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
50 #include <libcfs/posix/posix-types.h>
52 #include <lustre/ll_fiemap.h>
53 #if defined(__linux__)
54 #include <linux/lustre_user.h>
55 #elif defined(__APPLE__)
56 #include <darwin/lustre_user.h>
57 #elif defined(__WINNT__)
58 #include <winnt/lustre_user.h>
60 #error Unsupported operating system.
64 #define LL_SUPER_MAGIC 0x0BD00BD0
66 #ifndef FSFILT_IOC_GETFLAGS
67 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
68 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
69 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
70 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
71 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
72 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
73 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
76 /* FIEMAP flags supported by Lustre */
77 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
79 enum obd_statfs_state {
80 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
81 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
82 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
83 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
84 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
98 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
99 __u32 os_fprecreated; /* objs available now to the caller */
100 /* used in QoS code to find preferred
115 * FID is a cluster-wide unique identifier of a file or an object (stripe).
116 * FIDs are never reused.
120 * FID sequence. Sequence is a unit of migration: all files (objects)
121 * with FIDs from a given sequence are stored on the same server.
122 * Lustre should support 2^64 objects, so even if each sequence
123 * has only a single object we can still enumerate 2^64 objects.
126 /* FID number within sequence. */
129 * FID version, used to distinguish different versions (in the sense
130 * of snapshots, etc.) of the same file system object. Not currently
137 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
140 /* keep this one for compatibility */
141 struct filter_fid_old {
142 struct lu_fid ff_parent;
147 /* Userspace should treat lu_fid as opaque, and only use the following methods
148 * to print or parse them. Other functions (e.g. compare, swab) could be moved
149 * here from lustre_idl.h if needed. */
150 typedef struct lu_fid lustre_fid;
153 * Following struct for object attributes, that will be kept inode's EA.
154 * Introduced in 2.0 release (please see b15993, for details)
155 * Added to all objects since Lustre 2.4 as contains self FID
157 struct lustre_mdt_attrs {
159 * Bitfield for supported data in this structure. From enum lma_compat.
160 * lma_self_fid and lma_flags are always available.
164 * Per-file incompat feature list. Lustre version should support all
165 * flags set in this field. The supported feature mask is available in
169 /** FID of this inode */
170 struct lu_fid lma_self_fid;
174 * Prior to 2.4, the LMA structure also included SOM attributes which has since
175 * been moved to a dedicated xattr
176 * lma_flags was also removed because of lma_compat/incompat fields.
178 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
181 * OST object IDentifier.
189 struct lu_fid oi_fid;
193 #define DOSTID LPX64":"LPU64
194 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
197 * The ioctl naming rules:
198 * LL_* - works on the currently opened filehandle instead of parent dir
199 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
200 * *_MDC_* - gets/sets data related to MDC
201 * *_LOV_* - gets/sets data related to OSC/LOV
202 * *FILE* - called on parent dir and passes in a filename
203 * *STRIPE* - set/get lov_user_md
204 * *INFO - set/get lov_user_mds_data
206 /* see <lustre_lib.h> for ioctl numberss 101-150 */
207 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
208 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
209 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
210 /* LL_IOC_LOV_SETSTRIPE: See also OBD_IOC_LOV_SETSTRIPE */
211 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
212 /* LL_IOC_LOV_GETSTRIPE: See also OBD_IOC_LOV_GETSTRIPE */
213 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
214 /* LL_IOC_LOV_SETEA: See also OBD_IOC_LOV_SETEA */
215 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
216 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
217 #define LL_IOC_RECREATE_FID _IOW ('f', 157, struct lu_fid)
218 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
219 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
220 /* LL_IOC_QUOTACHECK: See also OBD_IOC_QUOTACHECK */
221 #define LL_IOC_QUOTACHECK _IOW ('f', 160, int)
222 /* LL_IOC_POLL_QUOTACHECK: See also OBD_IOC_POLL_QUOTACHECK */
223 #define LL_IOC_POLL_QUOTACHECK _IOR ('f', 161, struct if_quotacheck *)
224 /* LL_IOC_QUOTACTL: See also OBD_IOC_QUOTACTL */
225 #define LL_IOC_QUOTACTL _IOWR('f', 162, struct if_quotactl)
226 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
227 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
228 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
229 #define LL_IOC_RMTACL _IOW ('f', 167, long)
230 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
231 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
232 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
233 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
234 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
235 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
236 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
237 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
239 /* see <lustre_lib.h> for ioctl numbers 177-210 */
241 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
242 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
243 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
244 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
245 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
246 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
247 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
248 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
249 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
250 struct lustre_swap_layouts)
251 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
252 struct hsm_current_action)
253 /* see <lustre_lib.h> for ioctl numbers 221-232 */
255 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
256 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
257 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
259 #define LL_STATFS_LMV 1
260 #define LL_STATFS_LOV 2
261 #define LL_STATFS_NODELAY 4
263 #define IOC_MDC_TYPE 'i'
264 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
265 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
266 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
267 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
269 /* Keep these for backward compartability. */
270 #define LL_IOC_OBD_STATFS IOC_OBD_STATFS
271 #define IOC_MDC_GETSTRIPE IOC_MDC_GETFILESTRIPE
274 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
276 /* Hopefully O_LOV_DELAY_CREATE does not conflict with standard O_xxx flags.
277 * Previously it was defined as 0100000000 and conflicts with FMODE_NONOTIFY
278 * which was added since kernel 2.6.36, so we redefine it as 020000000.
279 * To be compatible with old version's statically linked binary, finally we
280 * define it as (020000000 | 0100000000).
282 #define O_LOV_DELAY_CREATE 0120000000
284 #define LL_FILE_IGNORE_LOCK 0x00000001
285 #define LL_FILE_GROUP_LOCKED 0x00000002
286 #define LL_FILE_READAHEA 0x00000004
287 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
288 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
289 #define LL_FILE_RMTACL 0x00000020
291 #define LOV_USER_MAGIC_V1 0x0BD10BD0
292 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
293 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
294 #define LOV_USER_MAGIC_V3 0x0BD30BD0
296 #define LMV_MAGIC_V1 0x0CD10CD0 /*normal stripe lmv magic */
297 #define LMV_USER_MAGIC 0x0CD20CD0 /*default lmv magic*/
299 #define LOV_PATTERN_RAID0 0x001
300 #define LOV_PATTERN_RAID1 0x002
301 #define LOV_PATTERN_FIRST 0x100
303 #define LOV_MAXPOOLNAME 16
304 #define LOV_POOLNAMEF "%.16s"
306 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
307 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
308 #define LOV_MAX_STRIPE_COUNT_OLD 160
309 /* This calculation is crafted so that input of 4096 will result in 160
310 * which in turn is equal to old maximal stripe count.
311 * XXX: In fact this is too simpified for now, what it also need is to get
312 * ea_type argument to clearly know how much space each stripe consumes.
314 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
315 * allocation that is sufficient for the current generation of systems.
317 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
318 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
319 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
320 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
322 #define lov_user_ost_data lov_user_ost_data_v1
323 struct lov_user_ost_data_v1 { /* per-stripe data structure */
324 struct ost_id l_ost_oi; /* OST object ID */
325 __u32 l_ost_gen; /* generation of this OST index */
326 __u32 l_ost_idx; /* OST index in LOV */
327 } __attribute__((packed));
329 #define lov_user_md lov_user_md_v1
330 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
331 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
332 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
333 struct ost_id lmm_oi; /* LOV object ID */
334 __u32 lmm_stripe_size; /* size of stripe in bytes */
335 __u16 lmm_stripe_count; /* num stripes in use for this object */
337 __u16 lmm_stripe_offset; /* starting stripe offset in
338 * lmm_objects, use when writing */
339 __u16 lmm_layout_gen; /* layout generation number
340 * used when reading */
342 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
343 } __attribute__((packed, __may_alias__));
345 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
346 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
347 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
348 struct ost_id lmm_oi; /* LOV object ID */
349 __u32 lmm_stripe_size; /* size of stripe in bytes */
350 __u16 lmm_stripe_count; /* num stripes in use for this object */
352 __u16 lmm_stripe_offset; /* starting stripe offset in
353 * lmm_objects, use when writing */
354 __u16 lmm_layout_gen; /* layout generation number
355 * used when reading */
357 char lmm_pool_name[LOV_MAXPOOLNAME]; /* pool name */
358 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
359 } __attribute__((packed));
361 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
363 if (lmm_magic == LOV_USER_MAGIC_V3)
364 return sizeof(struct lov_user_md_v3) +
365 stripes * sizeof(struct lov_user_ost_data_v1);
367 return sizeof(struct lov_user_md_v1) +
368 stripes * sizeof(struct lov_user_ost_data_v1);
371 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
372 * use this. It is unsafe to #define those values in this header as it
373 * is possible the application has already #included <sys/stat.h>. */
374 #ifdef HAVE_LOV_USER_MDS_DATA
375 #define lov_user_mds_data lov_user_mds_data_v1
376 struct lov_user_mds_data_v1 {
377 lstat_t lmd_st; /* MDS stat struct */
378 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
379 } __attribute__((packed));
381 struct lov_user_mds_data_v3 {
382 lstat_t lmd_st; /* MDS stat struct */
383 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
384 } __attribute__((packed));
387 /* keep this to be the same size as lov_user_ost_data_v1 */
388 struct lmv_user_mds_data {
389 struct lu_fid lum_fid;
397 LMV_DEFAULT_TYPE = 1,
400 #define lmv_user_md lmv_user_md_v1
401 struct lmv_user_md_v1 {
402 __u32 lum_magic; /* must be the first field */
403 __u32 lum_stripe_count; /* dirstripe count */
404 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
405 __u32 lum_hash_type; /* Dir stripe policy */
406 __u32 lum_type; /* LMV type: default or normal */
410 char lum_pool_name[LOV_MAXPOOLNAME];
411 struct lmv_user_mds_data lum_objects[0];
414 static inline int lmv_user_md_size(int stripes, int lmm_magic)
416 return sizeof(struct lmv_user_md) +
417 stripes * sizeof(struct lmv_user_mds_data);
420 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
422 struct ll_recreate_obj {
428 __u64 id; /* holds object id */
429 __u32 generation; /* holds object generation */
430 __u32 f_type; /* holds object type or stripe idx when passing it to
431 * OST for saving into EA. */
439 static inline int obd_uuid_equals(const struct obd_uuid *u1,
440 const struct obd_uuid *u2)
442 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
445 static inline int obd_uuid_empty(struct obd_uuid *uuid)
447 return uuid->uuid[0] == '\0';
450 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
452 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
453 uuid->uuid[sizeof(*uuid) - 1] = '\0';
456 /* For printf's only, make sure uuid is terminated */
457 static inline char *obd_uuid2str(struct obd_uuid *uuid)
459 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
460 /* Obviously not safe, but for printfs, no real harm done...
461 we're always null-terminated, even in a race. */
462 static char temp[sizeof(*uuid)];
463 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
464 temp[sizeof(*uuid) - 1] = '\0';
467 return (char *)(uuid->uuid);
470 /* Extract fsname from uuid (or target name) of a target
471 e.g. (myfs-OST0007_UUID -> myfs)
472 see also deuuidify. */
473 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
477 strncpy(buf, uuid, buflen - 1);
478 buf[buflen - 1] = '\0';
479 p = strrchr(buf, '-');
484 /* printf display format
485 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
486 #define FID_NOBRACE_LEN 40
487 #define FID_LEN (FID_NOBRACE_LEN + 2)
488 #define DFID_NOBRACE LPX64":0x%x:0x%x"
489 #define DFID "["DFID_NOBRACE"]"
495 /* scanf input parse format -- strip '[' first.
496 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
497 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
498 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
499 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
501 #define SFID "0x"LPX64i":0x%x:0x%x"
508 /********* Quotas **********/
510 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
511 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
512 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
513 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
514 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
515 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
516 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
517 /* lustre-specific control commands */
518 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
519 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
521 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
523 struct if_quotacheck {
525 struct obd_uuid obd_uuid;
528 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
531 #define N_PERMS_MAX 64
533 struct perm_downcall_data {
539 struct identity_downcall_data {
546 struct perm_downcall_data idd_perms[N_PERMS_MAX];
550 /* for non-mapped uid/gid */
551 #define NOBODY_UID 99
552 #define NOBODY_GID 99
554 #define INVALID_ID (-1)
563 #ifdef NEED_QUOTA_DEFS
565 #define QIF_BLIMITS 1
567 #define QIF_ILIMITS 4
571 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
572 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
573 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
574 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
577 #endif /* !__KERNEL__ */
579 /* lustre volatile file support
580 * file name header: .^L^S^T^R:volatile"
582 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
583 #define LUSTRE_VOLATILE_HDR_LEN 14
584 /* hdr + MDT index */
585 #define LUSTRE_VOLATILE_IDX LUSTRE_VOLATILE_HDR":%.4X:"
587 typedef enum lustre_quota_version {
589 } lustre_quota_version_t;
591 /* XXX: same as if_dqinfo struct in kernel */
599 /* XXX: same as if_dqblk struct in kernel, plus one padding */
601 __u64 dqb_bhardlimit;
602 __u64 dqb_bsoftlimit;
604 __u64 dqb_ihardlimit;
605 __u64 dqb_isoftlimit;
627 struct obd_dqinfo qc_dqinfo;
628 struct obd_dqblk qc_dqblk;
630 struct obd_uuid obd_uuid;
633 /* swap layout flags */
634 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
635 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
636 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
637 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
638 struct lustre_swap_layouts {
647 /********* Changelogs **********/
648 /** Changelog record types */
649 enum changelog_rec_type {
651 CL_CREATE = 1, /* namespace */
652 CL_MKDIR = 2, /* namespace */
653 CL_HARDLINK = 3, /* namespace */
654 CL_SOFTLINK = 4, /* namespace */
655 CL_MKNOD = 5, /* namespace */
656 CL_UNLINK = 6, /* namespace */
657 CL_RMDIR = 7, /* namespace */
658 CL_RENAME = 8, /* namespace */
659 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
660 CL_OPEN = 10, /* not currently used */
661 CL_CLOSE = 11, /* may be written to log only with mtime change */
662 CL_LAYOUT = 12, /* file layout/striping modified */
666 CL_HSM = 16, /* HSM specific events, see flags */
667 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
673 static inline const char *changelog_type2str(int type) {
674 static const char *changelog_str[] = {
675 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
676 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
677 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
680 if (type >= 0 && type < CL_LAST)
681 return changelog_str[type];
685 /* per-record flags */
686 #define CLF_VERSION 0x1000
687 #define CLF_EXT_VERSION 0x2000
688 #define CLF_FLAGSHIFT 12
689 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
690 #define CLF_VERMASK (~CLF_FLAGMASK)
691 /* Anything under the flagmask may be per-type (if desired) */
692 /* Flags for unlink */
693 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
694 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
695 /* HSM cleaning needed */
696 /* Flags for rename */
697 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink of target */
700 /* 12b used (from high weight to low weight):
705 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
706 #define CLF_HSM_ERR_H 6
707 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
708 #define CLF_HSM_EVENT_H 9
709 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
710 #define CLF_HSM_FLAG_H 11
711 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
712 #define CLF_HSM_SPARE_H 15
713 #define CLF_HSM_LAST 15
715 /* Remove bits higher than _h, then extract the value
716 * between _h and _l by shifting lower weigth to bit 0. */
717 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
718 >> (CLF_HSM_LAST - _h + _l))
720 #define CLF_HSM_SUCCESS 0x00
721 #define CLF_HSM_MAXERROR 0x7E
722 #define CLF_HSM_ERROVERFLOW 0x7F
724 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
726 /* 3 bits field => 8 values allowed */
738 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
740 return CLF_GET_BITS(flags, CLF_HSM_EVENT_H, CLF_HSM_EVENT_L);
743 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
745 *flags |= (he << CLF_HSM_EVENT_L);
748 static inline __u16 hsm_get_cl_flags(int flags)
750 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
753 static inline void hsm_set_cl_flags(int *flags, int bits)
755 *flags |= (bits << CLF_HSM_FLAG_L);
758 static inline int hsm_get_cl_error(int flags)
760 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
763 static inline void hsm_set_cl_error(int *flags, int error)
765 *flags |= (error << CLF_HSM_ERR_L);
768 #define CR_MAXSIZE cfs_size_round(2*NAME_MAX + 1 + sizeof(struct changelog_rec))
770 struct changelog_rec {
772 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
773 __u32 cr_type; /**< \a changelog_rec_type */
774 __u64 cr_index; /**< changelog record number */
775 __u64 cr_prev; /**< last index for this target fid */
778 lustre_fid cr_tfid; /**< target fid */
779 __u32 cr_markerflags; /**< CL_MARK flags */
781 lustre_fid cr_pfid; /**< parent fid */
782 char cr_name[0]; /**< last element */
783 } __attribute__((packed));
785 /* changelog_ext_rec is 2*sizeof(lu_fid) bigger than changelog_rec, to save
786 * space, only rename uses changelog_ext_rec, while others use changelog_rec to
789 struct changelog_ext_rec {
791 __u16 cr_flags; /**< (flags & CLF_FLAGMASK) |
793 __u32 cr_type; /**< \a changelog_rec_type */
794 __u64 cr_index; /**< changelog record number */
795 __u64 cr_prev; /**< last index for this target fid */
798 lustre_fid cr_tfid; /**< target fid */
799 __u32 cr_markerflags; /**< CL_MARK flags */
801 lustre_fid cr_pfid; /**< target parent fid */
802 lustre_fid cr_sfid; /**< source fid, or zero */
803 lustre_fid cr_spfid; /**< source parent fid, or zero */
804 char cr_name[0]; /**< last element */
805 } __attribute__((packed));
807 #define CHANGELOG_REC_EXTENDED(rec) \
808 (((rec)->cr_flags & CLF_VERMASK) == CLF_EXT_VERSION)
810 static inline int changelog_rec_size(struct changelog_rec *rec)
812 return CHANGELOG_REC_EXTENDED(rec) ? sizeof(struct changelog_ext_rec):
816 static inline char *changelog_rec_name(struct changelog_rec *rec)
818 return CHANGELOG_REC_EXTENDED(rec) ?
819 ((struct changelog_ext_rec *)rec)->cr_name: rec->cr_name;
822 static inline int changelog_rec_snamelen(struct changelog_ext_rec *rec)
824 return rec->cr_namelen - strlen(rec->cr_name) - 1;
827 static inline char *changelog_rec_sname(struct changelog_ext_rec *rec)
829 return rec->cr_name + strlen(rec->cr_name) + 1;
832 struct ioc_changelog {
839 enum changelog_message_type {
840 CL_RECORD = 10, /* message is a changelog_rec */
841 CL_EOF = 11, /* at end of current changelog */
844 /********* Misc **********/
846 struct ioc_data_version {
848 __u64 idv_flags; /* See LL_DV_xxx */
850 #define LL_DV_NOFLUSH 0x01 /* Do not take READ EXTENT LOCK before sampling
851 version. Dirty caches are left unchanged. */
854 # define offsetof(typ,memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
857 #define dot_lustre_name ".lustre"
860 /********* HSM **********/
862 /** HSM per-file state
863 * See HSM_FLAGS below.
866 HS_EXISTS = 0x00000001,
867 HS_DIRTY = 0x00000002,
868 HS_RELEASED = 0x00000004,
869 HS_ARCHIVED = 0x00000008,
870 HS_NORELEASE = 0x00000010,
871 HS_NOARCHIVE = 0x00000020,
872 HS_LOST = 0x00000040,
875 /* HSM user-setable flags. */
876 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
878 /* Other HSM flags. */
879 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
882 * All HSM-related possible flags that could be applied to a file.
883 * This should be kept in sync with hsm_states.
885 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
888 * HSMÂ request progress state
890 enum hsm_progress_states {
897 static inline char *hsm_progress_state2name(enum hsm_progress_states s)
900 case HPS_WAITING: return "waiting";
901 case HPS_RUNNING: return "running";
902 case HPS_DONE: return "done";
903 default: return "unknown";
910 } __attribute__((packed));
913 * Current HSM states of a Lustre file.
915 * This structure purpose is to be sent to user-space mainly. It describes the
916 * current HSM flags and in-progress action.
918 struct hsm_user_state {
919 /** Current HSM states, from enum hsm_states. */
921 __u32 hus_archive_id;
922 /** The current undergoing action, if there is one */
923 __u32 hus_in_progress_state;
924 __u32 hus_in_progress_action;
925 struct hsm_extent hus_in_progress_location;
926 char hus_extended_info[];
929 struct hsm_state_set_ioc {
930 struct lu_fid hssi_fid;
932 __u64 hssi_clearmask;
936 * This structure describes the current in-progress action for a file.
937 * it is retuned to user space and send over the wire
939 struct hsm_current_action {
940 /** The current undergoing action, if there is one */
941 /* state is one of hsm_progress_states */
943 /* action is one of hsm_user_action */
945 struct hsm_extent hca_location;
948 /***** HSM user requests ******/
949 /* User-generated (lfs/ioctl) request types */
950 enum hsm_user_action {
951 HUA_NONE = 1, /* no action (noop) */
952 HUA_ARCHIVE = 10, /* copy to hsm */
953 HUA_RESTORE = 11, /* prestage */
954 HUA_RELEASE = 12, /* drop ost objects */
955 HUA_REMOVE = 13, /* remove from archive */
956 HUA_CANCEL = 14 /* cancel a request */
959 static inline char *hsm_user_action2name(enum hsm_user_action a)
962 case HUA_NONE: return "NOOP";
963 case HUA_ARCHIVE: return "ARCHIVE";
964 case HUA_RESTORE: return "RESTORE";
965 case HUA_RELEASE: return "RELEASE";
966 case HUA_REMOVE: return "REMOVE";
967 case HUA_CANCEL: return "CANCEL";
968 default: return "UNKNOWN";
973 * List of hr_flags (bit field)
975 #define HSM_FORCE_ACTION 0x0001
976 /* used by CT, connot be set by user */
977 #define HSM_GHOST_COPY 0x0002
980 * Contains all the fixed part of struct hsm_user_request.
984 __u32 hr_action; /* enum hsm_user_action */
985 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
986 __u64 hr_flags; /* request flags */
987 __u32 hr_itemcount; /* item count in hur_user_item vector */
991 struct hsm_user_item {
993 struct hsm_extent hui_extent;
994 } __attribute__((packed));
996 struct hsm_user_request {
997 struct hsm_request hur_request;
998 struct hsm_user_item hur_user_item[0];
999 /* extra data blob at end of struct (after all
1000 * hur_user_items), only use helpers to access it
1002 } __attribute__((packed));
1004 /** Return pointer to data field in a hsm user request */
1005 static inline void *hur_data(struct hsm_user_request *hur)
1007 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1010 /** Compute the current length of the provided hsm_user_request. */
1011 static inline int hur_len(struct hsm_user_request *hur)
1013 return offsetof(struct hsm_user_request,
1014 hur_user_item[hur->hur_request.hr_itemcount]) +
1015 hur->hur_request.hr_data_len;
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 */
1055 } __attribute__((packed));
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++)
1078 cnt = snprintf(ptr, sz, "%.2X",
1079 (unsigned char)hai->hai_data[i]);
1087 /* Copytool action list */
1088 #define HAL_VERSION 1
1089 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1090 struct hsm_action_list {
1092 __u32 hal_count; /* number of hai's to follow */
1093 __u64 hal_compound_id; /* returned by coordinator */
1095 __u32 hal_archive_id; /* which archive backend */
1097 char hal_fsname[0]; /* null-terminated */
1098 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1099 boundaries. See hai_zero */
1100 } __attribute__((packed));
1102 #ifndef HAVE_CFS_SIZE_ROUND
1103 static inline int cfs_size_round (int val)
1105 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-> \
1117 /* Return pointer to next hai */
1118 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1120 return (struct hsm_action_item *)((char *)hai +
1121 cfs_size_round(hai->hai_len));
1124 /* Return size of an hsm_action_list */
1125 static inline int hal_size(struct hsm_action_list *hal)
1128 struct hsm_action_item *hai;
1130 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname));
1131 hai = hai_zero(hal);
1132 for (i = 0 ; i < hal->hal_count ; i++) {
1133 sz += cfs_size_round(hai->hai_len);
1134 hai = hai_next(hai);
1139 /* Copytool progress reporting */
1140 #define HP_FLAG_COMPLETED 0x01
1141 #define HP_FLAG_RETRY 0x02
1143 struct hsm_progress {
1146 struct hsm_extent hp_extent;
1148 __u16 hp_errval; /* positive val */
1153 * Use by copytool during any hsm request they handled.
1154 * This structure is initialized by llapi_hsm_copy_start()
1155 * which is an helper over the ioctl() interface
1156 * Store Lustre, internal use only, data.
1159 __u64 hc_data_version;
1161 __u16 hc_errval; /* positive val */
1163 struct hsm_action_item hc_hai;
1166 /** @} lustreuser */
1168 #endif /* _LUSTRE_USER_H */