4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2010, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_user.h
38 * Lustre public user-space interface definitions.
41 #ifndef _LUSTRE_USER_H
42 #define _LUSTRE_USER_H
44 /** \defgroup lustreuser lustreuser
51 #include <libcfs/posix/posix-types.h>
53 #include <lustre/ll_fiemap.h>
54 #if defined(__linux__)
55 #include <linux/lustre_user.h>
56 #elif defined(__APPLE__)
57 #include <darwin/lustre_user.h>
58 #elif defined(__WINNT__)
59 #include <winnt/lustre_user.h>
61 #error Unsupported operating system.
65 #define LUSTRE_ANONYMOUS_UNION_NAME u
67 #define LUSTRE_ANONYMOUS_UNION_NAME
71 #define LL_SUPER_MAGIC 0x0BD00BD0
73 #ifndef FSFILT_IOC_GETFLAGS
74 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
75 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
76 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
77 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
78 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
79 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
80 #define FSFILT_IOC_FIEMAP _IOWR('f', 11, struct ll_user_fiemap)
83 /* FIEMAP flags supported by Lustre */
84 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
86 enum obd_statfs_state {
87 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
88 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
89 OS_STATE_RDONLY_1 = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
90 OS_STATE_RDONLY_2 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
91 OS_STATE_RDONLY_3 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
105 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
106 __u32 os_fprecreated; /* objs available now to the caller */
107 /* used in QoS code to find preferred
122 * FID is a cluster-wide unique identifier of a file or an object (stripe).
123 * FIDs are never reused.
127 * FID sequence. Sequence is a unit of migration: all files (objects)
128 * with FIDs from a given sequence are stored on the same server.
129 * Lustre should support 2^64 objects, so even if each sequence
130 * has only a single object we can still enumerate 2^64 objects.
133 /* FID number within sequence. */
136 * FID version, used to distinguish different versions (in the sense
137 * of snapshots, etc.) of the same file system object. Not currently
143 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
144 * MDT-object's FID::f_ver, instead it is the OST-object index in its
145 * parent MDT-object's layout EA. */
146 #define f_stripe_idx f_ver
149 struct lu_fid ff_parent; /* ff_parent.f_ver == file stripe number */
152 /* keep this one for compatibility */
153 struct filter_fid_old {
154 struct lu_fid ff_parent;
159 /* Userspace should treat lu_fid as opaque, and only use the following methods
160 * to print or parse them. Other functions (e.g. compare, swab) could be moved
161 * here from lustre_idl.h if needed. */
162 typedef struct lu_fid lustre_fid;
165 * Following struct for object attributes, that will be kept inode's EA.
166 * Introduced in 2.0 release (please see b15993, for details)
167 * Added to all objects since Lustre 2.4 as contains self FID
169 struct lustre_mdt_attrs {
171 * Bitfield for supported data in this structure. From enum lma_compat.
172 * lma_self_fid and lma_flags are always available.
176 * Per-file incompat feature list. Lustre version should support all
177 * flags set in this field. The supported feature mask is available in
181 /** FID of this inode */
182 struct lu_fid lma_self_fid;
186 * Prior to 2.4, the LMA structure also included SOM attributes which has since
187 * been moved to a dedicated xattr
188 * lma_flags was also removed because of lma_compat/incompat fields.
190 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
193 * OST object IDentifier.
201 struct lu_fid oi_fid;
202 } LUSTRE_ANONYMOUS_UNION_NAME;
205 #define DOSTID LPX64":"LPU64
206 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
209 * The ioctl naming rules:
210 * LL_* - works on the currently opened filehandle instead of parent dir
211 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
212 * *_MDC_* - gets/sets data related to MDC
213 * *_LOV_* - gets/sets data related to OSC/LOV
214 * *FILE* - called on parent dir and passes in a filename
215 * *STRIPE* - set/get lov_user_md
216 * *INFO - set/get lov_user_mds_data
218 /* lustre_ioctl.h 101-150 */
219 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
220 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
221 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
222 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
223 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
224 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
225 #define LL_IOC_RECREATE_OBJ _IOW ('f', 157, long)
226 #define LL_IOC_RECREATE_FID _IOW ('f', 157, struct lu_fid)
227 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
228 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
229 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
230 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
231 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
232 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
233 #define IOC_LOV_GETINFO _IOWR('f', 165, struct lov_user_mds_data *)
234 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
235 #define LL_IOC_RMTACL _IOW ('f', 167, long)
236 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
237 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
238 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
239 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
240 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
241 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
242 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
243 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
244 /* lustre_ioctl.h 177-210 */
245 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
246 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
247 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
248 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
249 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
250 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
251 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
252 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
253 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
254 struct lustre_swap_layouts)
255 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
256 struct hsm_current_action)
257 /* lustre_ioctl.h 221-232 */
258 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
259 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
260 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
261 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
262 #define LL_IOC_GET_LEASE _IO('f', 244)
263 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
264 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
265 #define LL_IOC_MIGRATE _IOR('f', 247, int)
267 #define LL_STATFS_LMV 1
268 #define LL_STATFS_LOV 2
269 #define LL_STATFS_NODELAY 4
271 #define IOC_MDC_TYPE 'i'
272 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
273 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
274 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
275 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
277 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
279 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
280 * files, but are unlikely to be used in practice and are not harmful if
281 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
282 * devices and are safe for use on new files. See LU-4209. */
283 /* To be compatible with old statically linked binary we keep the check for
284 * the older 0100000000 flag. This is already removed upstream. LU-812. */
285 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
286 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
287 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
288 O_LOV_DELAY_CREATE_MASK)
290 #define LL_FILE_IGNORE_LOCK 0x00000001
291 #define LL_FILE_GROUP_LOCKED 0x00000002
292 #define LL_FILE_READAHEA 0x00000004
293 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
294 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
295 #define LL_FILE_RMTACL 0x00000020
297 #define LOV_USER_MAGIC_V1 0x0BD10BD0
298 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
299 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
300 #define LOV_USER_MAGIC_V3 0x0BD30BD0
302 #define LMV_USER_MAGIC 0x0CD30CD0 /*default lmv magic*/
304 #define LOV_PATTERN_RAID0 0x001
305 #define LOV_PATTERN_RAID1 0x002
306 #define LOV_PATTERN_FIRST 0x100
308 #define LOV_MAXPOOLNAME 16
309 #define LOV_POOLNAMEF "%.16s"
311 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
312 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
313 #define LOV_MAX_STRIPE_COUNT_OLD 160
314 /* This calculation is crafted so that input of 4096 will result in 160
315 * which in turn is equal to old maximal stripe count.
316 * XXX: In fact this is too simpified for now, what it also need is to get
317 * ea_type argument to clearly know how much space each stripe consumes.
319 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
320 * allocation that is sufficient for the current generation of systems.
322 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
323 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
324 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
325 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
327 #define XATTR_LUSTRE_PREFIX "lustre."
328 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
330 #define lov_user_ost_data lov_user_ost_data_v1
331 struct lov_user_ost_data_v1 { /* per-stripe data structure */
332 struct ost_id l_ost_oi; /* OST object ID */
333 __u32 l_ost_gen; /* generation of this OST index */
334 __u32 l_ost_idx; /* OST index in LOV */
335 } __attribute__((packed));
337 #define lov_user_md lov_user_md_v1
338 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
339 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
340 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
341 struct ost_id lmm_oi; /* LOV object ID */
342 __u32 lmm_stripe_size; /* size of stripe in bytes */
343 __u16 lmm_stripe_count; /* num stripes in use for this object */
345 __u16 lmm_stripe_offset; /* starting stripe offset in
346 * lmm_objects, use when writing */
347 __u16 lmm_layout_gen; /* layout generation number
348 * used when reading */
350 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
351 } __attribute__((packed, __may_alias__));
353 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
354 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
355 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
356 struct ost_id lmm_oi; /* LOV object ID */
357 __u32 lmm_stripe_size; /* size of stripe in bytes */
358 __u16 lmm_stripe_count; /* num stripes in use for this object */
360 __u16 lmm_stripe_offset; /* starting stripe offset in
361 * lmm_objects, use when writing */
362 __u16 lmm_layout_gen; /* layout generation number
363 * used when reading */
365 char lmm_pool_name[LOV_MAXPOOLNAME]; /* pool name */
366 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
367 } __attribute__((packed));
369 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
371 if (lmm_magic == LOV_USER_MAGIC_V3)
372 return sizeof(struct lov_user_md_v3) +
373 stripes * sizeof(struct lov_user_ost_data_v1);
375 return sizeof(struct lov_user_md_v1) +
376 stripes * sizeof(struct lov_user_ost_data_v1);
379 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
380 * use this. It is unsafe to #define those values in this header as it
381 * is possible the application has already #included <sys/stat.h>. */
382 #ifdef HAVE_LOV_USER_MDS_DATA
383 #define lov_user_mds_data lov_user_mds_data_v1
384 struct lov_user_mds_data_v1 {
385 lstat_t lmd_st; /* MDS stat struct */
386 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
387 } __attribute__((packed));
389 struct lov_user_mds_data_v3 {
390 lstat_t lmd_st; /* MDS stat struct */
391 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
392 } __attribute__((packed));
395 struct lmv_user_mds_data {
396 struct lu_fid lum_fid;
401 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
402 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
403 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
404 #define lmv_user_md lmv_user_md_v1
405 struct lmv_user_md_v1 {
406 __u32 lum_magic; /* must be the first field */
407 __u32 lum_stripe_count; /* dirstripe count */
408 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
409 __u32 lum_hash_type; /* Dir stripe policy */
410 __u32 lum_type; /* LMV type: default or normal */
414 char lum_pool_name[LOV_MAXPOOLNAME];
415 struct lmv_user_mds_data lum_objects[0];
416 } __attribute__((packed));
418 static inline int lmv_user_md_size(int stripes, int lmm_magic)
420 return sizeof(struct lmv_user_md) +
421 stripes * sizeof(struct lmv_user_mds_data);
424 extern void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
426 struct ll_recreate_obj {
432 __u64 id; /* holds object id */
433 __u32 generation; /* holds object generation */
434 __u32 f_type; /* holds object type or stripe idx when passing it to
435 * OST for saving into EA. */
443 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
444 const struct obd_uuid *u2)
446 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
449 static inline int obd_uuid_empty(struct obd_uuid *uuid)
451 return uuid->uuid[0] == '\0';
454 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
456 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
457 uuid->uuid[sizeof(*uuid) - 1] = '\0';
460 /* For printf's only, make sure uuid is terminated */
461 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
466 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
467 /* Obviously not safe, but for printfs, no real harm done...
468 we're always null-terminated, even in a race. */
469 static char temp[sizeof(*uuid)];
470 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
471 temp[sizeof(*uuid) - 1] = '\0';
474 return (char *)(uuid->uuid);
477 /* Extract fsname from uuid (or target name) of a target
478 e.g. (myfs-OST0007_UUID -> myfs)
479 see also deuuidify. */
480 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
484 strncpy(buf, uuid, buflen - 1);
485 buf[buflen - 1] = '\0';
486 p = strrchr(buf, '-');
491 /* printf display format
492 e.g. printf("file FID is "DFID"\n", PFID(fid)); */
493 #define FID_NOBRACE_LEN 40
494 #define FID_LEN (FID_NOBRACE_LEN + 2)
495 #define DFID_NOBRACE LPX64":0x%x:0x%x"
496 #define DFID "["DFID_NOBRACE"]"
502 /* scanf input parse format -- strip '[' first.
503 e.g. sscanf(fidstr, SFID, RFID(&fid)); */
504 /* #define SFID "0x"LPX64i":0x"LPSZX":0x"LPSZX""
505 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 4 has type 'unsigned int *'
506 liblustreapi.c:2893: warning: format '%lx' expects type 'long unsigned int *', but argument 5 has type 'unsigned int *'
508 #define SFID "0x"LPX64i":0x%x:0x%x"
515 /********* Quotas **********/
517 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
518 #define LUSTRE_Q_QUOTAON 0x800002 /* turn quotas on */
519 #define LUSTRE_Q_QUOTAOFF 0x800003 /* turn quotas off */
520 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
521 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
522 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
523 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
524 /* lustre-specific control commands */
525 #define LUSTRE_Q_INVALIDATE 0x80000b /* invalidate quota data */
526 #define LUSTRE_Q_FINVALIDATE 0x80000c /* invalidate filter quota data */
528 #define UGQUOTA 2 /* set both USRQUOTA and GRPQUOTA */
530 struct if_quotacheck {
532 struct obd_uuid obd_uuid;
535 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
538 #define N_PERMS_MAX 64
540 struct perm_downcall_data {
546 struct identity_downcall_data {
553 struct perm_downcall_data idd_perms[N_PERMS_MAX];
557 /* for non-mapped uid/gid */
558 #define NOBODY_UID 99
559 #define NOBODY_GID 99
561 #define INVALID_ID (-1)
570 #ifdef NEED_QUOTA_DEFS
572 #define QIF_BLIMITS 1
574 #define QIF_ILIMITS 4
578 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
579 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
580 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
581 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
584 #endif /* !__KERNEL__ */
586 /* lustre volatile file support
587 * file name header: .^L^S^T^R:volatile"
589 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
590 #define LUSTRE_VOLATILE_HDR_LEN 14
592 typedef enum lustre_quota_version {
594 } lustre_quota_version_t;
596 /* XXX: same as if_dqinfo struct in kernel */
604 /* XXX: same as if_dqblk struct in kernel, plus one padding */
606 __u64 dqb_bhardlimit;
607 __u64 dqb_bsoftlimit;
609 __u64 dqb_ihardlimit;
610 __u64 dqb_isoftlimit;
632 struct obd_dqinfo qc_dqinfo;
633 struct obd_dqblk qc_dqblk;
635 struct obd_uuid obd_uuid;
638 /* swap layout flags */
639 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
640 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
641 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
642 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
644 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
645 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
646 struct lustre_swap_layouts {
655 /********* Changelogs **********/
656 /** Changelog record types */
657 enum changelog_rec_type {
659 CL_CREATE = 1, /* namespace */
660 CL_MKDIR = 2, /* namespace */
661 CL_HARDLINK = 3, /* namespace */
662 CL_SOFTLINK = 4, /* namespace */
663 CL_MKNOD = 5, /* namespace */
664 CL_UNLINK = 6, /* namespace */
665 CL_RMDIR = 7, /* namespace */
666 CL_RENAME = 8, /* namespace */
667 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
668 CL_OPEN = 10, /* not currently used */
669 CL_CLOSE = 11, /* may be written to log only with mtime change */
670 CL_LAYOUT = 12, /* file layout/striping modified */
674 CL_HSM = 16, /* HSM specific events, see flags */
675 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
681 static inline const char *changelog_type2str(int type) {
682 static const char *changelog_str[] = {
683 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
684 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
685 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME",
688 if (type >= 0 && type < CL_LAST)
689 return changelog_str[type];
693 /* per-record flags */
694 #define CLF_VERSION 0x1000
695 #define CLF_EXT_VERSION 0x2000
696 #define CLF_FLAGSHIFT 12
697 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
698 #define CLF_VERMASK (~CLF_FLAGMASK)
699 /* Anything under the flagmask may be per-type (if desired) */
700 /* Flags for unlink */
701 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
702 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
703 /* HSM cleaning needed */
704 /* Flags for rename */
705 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
707 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
708 * has an archive in backend */
711 /* 12b used (from high weight to low weight):
716 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
717 #define CLF_HSM_ERR_H 6
718 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
719 #define CLF_HSM_EVENT_H 9
720 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
721 #define CLF_HSM_FLAG_H 11
722 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
723 #define CLF_HSM_SPARE_H 15
724 #define CLF_HSM_LAST 15
726 /* Remove bits higher than _h, then extract the value
727 * between _h and _l by shifting lower weigth to bit 0. */
728 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
729 >> (CLF_HSM_LAST - _h + _l))
731 #define CLF_HSM_SUCCESS 0x00
732 #define CLF_HSM_MAXERROR 0x7E
733 #define CLF_HSM_ERROVERFLOW 0x7F
735 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
737 /* 3 bits field => 8 values allowed */
749 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
751 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
755 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
757 *flags |= (he << CLF_HSM_EVENT_L);
760 static inline __u16 hsm_get_cl_flags(int flags)
762 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
765 static inline void hsm_set_cl_flags(int *flags, int bits)
767 *flags |= (bits << CLF_HSM_FLAG_L);
770 static inline int hsm_get_cl_error(int flags)
772 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
775 static inline void hsm_set_cl_error(int *flags, int error)
777 *flags |= (error << CLF_HSM_ERR_L);
780 #define CR_MAXSIZE cfs_size_round(2*NAME_MAX + 1 + \
781 sizeof(struct changelog_ext_rec))
783 struct changelog_rec {
785 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
786 __u32 cr_type; /**< \a changelog_rec_type */
787 __u64 cr_index; /**< changelog record number */
788 __u64 cr_prev; /**< last index for this target fid */
791 lustre_fid cr_tfid; /**< target fid */
792 __u32 cr_markerflags; /**< CL_MARK flags */
794 lustre_fid cr_pfid; /**< parent fid */
795 char cr_name[0]; /**< last element */
796 } __attribute__((packed));
798 /* changelog_ext_rec is 2*sizeof(lu_fid) bigger than changelog_rec, to save
799 * space, only rename uses changelog_ext_rec, while others use changelog_rec to
802 struct changelog_ext_rec {
804 __u16 cr_flags; /**< (flags & CLF_FLAGMASK) |
806 __u32 cr_type; /**< \a changelog_rec_type */
807 __u64 cr_index; /**< changelog record number */
808 __u64 cr_prev; /**< last index for this target fid */
811 lustre_fid cr_tfid; /**< target fid */
812 __u32 cr_markerflags; /**< CL_MARK flags */
814 lustre_fid cr_pfid; /**< target parent fid */
815 lustre_fid cr_sfid; /**< source fid, or zero */
816 lustre_fid cr_spfid; /**< source parent fid, or zero */
817 char cr_name[0]; /**< last element */
818 } __attribute__((packed));
820 #define CHANGELOG_REC_EXTENDED(rec) \
821 (((rec)->cr_flags & CLF_VERMASK) == CLF_EXT_VERSION)
823 static inline int changelog_rec_size(struct changelog_rec *rec)
825 return CHANGELOG_REC_EXTENDED(rec) ? sizeof(struct changelog_ext_rec):
829 static inline char *changelog_rec_name(struct changelog_rec *rec)
831 return CHANGELOG_REC_EXTENDED(rec) ?
832 ((struct changelog_ext_rec *)rec)->cr_name: rec->cr_name;
835 static inline int changelog_rec_snamelen(struct changelog_ext_rec *rec)
837 return rec->cr_namelen - strlen(rec->cr_name) - 1;
840 static inline char *changelog_rec_sname(struct changelog_ext_rec *rec)
842 return rec->cr_name + strlen(rec->cr_name) + 1;
845 struct ioc_changelog {
852 enum changelog_message_type {
853 CL_RECORD = 10, /* message is a changelog_rec */
854 CL_EOF = 11, /* at end of current changelog */
857 /********* Misc **********/
859 struct ioc_data_version {
861 __u64 idv_flags; /* See LL_DV_xxx */
863 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
864 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
867 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
870 #define dot_lustre_name ".lustre"
873 /********* HSM **********/
875 /** HSM per-file state
876 * See HSM_FLAGS below.
879 HS_EXISTS = 0x00000001,
880 HS_DIRTY = 0x00000002,
881 HS_RELEASED = 0x00000004,
882 HS_ARCHIVED = 0x00000008,
883 HS_NORELEASE = 0x00000010,
884 HS_NOARCHIVE = 0x00000020,
885 HS_LOST = 0x00000040,
888 /* HSM user-setable flags. */
889 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
891 /* Other HSM flags. */
892 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
895 * All HSM-related possible flags that could be applied to a file.
896 * This should be kept in sync with hsm_states.
898 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
901 * HSMÂ request progress state
903 enum hsm_progress_states {
910 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
913 case HPS_WAITING: return "waiting";
914 case HPS_RUNNING: return "running";
915 case HPS_DONE: return "done";
916 default: return "unknown";
923 } __attribute__((packed));
926 * Current HSM states of a Lustre file.
928 * This structure purpose is to be sent to user-space mainly. It describes the
929 * current HSM flags and in-progress action.
931 struct hsm_user_state {
932 /** Current HSM states, from enum hsm_states. */
934 __u32 hus_archive_id;
935 /** The current undergoing action, if there is one */
936 __u32 hus_in_progress_state;
937 __u32 hus_in_progress_action;
938 struct hsm_extent hus_in_progress_location;
939 char hus_extended_info[];
942 struct hsm_state_set_ioc {
943 struct lu_fid hssi_fid;
945 __u64 hssi_clearmask;
949 * This structure describes the current in-progress action for a file.
950 * it is retuned to user space and send over the wire
952 struct hsm_current_action {
953 /** The current undergoing action, if there is one */
954 /* state is one of hsm_progress_states */
956 /* action is one of hsm_user_action */
958 struct hsm_extent hca_location;
961 /***** HSM user requests ******/
962 /* User-generated (lfs/ioctl) request types */
963 enum hsm_user_action {
964 HUA_NONE = 1, /* no action (noop) */
965 HUA_ARCHIVE = 10, /* copy to hsm */
966 HUA_RESTORE = 11, /* prestage */
967 HUA_RELEASE = 12, /* drop ost objects */
968 HUA_REMOVE = 13, /* remove from archive */
969 HUA_CANCEL = 14 /* cancel a request */
972 static inline const char *hsm_user_action2name(enum hsm_user_action a)
975 case HUA_NONE: return "NOOP";
976 case HUA_ARCHIVE: return "ARCHIVE";
977 case HUA_RESTORE: return "RESTORE";
978 case HUA_RELEASE: return "RELEASE";
979 case HUA_REMOVE: return "REMOVE";
980 case HUA_CANCEL: return "CANCEL";
981 default: return "UNKNOWN";
986 * List of hr_flags (bit field)
988 #define HSM_FORCE_ACTION 0x0001
989 /* used by CT, connot be set by user */
990 #define HSM_GHOST_COPY 0x0002
993 * Contains all the fixed part of struct hsm_user_request.
997 __u32 hr_action; /* enum hsm_user_action */
998 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
999 __u64 hr_flags; /* request flags */
1000 __u32 hr_itemcount; /* item count in hur_user_item vector */
1004 struct hsm_user_item {
1006 struct hsm_extent hui_extent;
1007 } __attribute__((packed));
1009 struct hsm_user_request {
1010 struct hsm_request hur_request;
1011 struct hsm_user_item hur_user_item[0];
1012 /* extra data blob at end of struct (after all
1013 * hur_user_items), only use helpers to access it
1015 } __attribute__((packed));
1017 /** Return pointer to data field in a hsm user request */
1018 static inline void *hur_data(struct hsm_user_request *hur)
1020 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1023 /** Compute the current length of the provided hsm_user_request. */
1024 static inline int hur_len(struct hsm_user_request *hur)
1026 return offsetof(struct hsm_user_request, hur_user_item[0]) +
1027 hur->hur_request.hr_itemcount * sizeof(hur->hur_user_item[0]) +
1028 hur->hur_request.hr_data_len;
1031 /****** HSM RPCs to copytool *****/
1032 /* Message types the copytool may receive */
1033 enum hsm_message_type {
1034 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1037 /* Actions the copytool may be instructed to take for a given action_item */
1038 enum hsm_copytool_action {
1039 HSMA_NONE = 10, /* no action */
1040 HSMA_ARCHIVE = 20, /* arbitrary offset */
1046 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1049 case HSMA_NONE: return "NOOP";
1050 case HSMA_ARCHIVE: return "ARCHIVE";
1051 case HSMA_RESTORE: return "RESTORE";
1052 case HSMA_REMOVE: return "REMOVE";
1053 case HSMA_CANCEL: return "CANCEL";
1054 default: return "UNKNOWN";
1058 /* Copytool item action description */
1059 struct hsm_action_item {
1060 __u32 hai_len; /* valid size of this struct */
1061 __u32 hai_action; /* hsm_copytool_action, but use known size */
1062 lustre_fid hai_fid; /* Lustre FID to operated on */
1063 lustre_fid hai_dfid; /* fid used for data access */
1064 struct hsm_extent hai_extent; /* byte range to operate on */
1065 __u64 hai_cookie; /* action cookie from coordinator */
1066 __u64 hai_gid; /* grouplock id */
1067 char hai_data[0]; /* variable length */
1068 } __attribute__((packed));
1071 * helper function which print in hexa the first bytes of
1073 * \param hai [IN] record to print
1074 * \param buffer [OUT] output buffer
1075 * \param len [IN] max buffer len
1078 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1079 char *buffer, int len)
1081 int i, sz, data_len;
1086 data_len = hai->hai_len - sizeof(*hai);
1087 for (i = 0 ; (i < data_len) && (sz > 0) ; i++)
1091 cnt = snprintf(ptr, sz, "%.2X",
1092 (unsigned char)hai->hai_data[i]);
1100 /* Copytool action list */
1101 #define HAL_VERSION 1
1102 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1103 struct hsm_action_list {
1105 __u32 hal_count; /* number of hai's to follow */
1106 __u64 hal_compound_id; /* returned by coordinator */
1108 __u32 hal_archive_id; /* which archive backend */
1110 char hal_fsname[0]; /* null-terminated */
1111 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1112 boundaries. See hai_zero */
1113 } __attribute__((packed));
1115 #ifndef HAVE_CFS_SIZE_ROUND
1116 static inline int cfs_size_round (int val)
1118 return (val + 7) & (~0x7);
1120 #define HAVE_CFS_SIZE_ROUND
1123 /* Return pointer to first hai in action list */
1124 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1126 return (struct hsm_action_item *)(hal->hal_fsname +
1127 cfs_size_round(strlen(hal-> \
1131 /* Return pointer to next hai */
1132 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1134 return (struct hsm_action_item *)((char *)hai +
1135 cfs_size_round(hai->hai_len));
1138 /* Return size of an hsm_action_list */
1139 static inline size_t hal_size(struct hsm_action_list *hal)
1143 struct hsm_action_item *hai;
1145 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1146 hai = hai_first(hal);
1147 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1148 sz += cfs_size_round(hai->hai_len);
1154 * describe the attributes to be set on imported file
1156 struct hsm_user_import {
1165 __u32 hui_archive_id;
1168 /* Copytool progress reporting */
1169 #define HP_FLAG_COMPLETED 0x01
1170 #define HP_FLAG_RETRY 0x02
1172 struct hsm_progress {
1175 struct hsm_extent hp_extent;
1177 __u16 hp_errval; /* positive val */
1182 __u64 hc_data_version;
1184 __u16 hc_errval; /* positive val */
1186 struct hsm_action_item hc_hai;
1190 enum llapi_json_types {
1191 LLAPI_JSON_INTEGER = 1,
1197 struct llapi_json_item {
1206 struct llapi_json_item *lji_next;
1209 struct llapi_json_item_list {
1210 int ljil_item_count;
1211 struct llapi_json_item *ljil_items;
1214 /** @} lustreuser */
1216 #endif /* _LUSTRE_USER_H */