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>
54 #include <lustre/ll_fiemap.h>
55 #if defined(__linux__)
56 #include <linux/lustre_user.h>
57 #elif defined(__APPLE__)
58 #include <darwin/lustre_user.h>
59 #elif defined(__WINNT__)
60 #include <winnt/lustre_user.h>
62 #error Unsupported operating system.
66 #define LUSTRE_ANONYMOUS_UNION_NAME u
68 #define LUSTRE_ANONYMOUS_UNION_NAME
72 #define LL_SUPER_MAGIC 0x0BD00BD0
74 #ifndef FSFILT_IOC_GETFLAGS
75 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
76 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
77 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
78 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
79 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
80 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
81 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
84 /* FIEMAP flags supported by Lustre */
85 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
87 enum obd_statfs_state {
88 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
89 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
90 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
91 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
92 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
106 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
107 __u32 os_fprecreated; /* objs available now to the caller */
108 /* used in QoS code to find preferred
123 * FID is a cluster-wide unique identifier of a file or an object (stripe).
124 * FIDs are never reused.
128 * FID sequence. Sequence is a unit of migration: all files (objects)
129 * with FIDs from a given sequence are stored on the same server.
130 * Lustre should support 2^64 objects, so even if each sequence
131 * has only a single object we can still enumerate 2^64 objects.
134 /* FID number within sequence. */
137 * FID version, used to distinguish different versions (in the sense
138 * of snapshots, etc.) of the same file system object. Not currently
145 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
148 /* keep this one for compatibility */
149 struct filter_fid_old {
150 struct lu_fid ff_parent;
155 /* Userspace should treat lu_fid as opaque, and only use the following methods
156 * to print or parse them. Other functions (e.g. compare, swab) could be moved
157 * here from lustre_idl.h if needed. */
158 typedef struct lu_fid lustre_fid;
161 * Following struct for object attributes, that will be kept inode's EA.
162 * Introduced in 2.0 release (please see b15993, for details)
163 * Added to all objects since Lustre 2.4 as contains self FID
165 struct lustre_mdt_attrs {
167 * Bitfield for supported data in this structure. From enum lma_compat.
168 * lma_self_fid and lma_flags are always available.
172 * Per-file incompat feature list. Lustre version should support all
173 * flags set in this field. The supported feature mask is available in
177 /** FID of this inode */
178 struct lu_fid lma_self_fid;
182 * Prior to 2.4, the LMA structure also included SOM attributes which has since
183 * been moved to a dedicated xattr
184 * lma_flags was also removed because of lma_compat/incompat fields.
186 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
189 * OST object IDentifier.
197 struct lu_fid oi_fid;
198 } LUSTRE_ANONYMOUS_UNION_NAME;
201 #define DOSTID LPX64":"LPU64
202 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
205 * The ioctl naming rules:
206 * LL_* - works on the currently opened filehandle instead of parent dir
207 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
208 * *_MDC_* - gets/sets data related to MDC
209 * *_LOV_* - gets/sets data related to OSC/LOV
210 * *FILE* - called on parent dir and passes in a filename
211 * *STRIPE* - set/get lov_user_md
212 * *INFO - set/get lov_user_mds_data
214 /* see <lustre_lib.h> for ioctl numberss 101-150 */
215 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
216 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
217 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
218 /* LL_IOC_LOV_SETSTRIPE: See also OBD_IOC_LOV_SETSTRIPE */
219 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
220 /* LL_IOC_LOV_GETSTRIPE: See also OBD_IOC_LOV_GETSTRIPE */
221 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
222 /* LL_IOC_LOV_SETEA: See also OBD_IOC_LOV_SETEA */
223 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
224 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
225 #define LL_IOC_RECREATE_FID _IOW ('f', 157, struct lu_fid)
226 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
227 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
228 /* LL_IOC_QUOTACHECK: See also OBD_IOC_QUOTACHECK */
229 #define LL_IOC_QUOTACHECK _IOW ('f', 160, int)
230 /* LL_IOC_POLL_QUOTACHECK: See also OBD_IOC_POLL_QUOTACHECK */
231 #define LL_IOC_POLL_QUOTACHECK _IOR ('f', 161, struct if_quotacheck *)
232 /* LL_IOC_QUOTACTL: See also OBD_IOC_QUOTACTL */
233 #define LL_IOC_QUOTACTL _IOWR('f', 162, struct if_quotactl)
234 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
235 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
236 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
237 #define LL_IOC_RMTACL _IOW ('f', 167, long)
238 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
239 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
240 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
241 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
242 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
243 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
244 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
245 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
247 /* see <lustre_lib.h> for ioctl numbers 177-210 */
249 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
250 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
251 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
252 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
253 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
254 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
255 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
256 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
257 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
258 struct lustre_swap_layouts)
259 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
260 struct hsm_current_action)
261 /* see <lustre_lib.h> for ioctl numbers 221-232 */
263 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
264 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
265 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
266 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
267 #define LL_IOC_GET_LEASE _IO('f', 244)
268 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
270 #define LL_STATFS_LMV 1
271 #define LL_STATFS_LOV 2
272 #define LL_STATFS_NODELAY 4
274 #define IOC_MDC_TYPE 'i'
275 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
276 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
277 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
278 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
280 /* Keep these for backward compartability. */
281 #define LL_IOC_OBD_STATFS IOC_OBD_STATFS
282 #define IOC_MDC_GETSTRIPE IOC_MDC_GETFILESTRIPE
285 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
287 /* Hopefully O_LOV_DELAY_CREATE does not conflict with standard O_xxx flags.
288 * Previously it was defined as 0100000000 and conflicts with FMODE_NONOTIFY
289 * which was added since kernel 2.6.36, so we redefine it as 020000000.
290 * To be compatible with old version's statically linked binary, finally we
291 * define it as (020000000 | 0100000000).
293 #define O_LOV_DELAY_CREATE 0120000000
295 #define LL_FILE_IGNORE_LOCK 0x00000001
296 #define LL_FILE_GROUP_LOCKED 0x00000002
297 #define LL_FILE_READAHEA 0x00000004
298 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
299 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
300 #define LL_FILE_RMTACL 0x00000020
302 #define LOV_USER_MAGIC_V1 0x0BD10BD0
303 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
304 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
305 #define LOV_USER_MAGIC_V3 0x0BD30BD0
307 #define LMV_MAGIC_V1 0x0CD10CD0 /*normal stripe lmv magic */
308 #define LMV_USER_MAGIC 0x0CD20CD0 /*default lmv magic*/
310 #define LOV_PATTERN_RAID0 0x001
311 #define LOV_PATTERN_RAID1 0x002
312 #define LOV_PATTERN_FIRST 0x100
314 #define LOV_MAXPOOLNAME 16
315 #define LOV_POOLNAMEF "%.16s"
317 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
318 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
319 #define LOV_MAX_STRIPE_COUNT_OLD 160
320 /* This calculation is crafted so that input of 4096 will result in 160
321 * which in turn is equal to old maximal stripe count.
322 * XXX: In fact this is too simpified for now, what it also need is to get
323 * ea_type argument to clearly know how much space each stripe consumes.
325 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
326 * allocation that is sufficient for the current generation of systems.
328 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
329 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
330 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
331 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
333 #define XATTR_LUSTRE_PREFIX "lustre."
334 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
336 #define lov_user_ost_data lov_user_ost_data_v1
337 struct lov_user_ost_data_v1 { /* per-stripe data structure */
338 struct ost_id l_ost_oi; /* OST object ID */
339 __u32 l_ost_gen; /* generation of this OST index */
340 __u32 l_ost_idx; /* OST index in LOV */
341 } __attribute__((packed));
343 #define lov_user_md lov_user_md_v1
344 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
345 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
346 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
347 struct ost_id lmm_oi; /* LOV object ID */
348 __u32 lmm_stripe_size; /* size of stripe in bytes */
349 __u16 lmm_stripe_count; /* num stripes in use for this object */
351 __u16 lmm_stripe_offset; /* starting stripe offset in
352 * lmm_objects, use when writing */
353 __u16 lmm_layout_gen; /* layout generation number
354 * used when reading */
356 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
357 } __attribute__((packed, __may_alias__));
359 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
360 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
361 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
362 struct ost_id lmm_oi; /* LOV object ID */
363 __u32 lmm_stripe_size; /* size of stripe in bytes */
364 __u16 lmm_stripe_count; /* num stripes in use for this object */
366 __u16 lmm_stripe_offset; /* starting stripe offset in
367 * lmm_objects, use when writing */
368 __u16 lmm_layout_gen; /* layout generation number
369 * used when reading */
371 char lmm_pool_name[LOV_MAXPOOLNAME]; /* pool name */
372 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
373 } __attribute__((packed));
375 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
377 if (lmm_magic == LOV_USER_MAGIC_V3)
378 return sizeof(struct lov_user_md_v3) +
379 stripes * sizeof(struct lov_user_ost_data_v1);
381 return sizeof(struct lov_user_md_v1) +
382 stripes * sizeof(struct lov_user_ost_data_v1);
385 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
386 * use this. It is unsafe to #define those values in this header as it
387 * is possible the application has already #included <sys/stat.h>. */
388 #ifdef HAVE_LOV_USER_MDS_DATA
389 #define lov_user_mds_data lov_user_mds_data_v1
390 struct lov_user_mds_data_v1 {
391 lstat_t lmd_st; /* MDS stat struct */
392 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
393 } __attribute__((packed));
395 struct lov_user_mds_data_v3 {
396 lstat_t lmd_st; /* MDS stat struct */
397 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
398 } __attribute__((packed));
401 /* keep this to be the same size as lov_user_ost_data_v1 */
402 struct lmv_user_mds_data {
403 struct lu_fid lum_fid;
411 LMV_DEFAULT_TYPE = 1,
414 #define lmv_user_md lmv_user_md_v1
415 struct lmv_user_md_v1 {
416 __u32 lum_magic; /* must be the first field */
417 __u32 lum_stripe_count; /* dirstripe count */
418 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
419 __u32 lum_hash_type; /* Dir stripe policy */
420 __u32 lum_type; /* LMV type: default or normal */
424 char lum_pool_name[LOV_MAXPOOLNAME];
425 struct lmv_user_mds_data lum_objects[0];
428 static inline int lmv_user_md_size(int stripes, int lmm_magic)
430 return sizeof(struct lmv_user_md) +
431 stripes * sizeof(struct lmv_user_mds_data);
434 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
436 struct ll_recreate_obj {
442 __u64 id; /* holds object id */
443 __u32 generation; /* holds object generation */
444 __u32 f_type; /* holds object type or stripe idx when passing it to
445 * OST for saving into EA. */
453 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
454 const struct obd_uuid *u2)
456 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
459 static inline int obd_uuid_empty(struct obd_uuid *uuid)
461 return uuid->uuid[0] == '\0';
464 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
466 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
467 uuid->uuid[sizeof(*uuid) - 1] = '\0';
470 /* For printf's only, make sure uuid is terminated */
471 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
476 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
477 /* Obviously not safe, but for printfs, no real harm done...
478 we're always null-terminated, even in a race. */
479 static char temp[sizeof(*uuid)];
480 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
481 temp[sizeof(*uuid) - 1] = '\0';
484 return (char *)(uuid->uuid);
487 /* Extract fsname from uuid (or target name) of a target
488 e.g. (myfs-OST0007_UUID -> myfs)
489 see also deuuidify. */
490 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
494 strncpy(buf, uuid, buflen - 1);
495 buf[buflen - 1] = '\0';
496 p = strrchr(buf, '-');
501 /* printf display format
502 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
503 #define FID_NOBRACE_LEN 40
504 #define FID_LEN (FID_NOBRACE_LEN + 2)
505 #define DFID_NOBRACE LPX64":0x%x:0x%x"
506 #define DFID "["DFID_NOBRACE"]"
512 /* scanf input parse format -- strip '[' first.
513 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
514 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
515 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
516 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
518 #define SFID "0x"LPX64i":0x%x:0x%x"
525 /********* Quotas **********/
527 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
528 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
529 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
530 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
531 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
532 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
533 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
534 /* lustre-specific control commands */
535 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
536 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
538 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
540 struct if_quotacheck {
542 struct obd_uuid obd_uuid;
545 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
548 #define N_PERMS_MAX 64
550 struct perm_downcall_data {
556 struct identity_downcall_data {
563 struct perm_downcall_data idd_perms[N_PERMS_MAX];
567 /* for non-mapped uid/gid */
568 #define NOBODY_UID 99
569 #define NOBODY_GID 99
571 #define INVALID_ID (-1)
580 #ifdef NEED_QUOTA_DEFS
582 #define QIF_BLIMITS 1
584 #define QIF_ILIMITS 4
588 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
589 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
590 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
591 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
594 #endif /* !__KERNEL__ */
596 /* lustre volatile file support
597 * file name header: .^L^S^T^R:volatile"
599 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
600 #define LUSTRE_VOLATILE_HDR_LEN 14
601 /* hdr + MDT index */
602 #define LUSTRE_VOLATILE_IDX LUSTRE_VOLATILE_HDR":%.4X:"
604 typedef enum lustre_quota_version {
606 } lustre_quota_version_t;
608 /* XXX: same as if_dqinfo struct in kernel */
616 /* XXX: same as if_dqblk struct in kernel, plus one padding */
618 __u64 dqb_bhardlimit;
619 __u64 dqb_bsoftlimit;
621 __u64 dqb_ihardlimit;
622 __u64 dqb_isoftlimit;
644 struct obd_dqinfo qc_dqinfo;
645 struct obd_dqblk qc_dqblk;
647 struct obd_uuid obd_uuid;
650 /* swap layout flags */
651 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
652 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
653 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
654 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
656 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
657 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
658 struct lustre_swap_layouts {
667 /********* Changelogs **********/
668 /** Changelog record types */
669 enum changelog_rec_type {
671 CL_CREATE = 1, /* namespace */
672 CL_MKDIR = 2, /* namespace */
673 CL_HARDLINK = 3, /* namespace */
674 CL_SOFTLINK = 4, /* namespace */
675 CL_MKNOD = 5, /* namespace */
676 CL_UNLINK = 6, /* namespace */
677 CL_RMDIR = 7, /* namespace */
678 CL_RENAME = 8, /* namespace */
679 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
680 CL_OPEN = 10, /* not currently used */
681 CL_CLOSE = 11, /* may be written to log only with mtime change */
682 CL_LAYOUT = 12, /* file layout/striping modified */
686 CL_HSM = 16, /* HSM specific events, see flags */
687 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
693 static inline const char *changelog_type2str(int type) {
694 static const char *changelog_str[] = {
695 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
696 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
697 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
700 if (type >= 0 && type < CL_LAST)
701 return changelog_str[type];
705 /* per-record flags */
706 #define CLF_VERSION 0x1000
707 #define CLF_EXT_VERSION 0x2000
708 #define CLF_FLAGSHIFT 12
709 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
710 #define CLF_VERMASK (~CLF_FLAGMASK)
711 /* Anything under the flagmask may be per-type (if desired) */
712 /* Flags for unlink */
713 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
714 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
715 /* HSM cleaning needed */
716 /* Flags for rename */
717 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
719 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
720 * has an archive in backend */
723 /* 12b used (from high weight to low weight):
728 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
729 #define CLF_HSM_ERR_H 6
730 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
731 #define CLF_HSM_EVENT_H 9
732 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
733 #define CLF_HSM_FLAG_H 11
734 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
735 #define CLF_HSM_SPARE_H 15
736 #define CLF_HSM_LAST 15
738 /* Remove bits higher than _h, then extract the value
739 * between _h and _l by shifting lower weigth to bit 0. */
740 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
741 >> (CLF_HSM_LAST - _h + _l))
743 #define CLF_HSM_SUCCESS 0x00
744 #define CLF_HSM_MAXERROR 0x7E
745 #define CLF_HSM_ERROVERFLOW 0x7F
747 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
749 /* 3 bits field => 8 values allowed */
761 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
763 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
767 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
769 *flags |= (he << CLF_HSM_EVENT_L);
772 static inline __u16 hsm_get_cl_flags(int flags)
774 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
777 static inline void hsm_set_cl_flags(int *flags, int bits)
779 *flags |= (bits << CLF_HSM_FLAG_L);
782 static inline int hsm_get_cl_error(int flags)
784 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
787 static inline void hsm_set_cl_error(int *flags, int error)
789 *flags |= (error << CLF_HSM_ERR_L);
792 #define CR_MAXSIZE cfs_size_round(2*NAME_MAX + 1 + \
793 sizeof(struct changelog_ext_rec))
795 struct changelog_rec {
797 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
798 __u32 cr_type; /**< \a changelog_rec_type */
799 __u64 cr_index; /**< changelog record number */
800 __u64 cr_prev; /**< last index for this target fid */
803 lustre_fid cr_tfid; /**< target fid */
804 __u32 cr_markerflags; /**< CL_MARK flags */
806 lustre_fid cr_pfid; /**< parent fid */
807 char cr_name[0]; /**< last element */
808 } __attribute__((packed));
810 /* changelog_ext_rec is 2*sizeof(lu_fid) bigger than changelog_rec, to save
811 * space, only rename uses changelog_ext_rec, while others use changelog_rec to
814 struct changelog_ext_rec {
816 __u16 cr_flags; /**< (flags & CLF_FLAGMASK) |
818 __u32 cr_type; /**< \a changelog_rec_type */
819 __u64 cr_index; /**< changelog record number */
820 __u64 cr_prev; /**< last index for this target fid */
823 lustre_fid cr_tfid; /**< target fid */
824 __u32 cr_markerflags; /**< CL_MARK flags */
826 lustre_fid cr_pfid; /**< target parent fid */
827 lustre_fid cr_sfid; /**< source fid, or zero */
828 lustre_fid cr_spfid; /**< source parent fid, or zero */
829 char cr_name[0]; /**< last element */
830 } __attribute__((packed));
832 #define CHANGELOG_REC_EXTENDED(rec) \
833 (((rec)->cr_flags & CLF_VERMASK) == CLF_EXT_VERSION)
835 static inline int changelog_rec_size(struct changelog_rec *rec)
837 return CHANGELOG_REC_EXTENDED(rec) ? sizeof(struct changelog_ext_rec):
841 static inline char *changelog_rec_name(struct changelog_rec *rec)
843 return CHANGELOG_REC_EXTENDED(rec) ?
844 ((struct changelog_ext_rec *)rec)->cr_name: rec->cr_name;
847 static inline int changelog_rec_snamelen(struct changelog_ext_rec *rec)
849 return rec->cr_namelen - strlen(rec->cr_name) - 1;
852 static inline char *changelog_rec_sname(struct changelog_ext_rec *rec)
854 return rec->cr_name + strlen(rec->cr_name) + 1;
857 struct ioc_changelog {
864 enum changelog_message_type {
865 CL_RECORD = 10, /* message is a changelog_rec */
866 CL_EOF = 11, /* at end of current changelog */
869 /********* Misc **********/
871 struct ioc_data_version {
873 __u64 idv_flags; /* See LL_DV_xxx */
875 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
876 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
879 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
882 #define dot_lustre_name ".lustre"
885 /********* HSM **********/
887 /** HSM per-file state
888 * See HSM_FLAGS below.
891 HS_EXISTS = 0x00000001,
892 HS_DIRTY = 0x00000002,
893 HS_RELEASED = 0x00000004,
894 HS_ARCHIVED = 0x00000008,
895 HS_NORELEASE = 0x00000010,
896 HS_NOARCHIVE = 0x00000020,
897 HS_LOST = 0x00000040,
900 /* HSM user-setable flags. */
901 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
903 /* Other HSM flags. */
904 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
907 * All HSM-related possible flags that could be applied to a file.
908 * This should be kept in sync with hsm_states.
910 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
913 * HSMÂ request progress state
915 enum hsm_progress_states {
922 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
925 case HPS_WAITING: return "waiting";
926 case HPS_RUNNING: return "running";
927 case HPS_DONE: return "done";
928 default: return "unknown";
935 } __attribute__((packed));
938 * Current HSM states of a Lustre file.
940 * This structure purpose is to be sent to user-space mainly. It describes the
941 * current HSM flags and in-progress action.
943 struct hsm_user_state {
944 /** Current HSM states, from enum hsm_states. */
946 __u32 hus_archive_id;
947 /** The current undergoing action, if there is one */
948 __u32 hus_in_progress_state;
949 __u32 hus_in_progress_action;
950 struct hsm_extent hus_in_progress_location;
951 char hus_extended_info[];
954 struct hsm_state_set_ioc {
955 struct lu_fid hssi_fid;
957 __u64 hssi_clearmask;
961 * This structure describes the current in-progress action for a file.
962 * it is retuned to user space and send over the wire
964 struct hsm_current_action {
965 /** The current undergoing action, if there is one */
966 /* state is one of hsm_progress_states */
968 /* action is one of hsm_user_action */
970 struct hsm_extent hca_location;
973 /***** HSM user requests ******/
974 /* User-generated (lfs/ioctl) request types */
975 enum hsm_user_action {
976 HUA_NONE = 1, /* no action (noop) */
977 HUA_ARCHIVE = 10, /* copy to hsm */
978 HUA_RESTORE = 11, /* prestage */
979 HUA_RELEASE = 12, /* drop ost objects */
980 HUA_REMOVE = 13, /* remove from archive */
981 HUA_CANCEL = 14 /* cancel a request */
984 static inline const char *hsm_user_action2name(enum hsm_user_action a)
987 case HUA_NONE: return "NOOP";
988 case HUA_ARCHIVE: return "ARCHIVE";
989 case HUA_RESTORE: return "RESTORE";
990 case HUA_RELEASE: return "RELEASE";
991 case HUA_REMOVE: return "REMOVE";
992 case HUA_CANCEL: return "CANCEL";
993 default: return "UNKNOWN";
998 * List of hr_flags (bit field)
1000 #define HSM_FORCE_ACTION 0x0001
1001 /* used by CT, connot be set by user */
1002 #define HSM_GHOST_COPY 0x0002
1005 * Contains all the fixed part of struct hsm_user_request.
1008 struct hsm_request {
1009 __u32 hr_action; /* enum hsm_user_action */
1010 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1011 __u64 hr_flags; /* request flags */
1012 __u32 hr_itemcount; /* item count in hur_user_item vector */
1016 struct hsm_user_item {
1018 struct hsm_extent hui_extent;
1019 } __attribute__((packed));
1021 struct hsm_user_request {
1022 struct hsm_request hur_request;
1023 struct hsm_user_item hur_user_item[0];
1024 /* extra data blob at end of struct (after all
1025 * hur_user_items), only use helpers to access it
1027 } __attribute__((packed));
1029 /** Return pointer to data field in a hsm user request */
1030 static inline void *hur_data(struct hsm_user_request *hur)
1032 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1035 /** Compute the current length of the provided hsm_user_request. */
1036 static inline int hur_len(struct hsm_user_request *hur)
1038 return offsetof(struct hsm_user_request, hur_user_item[0]) +
1039 hur->hur_request.hr_itemcount * sizeof(hur->hur_user_item[0]) +
1040 hur->hur_request.hr_data_len;
1043 /****** HSM RPCs to copytool *****/
1044 /* Message types the copytool may receive */
1045 enum hsm_message_type {
1046 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1049 /* Actions the copytool may be instructed to take for a given action_item */
1050 enum hsm_copytool_action {
1051 HSMA_NONE = 10, /* no action */
1052 HSMA_ARCHIVE = 20, /* arbitrary offset */
1058 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1061 case HSMA_NONE: return "NOOP";
1062 case HSMA_ARCHIVE: return "ARCHIVE";
1063 case HSMA_RESTORE: return "RESTORE";
1064 case HSMA_REMOVE: return "REMOVE";
1065 case HSMA_CANCEL: return "CANCEL";
1066 default: return "UNKNOWN";
1070 /* Copytool item action description */
1071 struct hsm_action_item {
1072 __u32 hai_len; /* valid size of this struct */
1073 __u32 hai_action; /* hsm_copytool_action, but use known size */
1074 lustre_fid hai_fid; /* Lustre FID to operated on */
1075 lustre_fid hai_dfid; /* fid used for data access */
1076 struct hsm_extent hai_extent; /* byte range to operate on */
1077 __u64 hai_cookie; /* action cookie from coordinator */
1078 __u64 hai_gid; /* grouplock id */
1079 char hai_data[0]; /* variable length */
1080 } __attribute__((packed));
1083 * helper function which print in hexa the first bytes of
1085 * \param hai [IN] record to print
1086 * \param buffer [OUT] output buffer
1087 * \param len [IN] max buffer len
1090 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1091 char *buffer, int len)
1093 int i, sz, data_len;
1098 data_len = hai->hai_len - sizeof(*hai);
1099 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1103 cnt = snprintf(ptr, sz, "%.2X",
1104 (unsigned char)hai->hai_data[i]);
1112 /* Copytool action list */
1113 #define HAL_VERSION 1
1114 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1115 struct hsm_action_list {
1117 __u32 hal_count; /* number of hai's to follow */
1118 __u64 hal_compound_id; /* returned by coordinator */
1120 __u32 hal_archive_id; /* which archive backend */
1122 char hal_fsname[0]; /* null-terminated */
1123 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1124 boundaries. See hai_zero */
1125 } __attribute__((packed));
1127 #ifndef HAVE_CFS_SIZE_ROUND
1128 static inline int cfs_size_round (int val)
1130 return (val + 7) & (~0x7);
1132 #define HAVE_CFS_SIZE_ROUND
1135 /* Return pointer to first hai in action list */
1136 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1138 return (struct hsm_action_item *)(hal->hal_fsname +
1139 cfs_size_round(strlen(hal-> \
1142 /* Return pointer to next hai */
1143 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1145 return (struct hsm_action_item *)((char *)hai +
1146 cfs_size_round(hai->hai_len));
1149 /* Return size of an hsm_action_list */
1150 static inline size_t hal_size(struct hsm_action_list *hal)
1154 struct hsm_action_item *hai;
1156 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname));
1157 hai = hai_first(hal);
1158 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1159 sz += cfs_size_round(hai->hai_len);
1165 * describe the attributes to be set on imported file
1167 struct hsm_user_import {
1176 __u32 hui_archive_id;
1179 /* Copytool progress reporting */
1180 #define HP_FLAG_COMPLETED 0x01
1181 #define HP_FLAG_RETRY 0x02
1183 struct hsm_progress {
1186 struct hsm_extent hp_extent;
1188 __u16 hp_errval; /* positive val */
1193 __u64 hc_data_version;
1195 __u16 hc_errval; /* positive val */
1197 struct hsm_action_item hc_hai;
1200 /** @} lustreuser */
1202 #endif /* _LUSTRE_USER_H */