generation; /* holds object generation */
457 __u32 f_type; /* holds object type or stripe idx when passing it to
458 * OST for saving into EA. */
466 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
467 const struct obd_uuid *u2)
469 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
472 static inline int obd_uuid_empty(struct obd_uuid *uuid)
474 return uuid->uuid[0] == '\0';
477 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
479 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
480 uuid->uuid[sizeof(*uuid) - 1] = '\0';
483 /* For printf's only, make sure uuid is terminated */
484 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
489 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
490 /* Obviously not safe, but for printfs, no real harm done...
491 we're always null-terminated, even in a race. */
492 static char temp[sizeof(*uuid)];
493 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
494 temp[sizeof(*uuid) - 1] = '\0';
497 return (char *)(uuid->uuid);
500 /* Extract fsname from uuid (or target name) of a target
501 e.g. (myfs-OST0007_UUID -> myfs)
502 see also deuuidify. */
503 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
507 strncpy(buf, uuid, buflen - 1);
508 buf[buflen - 1] = '\0';
509 p = strrchr(buf, '-');
514 /* printf display format
515 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
516 #define FID_NOBRACE_LEN 40
517 #define FID_LEN (FID_NOBRACE_LEN + 2)
518 #define DFID_NOBRACE LPX64":0x%x:0x%x"
519 #define DFID "["DFID_NOBRACE"]"
525 /* scanf input parse format -- strip '[' first.
526 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
527 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
528 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
529 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
531 #define SFID "0x"LPX64i":0x%x:0x%x"
538 /********* Quotas **********/
540 #define LUSTRE_QUOTABLOCK_BITS 10
541 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
543 static inline __u64 lustre_stoqb(size_t space)
545 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
548 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
549 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
550 #define Q_GETOINFO 0x800102 /* get obd quota info */
551 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
552 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
554 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
555 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
556 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
557 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
558 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
559 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
560 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
561 /* lustre-specific control commands */
562 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
563 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
565 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
567 struct if_quotacheck {
569 struct obd_uuid obd_uuid;
572 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
575 #define N_PERMS_MAX 64
577 struct perm_downcall_data {
583 struct identity_downcall_data {
590 struct perm_downcall_data idd_perms[N_PERMS_MAX];
594 /* for non-mapped uid/gid */
595 #define NOBODY_UID 99
596 #define NOBODY_GID 99
598 #define INVALID_ID (-1)
607 #ifdef NEED_QUOTA_DEFS
609 #define QIF_BLIMITS 1
611 #define QIF_ILIMITS 4
615 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
616 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
617 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
618 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
621 #endif /* !__KERNEL__ */
623 /* lustre volatile file support
624 * file name header: .^L^S^T^R:volatile"
626 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
627 #define LUSTRE_VOLATILE_HDR_LEN 14
629 typedef enum lustre_quota_version {
631 } lustre_quota_version_t;
633 /* XXX: same as if_dqinfo struct in kernel */
641 /* XXX: same as if_dqblk struct in kernel, plus one padding */
643 __u64 dqb_bhardlimit;
644 __u64 dqb_bsoftlimit;
646 __u64 dqb_ihardlimit;
647 __u64 dqb_isoftlimit;
669 struct obd_dqinfo qc_dqinfo;
670 struct obd_dqblk qc_dqblk;
672 struct obd_uuid obd_uuid;
675 /* swap layout flags */
676 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
677 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
678 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
679 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
681 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
682 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
683 struct lustre_swap_layouts {
692 /********* Changelogs **********/
693 /** Changelog record types */
694 enum changelog_rec_type {
696 CL_CREATE = 1, /* namespace */
697 CL_MKDIR = 2, /* namespace */
698 CL_HARDLINK = 3, /* namespace */
699 CL_SOFTLINK = 4, /* namespace */
700 CL_MKNOD = 5, /* namespace */
701 CL_UNLINK = 6, /* namespace */
702 CL_RMDIR = 7, /* namespace */
703 CL_RENAME = 8, /* namespace */
704 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
705 CL_OPEN = 10, /* not currently used */
706 CL_CLOSE = 11, /* may be written to log only with mtime change */
707 CL_LAYOUT = 12, /* file layout/striping modified */
711 CL_HSM = 16, /* HSM specific events, see flags */
712 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
718 static inline const char *changelog_type2str(int type) {
719 static const char *changelog_str[] = {
720 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
721 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
722 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
725 if (type >= 0 && type < CL_LAST)
726 return changelog_str[type];
730 /* per-record flags */
731 #define CLF_VERSION 0x1000
732 #define CLF_EXT_VERSION 0x2000
733 #define CLF_FLAGSHIFT 12
734 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
735 #define CLF_VERMASK (~CLF_FLAGMASK)
736 /* Anything under the flagmask may be per-type (if desired) */
737 /* Flags for unlink */
738 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
739 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
740 /* HSM cleaning needed */
741 /* Flags for rename */
742 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
744 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
745 * has an archive in backend */
748 /* 12b used (from high weight to low weight):
753 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
754 #define CLF_HSM_ERR_H 6
755 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
756 #define CLF_HSM_EVENT_H 9
757 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
758 #define CLF_HSM_FLAG_H 11
759 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
760 #define CLF_HSM_SPARE_H 15
761 #define CLF_HSM_LAST 15
763 /* Remove bits higher than _h, then extract the value
764 * between _h and _l by shifting lower weigth to bit 0. */
765 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
766 >> (CLF_HSM_LAST - _h + _l))
768 #define CLF_HSM_SUCCESS 0x00
769 #define CLF_HSM_MAXERROR 0x7E
770 #define CLF_HSM_ERROVERFLOW 0x7F
772 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
774 /* 3 bits field => 8 values allowed */
786 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
788 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
792 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
794 *flags |= (he << CLF_HSM_EVENT_L);
797 static inline __u16 hsm_get_cl_flags(int flags)
799 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
802 static inline void hsm_set_cl_flags(int *flags, int bits)
804 *flags |= (bits << CLF_HSM_FLAG_L);
807 static inline int hsm_get_cl_error(int flags)
809 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
812 static inline void hsm_set_cl_error(int *flags, int error)
814 *flags |= (error << CLF_HSM_ERR_L);
817 #define CR_MAXSIZE cfs_size_round(2*NAME_MAX + 1 + \
818 sizeof(struct changelog_ext_rec))
820 struct changelog_rec {
822 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
823 __u32 cr_type; /**< \a changelog_rec_type */
824 __u64 cr_index; /**< changelog record number */
825 __u64 cr_prev; /**< last index for this target fid */
828 lustre_fid cr_tfid; /**< target fid */
829 __u32 cr_markerflags; /**< CL_MARK flags */
831 lustre_fid cr_pfid; /**< parent fid */
832 char cr_name[0]; /**< last element */
833 } __attribute__((packed));
835 /* changelog_ext_rec is 2*sizeof(lu_fid) bigger than changelog_rec, to save
836 * space, only rename uses changelog_ext_rec, while others use changelog_rec to
839 struct changelog_ext_rec {
841 __u16 cr_flags; /**< (flags & CLF_FLAGMASK) |
843 __u32 cr_type; /**< \a changelog_rec_type */
844 __u64 cr_index; /**< changelog record number */
845 __u64 cr_prev; /**< last index for this target fid */
848 lustre_fid cr_tfid; /**< target fid */
849 __u32 cr_markerflags; /**< CL_MARK flags */
851 lustre_fid cr_pfid; /**< target parent fid */
852 lustre_fid cr_sfid; /**< source fid, or zero */
853 lustre_fid cr_spfid; /**< source parent fid, or zero */
854 char cr_name[0]; /**< last element */
855 } __attribute__((packed));
857 #define CHANGELOG_REC_EXTENDED(rec) \
858 (((rec)->cr_flags & CLF_VERMASK) == CLF_EXT_VERSION)
860 static inline int changelog_rec_size(struct changelog_rec *rec)
862 return CHANGELOG_REC_EXTENDED(rec) ? sizeof(struct changelog_ext_rec):
866 static inline char *changelog_rec_name(struct changelog_rec *rec)
868 return CHANGELOG_REC_EXTENDED(rec) ?
869 ((struct changelog_ext_rec *)rec)->cr_name: rec->cr_name;
872 static inline int changelog_rec_snamelen(struct changelog_ext_rec *rec)
874 return rec->cr_namelen - strlen(rec->cr_name) - 1;
877 static inline char *changelog_rec_sname(struct changelog_ext_rec *rec)
879 return rec->cr_name + strlen(rec->cr_name) + 1;
882 struct ioc_changelog {
889 enum changelog_message_type {
890 CL_RECORD = 10, /* message is a changelog_rec */
891 CL_EOF = 11, /* at end of current changelog */
894 /********* Misc **********/
896 struct ioc_data_version {
898 __u64 idv_flags; /* See LL_DV_xxx */
900 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
901 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
904 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
907 #define dot_lustre_name ".lustre"
910 /********* HSM **********/
912 /** HSM per-file state
913 * See HSM_FLAGS below.
916 HS_EXISTS = 0x00000001,
917 HS_DIRTY = 0x00000002,
918 HS_RELEASED = 0x00000004,
919 HS_ARCHIVED = 0x00000008,
920 HS_NORELEASE = 0x00000010,
921 HS_NOARCHIVE = 0x00000020,
922 HS_LOST = 0x00000040,
925 /* HSM user-setable flags. */
926 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
928 /* Other HSM flags. */
929 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
932 * All HSM-related possible flags that could be applied to a file.
933 * This should be kept in sync with hsm_states.
935 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
938 * HSMÂ request progress state
940 enum hsm_progress_states {
947 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
950 case HPS_WAITING: return "waiting";
951 case HPS_RUNNING: return "running";
952 case HPS_DONE: return "done";
953 default: return "unknown";
960 } __attribute__((packed));
963 * Current HSM states of a Lustre file.
965 * This structure purpose is to be sent to user-space mainly. It describes the
966 * current HSM flags and in-progress action.
968 struct hsm_user_state {
969 /** Current HSM states, from enum hsm_states. */
971 __u32 hus_archive_id;
972 /** The current undergoing action, if there is one */
973 __u32 hus_in_progress_state;
974 __u32 hus_in_progress_action;
975 struct hsm_extent hus_in_progress_location;
976 char hus_extended_info[];
979 struct hsm_state_set_ioc {
980 struct lu_fid hssi_fid;
982 __u64 hssi_clearmask;
986 * This structure describes the current in-progress action for a file.
987 * it is retuned to user space and send over the wire
989 struct hsm_current_action {
990 /** The current undergoing action, if there is one */
991 /* state is one of hsm_progress_states */
993 /* action is one of hsm_user_action */
995 struct hsm_extent hca_location;
998 /***** HSM user requests ******/
999 /* User-generated (lfs/ioctl) request types */
1000 enum hsm_user_action {
1001 HUA_NONE = 1, /* no action (noop) */
1002 HUA_ARCHIVE = 10, /* copy to hsm */
1003 HUA_RESTORE = 11, /* prestage */
1004 HUA_RELEASE = 12, /* drop ost objects */
1005 HUA_REMOVE = 13, /* remove from archive */
1006 HUA_CANCEL = 14 /* cancel a request */
1009 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1012 case HUA_NONE: return "NOOP";
1013 case HUA_ARCHIVE: return "ARCHIVE";
1014 case HUA_RESTORE: return "RESTORE";
1015 case HUA_RELEASE: return "RELEASE";
1016 case HUA_REMOVE: return "REMOVE";
1017 case HUA_CANCEL: return "CANCEL";
1018 default: return "UNKNOWN";
1023 * List of hr_flags (bit field)
1025 #define HSM_FORCE_ACTION 0x0001
1026 /* used by CT, connot be set by user */
1027 #define HSM_GHOST_COPY 0x0002
1030 * Contains all the fixed part of struct hsm_user_request.
1033 struct hsm_request {
1034 __u32 hr_action; /* enum hsm_user_action */
1035 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1036 __u64 hr_flags; /* request flags */
1037 __u32 hr_itemcount; /* item count in hur_user_item vector */
1041 struct hsm_user_item {
1043 struct hsm_extent hui_extent;
1044 } __attribute__((packed));
1046 struct hsm_user_request {
1047 struct hsm_request hur_request;
1048 struct hsm_user_item hur_user_item[0];
1049 /* extra data blob at end of struct (after all
1050 * hur_user_items), only use helpers to access it
1052 } __attribute__((packed));
1054 /** Return pointer to data field in a hsm user request */
1055 static inline void *hur_data(struct hsm_user_request *hur)
1057 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1061 * Compute the current length of the provided hsm_user_request. This returns -1
1062 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1064 * return -1 on bounds check error.
1066 static inline ssize_t hur_len(struct hsm_user_request *hur)
1070 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1071 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1072 (__u64)hur->hur_request.hr_itemcount *
1073 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1075 if (size != (ssize_t)size)
1081 /****** HSM RPCs to copytool *****/
1082 /* Message types the copytool may receive */
1083 enum hsm_message_type {
1084 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1087 /* Actions the copytool may be instructed to take for a given action_item */
1088 enum hsm_copytool_action {
1089 HSMA_NONE = 10, /* no action */
1090 HSMA_ARCHIVE = 20, /* arbitrary offset */
1096 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1099 case HSMA_NONE: return "NOOP";
1100 case HSMA_ARCHIVE: return "ARCHIVE";
1101 case HSMA_RESTORE: return "RESTORE";
1102 case HSMA_REMOVE: return "REMOVE";
1103 case HSMA_CANCEL: return "CANCEL";
1104 default: return "UNKNOWN";
1108 /* Copytool item action description */
1109 struct hsm_action_item {
1110 __u32 hai_len; /* valid size of this struct */
1111 __u32 hai_action; /* hsm_copytool_action, but use known size */
1112 lustre_fid hai_fid; /* Lustre FID to operated on */
1113 lustre_fid hai_dfid; /* fid used for data access */
1114 struct hsm_extent hai_extent; /* byte range to operate on */
1115 __u64 hai_cookie; /* action cookie from coordinator */
1116 __u64 hai_gid; /* grouplock id */
1117 char hai_data[0]; /* variable length */
1118 } __attribute__((packed));
1121 * helper function which print in hexa the first bytes of
1123 * \param hai [IN] record to print
1124 * \param buffer [OUT] output buffer
1125 * \param len [IN] max buffer len
1128 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1129 char *buffer, int len)
1131 int i, sz, data_len;
1136 data_len = hai->hai_len - sizeof(*hai);
1137 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1141 cnt = snprintf(ptr, sz, "%.2X",
1142 (unsigned char)hai->hai_data[i]);
1150 /* Copytool action list */
1151 #define HAL_VERSION 1
1152 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1153 struct hsm_action_list {
1155 __u32 hal_count; /* number of hai's to follow */
1156 __u64 hal_compound_id; /* returned by coordinator */
1158 __u32 hal_archive_id; /* which archive backend */
1160 char hal_fsname[0]; /* null-terminated */
1161 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1162 boundaries. See hai_zero */
1163 } __attribute__((packed));
1165 #ifndef HAVE_CFS_SIZE_ROUND
1166 static inline int cfs_size_round (int val)
1168 return (val + 7) & (~0x7);
1170 #define HAVE_CFS_SIZE_ROUND
1173 /* Return pointer to first hai in action list */
1174 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1176 return (struct hsm_action_item *)(hal->hal_fsname +
1177 cfs_size_round(strlen(hal-> \
1181 /* Return pointer to next hai */
1182 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1184 return (struct hsm_action_item *)((char *)hai +
1185 cfs_size_round(hai->hai_len));
1188 /* Return size of an hsm_action_list */
1189 static inline size_t hal_size(struct hsm_action_list *hal)
1193 struct hsm_action_item *hai;
1195 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1196 hai = hai_first(hal);
1197 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1198 sz += cfs_size_round(hai->hai_len);
1204 * describe the attributes to be set on imported file
1206 struct hsm_user_import {
1215 __u32 hui_archive_id;
1218 /* Copytool progress reporting */
1219 #define HP_FLAG_COMPLETED 0x01
1220 #define HP_FLAG_RETRY 0x02
1222 struct hsm_progress {
1225 struct hsm_extent hp_extent;
1227 __u16 hp_errval; /* positive val */
1232 __u64 hc_data_version;
1234 __u16 hc_errval; /* positive val */
1236 struct hsm_action_item hc_hai;
1240 enum llapi_json_types {
1241 LLAPI_JSON_INTEGER = 1,
1247 struct llapi_json_item {
1256 struct llapi_json_item *lji_next;
1259 struct llapi_json_item_list {
1260 int ljil_item_count;
1261 struct llapi_json_item *ljil_items;
1264 /** @} lustreuser */
1266 #endif /* _LUSTRE_USER_H */