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, 2012, 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 "ll_fiemap.h"
50 #include "../linux/lustre_user.h"
53 #define LL_SUPER_MAGIC 0x0BD00BD0
55 #ifndef FSFILT_IOC_GETFLAGS
56 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
57 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
58 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
59 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
60 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
61 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
62 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
65 /* FIEMAP flags supported by Lustre */
66 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
68 enum obd_statfs_state {
69 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
70 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
71 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
72 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
73 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
87 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
88 __u32 os_fprecreated; /* objs available now to the caller */
89 /* used in QoS code to find preferred
104 * FID is a cluster-wide unique identifier of a file or an object (stripe).
105 * FIDs are never reused.
109 * FID sequence. Sequence is a unit of migration: all files (objects)
110 * with FIDs from a given sequence are stored on the same server.
111 * Lustre should support 2^64 objects, so even if each sequence
112 * has only a single object we can still enumerate 2^64 objects.
115 /* FID number within sequence. */
118 * FID version, used to distinguish different versions (in the sense
119 * of snapshots, etc.) of the same file system object. Not currently
126 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
129 /* keep this one for compatibility */
130 struct filter_fid_old {
131 struct lu_fid ff_parent;
136 /* Userspace should treat lu_fid as opaque, and only use the following methods
137 * to print or parse them. Other functions (e.g. compare, swab) could be moved
138 * here from lustre_idl.h if needed. */
139 typedef struct lu_fid lustre_fid;
142 * Following struct for object attributes, that will be kept inode's EA.
143 * Introduced in 2.0 release (please see b15993, for details)
144 * Added to all objects since Lustre 2.4 as contains self FID
146 struct lustre_mdt_attrs {
148 * Bitfield for supported data in this structure. From enum lma_compat.
149 * lma_self_fid and lma_flags are always available.
153 * Per-file incompat feature list. Lustre version should support all
154 * flags set in this field. The supported feature mask is available in
158 /** FID of this inode */
159 struct lu_fid lma_self_fid;
163 * Prior to 2.4, the LMA structure also included SOM attributes which has since
164 * been moved to a dedicated xattr
165 * lma_flags was also removed because of lma_compat/incompat fields.
167 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
170 * OST object IDentifier.
178 struct lu_fid oi_fid;
182 #define DOSTID "%#llx:%llu"
183 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
186 * The ioctl naming rules:
187 * LL_* - works on the currently opened filehandle instead of parent dir
188 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
189 * *_MDC_* - gets/sets data related to MDC
190 * *_LOV_* - gets/sets data related to OSC/LOV
191 * *FILE* - called on parent dir and passes in a filename
192 * *STRIPE* - set/get lov_user_md
193 * *INFO - set/get lov_user_mds_data
195 /* see <lustre_lib.h> for ioctl numberss 101-150 */
196 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
197 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
198 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
199 /* LL_IOC_LOV_SETSTRIPE: See also OBD_IOC_LOV_SETSTRIPE */
200 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
201 /* LL_IOC_LOV_GETSTRIPE: See also OBD_IOC_LOV_GETSTRIPE */
202 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
203 /* LL_IOC_LOV_SETEA: See also OBD_IOC_LOV_SETEA */
204 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
205 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
206 #define LL_IOC_RECREATE_FID _IOW ('f', 157, struct lu_fid)
207 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
208 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
209 /* LL_IOC_QUOTACHECK: See also OBD_IOC_QUOTACHECK */
210 #define LL_IOC_QUOTACHECK _IOW ('f', 160, int)
211 /* LL_IOC_POLL_QUOTACHECK: See also OBD_IOC_POLL_QUOTACHECK */
212 #define LL_IOC_POLL_QUOTACHECK _IOR ('f', 161, struct if_quotacheck *)
213 /* LL_IOC_QUOTACTL: See also OBD_IOC_QUOTACTL */
214 #define LL_IOC_QUOTACTL _IOWR('f', 162, struct if_quotactl)
215 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
216 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
217 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
218 #define LL_IOC_RMTACL _IOW ('f', 167, long)
219 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
220 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
221 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
222 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
223 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
224 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
225 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
226 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
228 /* see <lustre_lib.h> for ioctl numbers 177-210 */
230 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
231 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
232 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
233 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
234 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
235 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
236 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
237 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
238 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
239 struct lustre_swap_layouts)
240 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
241 struct hsm_current_action)
242 /* see <lustre_lib.h> for ioctl numbers 221-232 */
244 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
245 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
246 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
247 #define LL_IOC_GET_LEASE _IO('f', 244)
248 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
250 #define LL_STATFS_LMV 1
251 #define LL_STATFS_LOV 2
252 #define LL_STATFS_NODELAY 4
254 #define IOC_MDC_TYPE 'i'
255 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
256 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
257 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
258 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
260 /* Keep these for backward compartability. */
261 #define LL_IOC_OBD_STATFS IOC_OBD_STATFS
262 #define IOC_MDC_GETSTRIPE IOC_MDC_GETFILESTRIPE
264 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
266 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
267 * files, but are unlikely to be used in practice and are not harmful if
268 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
269 * devices and are safe for use on new files (See LU-812, LU-4209). */
270 #define O_LOV_DELAY_CREATE (O_NOCTTY | FASYNC)
272 #define LL_FILE_IGNORE_LOCK 0x00000001
273 #define LL_FILE_GROUP_LOCKED 0x00000002
274 #define LL_FILE_READAHEA 0x00000004
275 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
276 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
277 #define LL_FILE_RMTACL 0x00000020
279 #define LOV_USER_MAGIC_V1 0x0BD10BD0
280 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
281 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
282 #define LOV_USER_MAGIC_V3 0x0BD30BD0
284 #define LMV_MAGIC_V1 0x0CD10CD0 /*normal stripe lmv magic */
285 #define LMV_USER_MAGIC 0x0CD20CD0 /*default lmv magic*/
287 #define LOV_PATTERN_RAID0 0x001
288 #define LOV_PATTERN_RAID1 0x002
289 #define LOV_PATTERN_FIRST 0x100
291 #define LOV_MAXPOOLNAME 16
292 #define LOV_POOLNAMEF "%.16s"
294 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
295 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
296 #define LOV_MAX_STRIPE_COUNT_OLD 160
297 /* This calculation is crafted so that input of 4096 will result in 160
298 * which in turn is equal to old maximal stripe count.
299 * XXX: In fact this is too simplified for now, what it also need is to get
300 * ea_type argument to clearly know how much space each stripe consumes.
302 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
303 * allocation that is sufficient for the current generation of systems.
305 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
306 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
307 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
308 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
310 #define lov_user_ost_data lov_user_ost_data_v1
311 struct lov_user_ost_data_v1 { /* per-stripe data structure */
312 struct ost_id l_ost_oi; /* OST object ID */
313 __u32 l_ost_gen; /* generation of this OST index */
314 __u32 l_ost_idx; /* OST index in LOV */
317 #define lov_user_md lov_user_md_v1
318 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
319 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
320 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
321 struct ost_id lmm_oi; /* LOV object ID */
322 __u32 lmm_stripe_size; /* size of stripe in bytes */
323 __u16 lmm_stripe_count; /* num stripes in use for this object */
325 __u16 lmm_stripe_offset; /* starting stripe offset in
326 * lmm_objects, use when writing */
327 __u16 lmm_layout_gen; /* layout generation number
328 * used when reading */
330 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
331 } __attribute__((packed, __may_alias__));
333 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
334 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
335 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
336 struct ost_id lmm_oi; /* LOV object ID */
337 __u32 lmm_stripe_size; /* size of stripe in bytes */
338 __u16 lmm_stripe_count; /* num stripes in use for this object */
340 __u16 lmm_stripe_offset; /* starting stripe offset in
341 * lmm_objects, use when writing */
342 __u16 lmm_layout_gen; /* layout generation number
343 * used when reading */
345 char lmm_pool_name[LOV_MAXPOOLNAME]; /* pool name */
346 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
349 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
351 if (lmm_magic == LOV_USER_MAGIC_V3)
352 return sizeof(struct lov_user_md_v3) +
353 stripes * sizeof(struct lov_user_ost_data_v1);
355 return sizeof(struct lov_user_md_v1) +
356 stripes * sizeof(struct lov_user_ost_data_v1);
359 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
360 * use this. It is unsafe to #define those values in this header as it
361 * is possible the application has already #included <sys/stat.h>. */
362 #ifdef HAVE_LOV_USER_MDS_DATA
363 #define lov_user_mds_data lov_user_mds_data_v1
364 struct lov_user_mds_data_v1 {
365 lstat_t lmd_st; /* MDS stat struct */
366 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
369 struct lov_user_mds_data_v3 {
370 lstat_t lmd_st; /* MDS stat struct */
371 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
375 /* keep this to be the same size as lov_user_ost_data_v1 */
376 struct lmv_user_mds_data {
377 struct lu_fid lum_fid;
385 LMV_DEFAULT_TYPE = 1,
388 #define lmv_user_md lmv_user_md_v1
389 struct lmv_user_md_v1 {
390 __u32 lum_magic; /* must be the first field */
391 __u32 lum_stripe_count; /* dirstripe count */
392 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
393 __u32 lum_hash_type; /* Dir stripe policy */
394 __u32 lum_type; /* LMV type: default or normal */
398 char lum_pool_name[LOV_MAXPOOLNAME];
399 struct lmv_user_mds_data lum_objects[0];
402 static inline int lmv_user_md_size(int stripes, int lmm_magic)
404 return sizeof(struct lmv_user_md) +
405 stripes * sizeof(struct lmv_user_mds_data);
408 struct ll_recreate_obj {
414 __u64 id; /* holds object id */
415 __u32 generation; /* holds object generation */
416 __u32 f_type; /* holds object type or stripe idx when passing it to
417 * OST for saving into EA. */
425 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
426 const struct obd_uuid *u2)
428 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
431 static inline int obd_uuid_empty(struct obd_uuid *uuid)
433 return uuid->uuid[0] == '\0';
436 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
438 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
439 uuid->uuid[sizeof(*uuid) - 1] = '\0';
442 /* For printf's only, make sure uuid is terminated */
443 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
445 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
446 /* Obviously not safe, but for printfs, no real harm done...
447 we're always null-terminated, even in a race. */
448 static char temp[sizeof(*uuid)];
450 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
451 temp[sizeof(*uuid) - 1] = '\0';
454 return (char *)(uuid->uuid);
457 /* Extract fsname from uuid (or target name) of a target
458 e.g. (myfs-OST0007_UUID -> myfs)
459 see also deuuidify. */
460 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
464 strncpy(buf, uuid, buflen - 1);
465 buf[buflen - 1] = '\0';
466 p = strrchr(buf, '-');
471 /* printf display format
472 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
473 #define FID_NOBRACE_LEN 40
474 #define FID_LEN (FID_NOBRACE_LEN + 2)
475 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
476 #define DFID "["DFID_NOBRACE"]"
482 /* scanf input parse format -- strip '[' first.
483 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
484 #define SFID "0x%llx:0x%x:0x%x"
490 /********* Quotas **********/
492 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
493 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
494 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
495 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
496 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
497 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
498 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
499 /* lustre-specific control commands */
500 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
501 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
503 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
505 struct if_quotacheck {
507 struct obd_uuid obd_uuid;
510 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
513 #define N_PERMS_MAX 64
515 struct perm_downcall_data {
521 struct identity_downcall_data {
528 struct perm_downcall_data idd_perms[N_PERMS_MAX];
532 /* for non-mapped uid/gid */
533 #define NOBODY_UID 99
534 #define NOBODY_GID 99
536 #define INVALID_ID (-1)
545 #ifdef NEED_QUOTA_DEFS
547 #define QIF_BLIMITS 1
549 #define QIF_ILIMITS 4
553 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
554 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
555 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
556 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
559 #endif /* !__KERNEL__ */
561 /* lustre volatile file support
562 * file name header: .^L^S^T^R:volatile"
564 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
565 #define LUSTRE_VOLATILE_HDR_LEN 14
566 /* hdr + MDT index */
567 #define LUSTRE_VOLATILE_IDX LUSTRE_VOLATILE_HDR":%.4X:"
569 typedef enum lustre_quota_version {
571 } lustre_quota_version_t;
573 /* XXX: same as if_dqinfo struct in kernel */
581 /* XXX: same as if_dqblk struct in kernel, plus one padding */
583 __u64 dqb_bhardlimit;
584 __u64 dqb_bsoftlimit;
586 __u64 dqb_ihardlimit;
587 __u64 dqb_isoftlimit;
609 struct obd_dqinfo qc_dqinfo;
610 struct obd_dqblk qc_dqblk;
612 struct obd_uuid obd_uuid;
615 /* swap layout flags */
616 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
617 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
618 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
619 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
621 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
622 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
623 struct lustre_swap_layouts {
631 /********* Changelogs **********/
632 /** Changelog record types */
633 enum changelog_rec_type {
635 CL_CREATE = 1, /* namespace */
636 CL_MKDIR = 2, /* namespace */
637 CL_HARDLINK = 3, /* namespace */
638 CL_SOFTLINK = 4, /* namespace */
639 CL_MKNOD = 5, /* namespace */
640 CL_UNLINK = 6, /* namespace */
641 CL_RMDIR = 7, /* namespace */
642 CL_RENAME = 8, /* namespace */
643 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
644 CL_OPEN = 10, /* not currently used */
645 CL_CLOSE = 11, /* may be written to log only with mtime change */
646 CL_LAYOUT = 12, /* file layout/striping modified */
650 CL_HSM = 16, /* HSM specific events, see flags */
651 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
657 static inline const char *changelog_type2str(int type)
659 static const char *changelog_str[] = {
660 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
661 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
662 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
665 if (type >= 0 && type < CL_LAST)
666 return changelog_str[type];
670 /* per-record flags */
671 #define CLF_VERSION 0x1000
672 #define CLF_EXT_VERSION 0x2000
673 #define CLF_FLAGSHIFT 12
674 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
675 #define CLF_VERMASK (~CLF_FLAGMASK)
676 /* Anything under the flagmask may be per-type (if desired) */
677 /* Flags for unlink */
678 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
679 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
680 /* HSM cleaning needed */
681 /* Flags for rename */
682 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink of target */
685 /* 12b used (from high weight to low weight):
690 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
691 #define CLF_HSM_ERR_H 6
692 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
693 #define CLF_HSM_EVENT_H 9
694 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
695 #define CLF_HSM_FLAG_H 11
696 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
697 #define CLF_HSM_SPARE_H 15
698 #define CLF_HSM_LAST 15
700 /* Remove bits higher than _h, then extract the value
701 * between _h and _l by shifting lower weigth to bit 0. */
702 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
703 >> (CLF_HSM_LAST - _h + _l))
705 #define CLF_HSM_SUCCESS 0x00
706 #define CLF_HSM_MAXERROR 0x7E
707 #define CLF_HSM_ERROVERFLOW 0x7F
709 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
711 /* 3 bits field => 8 values allowed */
723 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
725 return CLF_GET_BITS(flags, CLF_HSM_EVENT_H, CLF_HSM_EVENT_L);
728 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
730 *flags |= (he << CLF_HSM_EVENT_L);
733 static inline __u16 hsm_get_cl_flags(int flags)
735 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
738 static inline void hsm_set_cl_flags(int *flags, int bits)
740 *flags |= (bits << CLF_HSM_FLAG_L);
743 static inline int hsm_get_cl_error(int flags)
745 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
748 static inline void hsm_set_cl_error(int *flags, int error)
750 *flags |= (error << CLF_HSM_ERR_L);
753 #define CR_MAXSIZE cfs_size_round(2*NAME_MAX + 1 + \
754 sizeof(struct changelog_ext_rec))
756 struct changelog_rec {
758 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
759 __u32 cr_type; /**< \a changelog_rec_type */
760 __u64 cr_index; /**< changelog record number */
761 __u64 cr_prev; /**< last index for this target fid */
764 lustre_fid cr_tfid; /**< target fid */
765 __u32 cr_markerflags; /**< CL_MARK flags */
767 lustre_fid cr_pfid; /**< parent fid */
768 char cr_name[0]; /**< last element */
771 /* changelog_ext_rec is 2*sizeof(lu_fid) bigger than changelog_rec, to save
772 * space, only rename uses changelog_ext_rec, while others use changelog_rec to
775 struct changelog_ext_rec {
777 __u16 cr_flags; /**< (flags & CLF_FLAGMASK) |
779 __u32 cr_type; /**< \a changelog_rec_type */
780 __u64 cr_index; /**< changelog record number */
781 __u64 cr_prev; /**< last index for this target fid */
784 lustre_fid cr_tfid; /**< target fid */
785 __u32 cr_markerflags; /**< CL_MARK flags */
787 lustre_fid cr_pfid; /**< target parent fid */
788 lustre_fid cr_sfid; /**< source fid, or zero */
789 lustre_fid cr_spfid; /**< source parent fid, or zero */
790 char cr_name[0]; /**< last element */
793 #define CHANGELOG_REC_EXTENDED(rec) \
794 (((rec)->cr_flags & CLF_VERMASK) == CLF_EXT_VERSION)
796 static inline int changelog_rec_size(struct changelog_rec *rec)
798 return CHANGELOG_REC_EXTENDED(rec) ? sizeof(struct changelog_ext_rec) :
802 static inline char *changelog_rec_name(struct changelog_rec *rec)
804 return CHANGELOG_REC_EXTENDED(rec) ?
805 ((struct changelog_ext_rec *)rec)->cr_name : rec->cr_name;
808 static inline int changelog_rec_snamelen(struct changelog_ext_rec *rec)
810 return rec->cr_namelen - strlen(rec->cr_name) - 1;
813 static inline char *changelog_rec_sname(struct changelog_ext_rec *rec)
815 return rec->cr_name + strlen(rec->cr_name) + 1;
818 struct ioc_changelog {
825 enum changelog_message_type {
826 CL_RECORD = 10, /* message is a changelog_rec */
827 CL_EOF = 11, /* at end of current changelog */
830 /********* Misc **********/
832 struct ioc_data_version {
834 __u64 idv_flags; /* See LL_DV_xxx */
837 #define LL_DV_NOFLUSH 0x01 /* Do not take READ EXTENT LOCK before sampling
838 version. Dirty caches are left unchanged. */
841 # define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
844 #define dot_lustre_name ".lustre"
846 /********* HSM **********/
848 /** HSM per-file state
849 * See HSM_FLAGS below.
852 HS_EXISTS = 0x00000001,
853 HS_DIRTY = 0x00000002,
854 HS_RELEASED = 0x00000004,
855 HS_ARCHIVED = 0x00000008,
856 HS_NORELEASE = 0x00000010,
857 HS_NOARCHIVE = 0x00000020,
858 HS_LOST = 0x00000040,
861 /* HSM user-setable flags. */
862 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
864 /* Other HSM flags. */
865 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
868 * All HSM-related possible flags that could be applied to a file.
869 * This should be kept in sync with hsm_states.
871 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
874 * HSMÂ request progress state
876 enum hsm_progress_states {
884 static inline char *hsm_progress_state2name(enum hsm_progress_states s)
887 case HPS_WAITING: return "waiting";
888 case HPS_RUNNING: return "running";
889 case HPS_DONE: return "done";
890 default: return "unknown";
900 * Current HSM states of a Lustre file.
902 * This structure purpose is to be sent to user-space mainly. It describes the
903 * current HSM flags and in-progress action.
905 struct hsm_user_state {
906 /** Current HSM states, from enum hsm_states. */
908 __u32 hus_archive_id;
909 /** The current undergoing action, if there is one */
910 __u32 hus_in_progress_state;
911 __u32 hus_in_progress_action;
912 struct hsm_extent hus_in_progress_location;
913 char hus_extended_info[];
916 struct hsm_state_set_ioc {
917 struct lu_fid hssi_fid;
919 __u64 hssi_clearmask;
923 * This structure describes the current in-progress action for a file.
924 * it is returned to user space and send over the wire
926 struct hsm_current_action {
927 /** The current undergoing action, if there is one */
928 /* state is one of hsm_progress_states */
930 /* action is one of hsm_user_action */
932 struct hsm_extent hca_location;
935 /***** HSM user requests ******/
936 /* User-generated (lfs/ioctl) request types */
937 enum hsm_user_action {
938 HUA_NONE = 1, /* no action (noop) */
939 HUA_ARCHIVE = 10, /* copy to hsm */
940 HUA_RESTORE = 11, /* prestage */
941 HUA_RELEASE = 12, /* drop ost objects */
942 HUA_REMOVE = 13, /* remove from archive */
943 HUA_CANCEL = 14 /* cancel a request */
946 static inline char *hsm_user_action2name(enum hsm_user_action a)
949 case HUA_NONE: return "NOOP";
950 case HUA_ARCHIVE: return "ARCHIVE";
951 case HUA_RESTORE: return "RESTORE";
952 case HUA_RELEASE: return "RELEASE";
953 case HUA_REMOVE: return "REMOVE";
954 case HUA_CANCEL: return "CANCEL";
955 default: return "UNKNOWN";
960 * List of hr_flags (bit field)
962 #define HSM_FORCE_ACTION 0x0001
963 /* used by CT, connot be set by user */
964 #define HSM_GHOST_COPY 0x0002
967 * Contains all the fixed part of struct hsm_user_request.
971 __u32 hr_action; /* enum hsm_user_action */
972 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
973 __u64 hr_flags; /* request flags */
974 __u32 hr_itemcount; /* item count in hur_user_item vector */
978 struct hsm_user_item {
980 struct hsm_extent hui_extent;
983 struct hsm_user_request {
984 struct hsm_request hur_request;
985 struct hsm_user_item hur_user_item[0];
986 /* extra data blob at end of struct (after all
987 * hur_user_items), only use helpers to access it
991 /** Return pointer to data field in a hsm user request */
992 static inline void *hur_data(struct hsm_user_request *hur)
994 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
998 * Compute the current length of the provided hsm_user_request. This returns -1
999 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1001 * return -1 on bounds check error.
1003 static inline ssize_t hur_len(struct hsm_user_request *hur)
1007 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1008 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1009 (__u64)hur->hur_request.hr_itemcount *
1010 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1012 if (size != (ssize_t)size)
1018 /****** HSM RPCs to copytool *****/
1019 /* Message types the copytool may receive */
1020 enum hsm_message_type {
1021 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1024 /* Actions the copytool may be instructed to take for a given action_item */
1025 enum hsm_copytool_action {
1026 HSMA_NONE = 10, /* no action */
1027 HSMA_ARCHIVE = 20, /* arbitrary offset */
1033 static inline char *hsm_copytool_action2name(enum hsm_copytool_action a)
1036 case HSMA_NONE: return "NOOP";
1037 case HSMA_ARCHIVE: return "ARCHIVE";
1038 case HSMA_RESTORE: return "RESTORE";
1039 case HSMA_REMOVE: return "REMOVE";
1040 case HSMA_CANCEL: return "CANCEL";
1041 default: return "UNKNOWN";
1045 /* Copytool item action description */
1046 struct hsm_action_item {
1047 __u32 hai_len; /* valid size of this struct */
1048 __u32 hai_action; /* hsm_copytool_action, but use known size */
1049 lustre_fid hai_fid; /* Lustre FID to operated on */
1050 lustre_fid hai_dfid; /* fid used for data access */
1051 struct hsm_extent hai_extent; /* byte range to operate on */
1052 __u64 hai_cookie; /* action cookie from coordinator */
1053 __u64 hai_gid; /* grouplock id */
1054 char hai_data[0]; /* variable length */
1058 * helper function which print in hexa the first bytes of
1060 * \param hai [IN] record to print
1061 * \param buffer [OUT] output buffer
1062 * \param len [IN] max buffer len
1065 static inline char *hai_dump_data_field(struct hsm_action_item *hai,
1066 char *buffer, int len)
1068 int i, sz, data_len;
1073 data_len = hai->hai_len - sizeof(*hai);
1074 for (i = 0 ; (i < data_len) && (sz > 0) ; i++) {
1077 cnt = snprintf(ptr, sz, "%.2X",
1078 (unsigned char)hai->hai_data[i]);
1086 /* Copytool action list */
1087 #define HAL_VERSION 1
1088 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1089 struct hsm_action_list {
1091 __u32 hal_count; /* number of hai's to follow */
1092 __u64 hal_compound_id; /* returned by coordinator */
1094 __u32 hal_archive_id; /* which archive backend */
1096 char hal_fsname[0]; /* null-terminated */
1097 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1098 boundaries. See hai_zero */
1101 #ifndef HAVE_CFS_SIZE_ROUND
1102 static inline int cfs_size_round (int val)
1104 return (val + 7) & (~0x7);
1107 #define HAVE_CFS_SIZE_ROUND
1110 /* Return pointer to first hai in action list */
1111 static inline struct hsm_action_item *hai_zero(struct hsm_action_list *hal)
1113 return (struct hsm_action_item *)(hal->hal_fsname +
1114 cfs_size_round(strlen(hal-> \
1119 /* Return pointer to next hai */
1120 static inline struct hsm_action_item *hai_next(struct hsm_action_item *hai)
1122 return (struct hsm_action_item *)((char *)hai +
1123 cfs_size_round(hai->hai_len));
1126 /* Return size of an hsm_action_list */
1127 static inline int hal_size(struct hsm_action_list *hal)
1130 struct hsm_action_item *hai;
1132 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1133 hai = hai_zero(hal);
1134 for (i = 0; i < hal->hal_count; i++, hai = hai_next(hai))
1135 sz += cfs_size_round(hai->hai_len);
1141 * describe the attributes to be set on imported file
1143 struct hsm_user_import {
1152 __u32 hui_archive_id;
1155 /* Copytool progress reporting */
1156 #define HP_FLAG_COMPLETED 0x01
1157 #define HP_FLAG_RETRY 0x02
1159 struct hsm_progress {
1162 struct hsm_extent hp_extent;
1164 __u16 hp_errval; /* positive val */
1169 __u64 hc_data_version;
1171 __u16 hc_errval; /* positive val */
1173 struct hsm_action_item hc_hai;
1176 /** @} lustreuser */
1178 #endif /* _LUSTRE_USER_H */