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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_user.h
34 * Lustre public user-space interface definitions.
37 #ifndef _LUSTRE_USER_H
38 #define _LUSTRE_USER_H
40 /** \defgroup lustreuser lustreuser
45 #include <linux/types.h>
48 # include <linux/fs.h>
49 # include <linux/quota.h>
50 # include <linux/string.h> /* snprintf() */
51 # include <linux/version.h>
52 # include <uapi/linux/lustre/lustre_fiemap.h>
53 #else /* !__KERNEL__ */
54 # define NEED_QUOTA_DEFS
57 # include <stdio.h> /* snprintf() */
59 # include <sys/quota.h>
60 # include <sys/stat.h>
61 # include <linux/lustre/lustre_fiemap.h>
62 #endif /* __KERNEL__ */
64 #if defined(__cplusplus)
69 * This is a temporary solution of adding quota type.
70 * Should be removed as soon as system header is updated.
73 #define LL_MAXQUOTAS 3
75 #define INITQFNAMES { \
76 "user", /* USRQUOTA */ \
77 "group", /* GRPQUOTA */ \
78 "project", /* PRJQUOTA */ \
83 #if defined(__x86_64__) || defined(__ia64__) || defined(__ppc64__) || \
84 defined(__craynv) || defined(__mips64__) || defined(__powerpc64__) || \
86 typedef struct stat lstat_t;
87 # define lstat_f lstat
88 # define fstat_f fstat
89 # define fstatat_f fstatat
90 # define HAVE_LOV_USER_MDS_DATA
91 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
92 typedef struct stat64 lstat_t;
93 # define lstat_f lstat64
94 # define fstat_f fstat64
95 # define fstatat_f fstatat64
96 # define HAVE_LOV_USER_MDS_DATA
99 #define LUSTRE_EOF 0xffffffffffffffffULL
102 #define LL_SUPER_MAGIC 0x0BD00BD0
104 #ifndef FSFILT_IOC_GETFLAGS
105 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
106 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
107 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
108 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
109 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
110 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
113 /* FIEMAP flags supported by Lustre */
114 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
116 enum obd_statfs_state {
117 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
118 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
119 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
120 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
134 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
135 __u32 os_fprecreated; /* objs available now to the caller */
136 /* used in QoS code to find preferred
151 * FID is a cluster-wide unique identifier of a file or an object (stripe).
152 * FIDs are never reused.
156 * FID sequence. Sequence is a unit of migration: all files (objects)
157 * with FIDs from a given sequence are stored on the same server.
158 * Lustre should support 2^64 objects, so even if each sequence
159 * has only a single object we can still enumerate 2^64 objects.
162 /* FID number within sequence. */
165 * FID version, used to distinguish different versions (in the sense
166 * of snapshots, etc.) of the same file system object. Not currently
172 static inline bool fid_is_zero(const struct lu_fid *fid)
174 return fid->f_seq == 0 && fid->f_oid == 0;
177 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
178 * MDT-object's FID::f_ver, instead it is the OST-object index in its
179 * parent MDT-object's layout EA. */
180 #define f_stripe_idx f_ver
183 __u32 ol_stripe_size;
184 __u32 ol_stripe_count;
188 } __attribute__((packed));
190 /* keep this one for compatibility */
191 struct filter_fid_old {
192 struct lu_fid ff_parent;
198 struct lu_fid ff_parent;
199 struct ost_layout ff_layout;
200 } __attribute__((packed));
202 /* Userspace should treat lu_fid as opaque, and only use the following methods
203 * to print or parse them. Other functions (e.g. compare, swab) could be moved
204 * here from lustre_idl.h if needed. */
208 LMAC_HSM = 0x00000001,
209 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
210 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
211 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
212 * under /O/<seq>/d<x>. */
213 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
214 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
218 * Masks for all features that should be supported by a Lustre version to
219 * access a specific file.
220 * This information is stored in lustre_mdt_attrs::lma_incompat.
223 LMAI_RELEASED = 0x00000001, /* file is released */
224 LMAI_AGENT = 0x00000002, /* agent inode */
225 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
226 is on the remote MDT */
227 LMAI_STRIPED = 0x00000008, /* striped directory inode */
228 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
229 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
230 LMAI_STRIPED | LMAI_ORPHAN)
235 * Following struct for object attributes, that will be kept inode's EA.
236 * Introduced in 2.0 release (please see b15993, for details)
237 * Added to all objects since Lustre 2.4 as contains self FID
239 struct lustre_mdt_attrs {
241 * Bitfield for supported data in this structure. From enum lma_compat.
242 * lma_self_fid and lma_flags are always available.
246 * Per-file incompat feature list. Lustre version should support all
247 * flags set in this field. The supported feature mask is available in
251 /** FID of this inode */
252 struct lu_fid lma_self_fid;
255 struct lustre_ost_attrs {
256 /* Use lustre_mdt_attrs directly for now, need a common header
257 * structure if want to change lustre_mdt_attrs in future. */
258 struct lustre_mdt_attrs loa_lma;
260 /* Below five elements are for OST-object's PFID EA, the
261 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
262 * and the stripe_index (low 16 bits), the size should not exceed
263 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
264 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
265 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
266 * loa_comp_* elements are valid. */
267 struct lu_fid loa_parent_fid;
268 __u32 loa_stripe_size;
270 __u64 loa_comp_start;
275 * Prior to 2.4, the LMA structure also included SOM attributes which has since
276 * been moved to a dedicated xattr
277 * lma_flags was also removed because of lma_compat/incompat fields.
279 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
282 * OST object IDentifier.
290 struct lu_fid oi_fid;
294 #define DOSTID "%#llx:%llu"
295 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
296 ((unsigned long long)ostid_id(oi))
298 struct ll_futimes_3 {
300 __u64 lfu_atime_nsec;
302 __u64 lfu_mtime_nsec;
304 __u64 lfu_ctime_nsec;
308 * The ioctl naming rules:
309 * LL_* - works on the currently opened filehandle instead of parent dir
310 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
311 * *_MDC_* - gets/sets data related to MDC
312 * *_LOV_* - gets/sets data related to OSC/LOV
313 * *FILE* - called on parent dir and passes in a filename
314 * *STRIPE* - set/get lov_user_md
315 * *INFO - set/get lov_user_mds_data
317 /* lustre_ioctl.h 101-150 */
318 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
319 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
320 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
321 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
322 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
323 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
324 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
325 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
326 /* LL_IOC_RECREATE_OBJ 157 obsolete */
327 /* LL_IOC_RECREATE_FID 157 obsolete */
328 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
329 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
330 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
331 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
332 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
333 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
334 /* IOC_LOV_GETINFO 165 obsolete */
335 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
336 /* LL_IOC_RMTACL 167 obsolete */
337 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
338 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
339 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
340 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
341 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
342 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
343 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
344 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
345 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
346 /* lustre_ioctl.h 177-210 */
347 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
348 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
349 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
350 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
351 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
352 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
353 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
354 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
355 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
356 struct lustre_swap_layouts)
357 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
358 struct hsm_current_action)
359 /* lustre_ioctl.h 221-232 */
360 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
361 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
362 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
363 #define LL_IOC_SET_LEASE _IOWR('f', 243, long)
364 #define LL_IOC_GET_LEASE _IO('f', 244)
365 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
366 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
367 #define LL_IOC_MIGRATE _IOR('f', 247, int)
368 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
369 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
370 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
372 #ifndef FS_IOC_FSGETXATTR
374 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
377 __u32 fsx_xflags; /* xflags field value (get/set) */
378 __u32 fsx_extsize; /* extsize field value (get/set)*/
379 __u32 fsx_nextents; /* nextents field value (get) */
380 __u32 fsx_projid; /* project identifier (get/set) */
381 unsigned char fsx_pad[12];
383 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
384 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
386 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
387 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
390 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
392 LL_LEASE_RDLCK = 0x1,
393 LL_LEASE_WRLCK = 0x2,
394 LL_LEASE_UNLCK = 0x4,
397 #define LL_STATFS_LMV 1
398 #define LL_STATFS_LOV 2
399 #define LL_STATFS_NODELAY 4
401 #define IOC_MDC_TYPE 'i'
402 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
403 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
404 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
405 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
407 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
409 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
410 * files, but are unlikely to be used in practice and are not harmful if
411 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
412 * devices and are safe for use on new files. See LU-4209. */
413 /* To be compatible with old statically linked binary we keep the check for
414 * the older 0100000000 flag. This is already removed upstream. LU-812. */
415 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
416 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
417 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
418 O_LOV_DELAY_CREATE_MASK)
420 #define LL_FILE_IGNORE_LOCK 0x00000001
421 #define LL_FILE_GROUP_LOCKED 0x00000002
422 #define LL_FILE_READAHEA 0x00000004
423 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
424 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
426 #define LOV_USER_MAGIC_V1 0x0BD10BD0
427 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
428 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
429 #define LOV_USER_MAGIC_V3 0x0BD30BD0
430 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
431 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
432 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
434 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
435 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
437 #define LOV_PATTERN_NONE 0x000
438 #define LOV_PATTERN_RAID0 0x001
439 #define LOV_PATTERN_RAID1 0x002
440 #define LOV_PATTERN_MDT 0x100
441 #define LOV_PATTERN_CMOBD 0x200
443 #define LOV_PATTERN_F_MASK 0xffff0000
444 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
445 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
446 #define LOV_PATTERN_DEFAULT 0xffffffff
448 static inline bool lov_pattern_supported(__u32 pattern)
450 return pattern == LOV_PATTERN_RAID0 ||
451 pattern == LOV_PATTERN_MDT ||
452 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_F_RELEASED);
455 #define LOV_MAXPOOLNAME 15
456 #define LOV_POOLNAMEF "%.15s"
458 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
459 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
460 #define LOV_MAX_STRIPE_COUNT_OLD 160
461 /* This calculation is crafted so that input of 4096 will result in 160
462 * which in turn is equal to old maximal stripe count.
463 * XXX: In fact this is too simpified for now, what it also need is to get
464 * ea_type argument to clearly know how much space each stripe consumes.
466 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
467 * allocation that is sufficient for the current generation of systems.
469 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
470 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
471 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
472 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
474 #define XATTR_LUSTRE_PREFIX "lustre."
475 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
477 #define lov_user_ost_data lov_user_ost_data_v1
478 struct lov_user_ost_data_v1 { /* per-stripe data structure */
479 struct ost_id l_ost_oi; /* OST object ID */
480 __u32 l_ost_gen; /* generation of this OST index */
481 __u32 l_ost_idx; /* OST index in LOV */
482 } __attribute__((packed));
484 #define lov_user_md lov_user_md_v1
485 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
486 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
487 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
488 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
489 __u32 lmm_stripe_size; /* size of stripe in bytes */
490 __u16 lmm_stripe_count; /* num stripes in use for this object */
492 __u16 lmm_stripe_offset; /* starting stripe offset in
493 * lmm_objects, use when writing */
494 __u16 lmm_layout_gen; /* layout generation number
495 * used when reading */
497 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
498 } __attribute__((packed, __may_alias__));
500 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
501 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
502 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
503 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
504 __u32 lmm_stripe_size; /* size of stripe in bytes */
505 __u16 lmm_stripe_count; /* num stripes in use for this object */
507 __u16 lmm_stripe_offset; /* starting stripe offset in
508 * lmm_objects, use when writing */
509 __u16 lmm_layout_gen; /* layout generation number
510 * used when reading */
512 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
513 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
514 } __attribute__((packed));
521 #define DEXT "[ %#llx , %#llx )"
522 #define PEXT(ext) (ext)->e_start, (ext)->e_end
524 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
525 struct lu_extent *e2)
527 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
530 enum lov_comp_md_entry_flags {
531 LCME_FL_PRIMARY = 0x00000001, /* Not used */
532 LCME_FL_STALE = 0x00000002, /* Not used */
533 LCME_FL_OFFLINE = 0x00000004, /* Not used */
534 LCME_FL_PREFERRED = 0x00000008, /* Not used */
535 LCME_FL_INIT = 0x00000010, /* instantiated */
536 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
537 won't be stored on disk */
540 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT)
542 /* lcme_id can be specified as certain flags, and the the first
543 * bit of lcme_id is used to indicate that the ID is representing
544 * certain LCME_FL_* but not a real ID. Which implies we can have
545 * at most 31 flags (see LCME_FL_XXX). */
548 LCME_ID_MAX = 0x7FFFFFFF,
549 LCME_ID_ALL = 0xFFFFFFFF,
550 LCME_ID_NOT_ID = LCME_FL_NEG
553 #define LCME_ID_MASK LCME_ID_MAX
555 struct lov_comp_md_entry_v1 {
556 __u32 lcme_id; /* unique id of component */
557 __u32 lcme_flags; /* LCME_FL_XXX */
558 struct lu_extent lcme_extent; /* file extent for component */
559 __u32 lcme_offset; /* offset of component blob,
560 start from lov_comp_md_v1 */
561 __u32 lcme_size; /* size of component blob */
562 __u64 lcme_padding[2];
563 } __attribute__((packed));
565 enum lov_comp_md_flags;
567 struct lov_comp_md_v1 {
568 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
569 __u32 lcm_size; /* overall size including this struct */
570 __u32 lcm_layout_gen;
572 __u16 lcm_entry_count;
575 struct lov_comp_md_entry_v1 lcm_entries[0];
576 } __attribute__((packed));
578 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
580 if (stripes == (__u16)-1)
583 if (lmm_magic == LOV_USER_MAGIC_V1)
584 return sizeof(struct lov_user_md_v1) +
585 stripes * sizeof(struct lov_user_ost_data_v1);
586 return sizeof(struct lov_user_md_v3) +
587 stripes * sizeof(struct lov_user_ost_data_v1);
590 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
591 * use this. It is unsafe to #define those values in this header as it
592 * is possible the application has already #included <sys/stat.h>. */
593 #ifdef HAVE_LOV_USER_MDS_DATA
594 #define lov_user_mds_data lov_user_mds_data_v1
595 struct lov_user_mds_data_v1 {
596 lstat_t lmd_st; /* MDS stat struct */
597 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
598 } __attribute__((packed));
600 struct lov_user_mds_data_v3 {
601 lstat_t lmd_st; /* MDS stat struct */
602 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
603 } __attribute__((packed));
606 struct lmv_user_mds_data {
607 struct lu_fid lum_fid;
613 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
614 LMV_HASH_TYPE_ALL_CHARS = 1,
615 LMV_HASH_TYPE_FNV_1A_64 = 2,
619 #define LMV_HASH_NAME_ALL_CHARS "all_char"
620 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
622 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
624 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
625 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
626 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
627 #define lmv_user_md lmv_user_md_v1
628 struct lmv_user_md_v1 {
629 __u32 lum_magic; /* must be the first field */
630 __u32 lum_stripe_count; /* dirstripe count */
631 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
632 __u32 lum_hash_type; /* Dir stripe policy */
633 __u32 lum_type; /* LMV type: default or normal */
637 char lum_pool_name[LOV_MAXPOOLNAME + 1];
638 struct lmv_user_mds_data lum_objects[0];
639 } __attribute__((packed));
641 static inline int lmv_user_md_size(int stripes, int lmm_magic)
643 return sizeof(struct lmv_user_md) +
644 stripes * sizeof(struct lmv_user_mds_data);
647 struct ll_recreate_obj {
653 __u64 id; /* holds object id */
654 __u32 generation; /* holds object generation */
655 __u32 f_type; /* holds object type or stripe idx when passing it to
656 * OST for saving into EA. */
664 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
665 const struct obd_uuid *u2)
667 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
670 static inline int obd_uuid_empty(struct obd_uuid *uuid)
672 return uuid->uuid[0] == '\0';
675 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
677 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
678 uuid->uuid[sizeof(*uuid) - 1] = '\0';
681 /* For printf's only, make sure uuid is terminated */
682 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
687 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
688 /* Obviously not safe, but for printfs, no real harm done...
689 we're always null-terminated, even in a race. */
690 static char temp[sizeof(*uuid)];
691 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
692 temp[sizeof(*uuid) - 1] = '\0';
695 return (char *)(uuid->uuid);
698 #define LUSTRE_MAXFSNAME 8
700 /* Extract fsname from uuid (or target name) of a target
701 e.g. (myfs-OST0007_UUID -> myfs)
702 see also deuuidify. */
703 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
707 strncpy(buf, uuid, buflen - 1);
708 buf[buflen - 1] = '\0';
709 p = strrchr(buf, '-');
714 /* printf display format for Lustre FIDs
715 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
716 #define FID_NOBRACE_LEN 40
717 #define FID_LEN (FID_NOBRACE_LEN + 2)
718 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
719 #define DFID "["DFID_NOBRACE"]"
720 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
722 /* scanf input parse format for fids in DFID_NOBRACE format
723 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
724 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
725 #define SFID "0x%llx:0x%x:0x%x"
726 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
728 /********* Quotas **********/
730 #define LUSTRE_QUOTABLOCK_BITS 10
731 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
733 static inline __u64 lustre_stoqb(size_t space)
735 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
738 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
739 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
740 #define Q_GETOINFO 0x800102 /* get obd quota info */
741 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
742 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
744 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
745 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
746 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
747 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
748 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
749 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
750 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
751 /* lustre-specific control commands */
752 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
753 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
755 #define ALLQUOTA 255 /* set all quota */
756 static inline char *qtype_name(int qtype)
769 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
772 #define N_PERMS_MAX 64
774 struct perm_downcall_data {
780 struct identity_downcall_data {
787 struct perm_downcall_data idd_perms[N_PERMS_MAX];
791 #ifdef NEED_QUOTA_DEFS
793 #define QIF_BLIMITS 1
795 #define QIF_ILIMITS 4
799 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
800 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
801 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
802 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
805 #endif /* !__KERNEL__ */
807 /* lustre volatile file support
808 * file name header: .^L^S^T^R:volatile"
810 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
811 #define LUSTRE_VOLATILE_HDR_LEN 14
813 typedef enum lustre_quota_version {
815 } lustre_quota_version_t;
817 /* XXX: same as if_dqinfo struct in kernel */
825 /* XXX: same as if_dqblk struct in kernel, plus one padding */
827 __u64 dqb_bhardlimit;
828 __u64 dqb_bsoftlimit;
830 __u64 dqb_ihardlimit;
831 __u64 dqb_isoftlimit;
853 struct obd_dqinfo qc_dqinfo;
854 struct obd_dqblk qc_dqblk;
856 struct obd_uuid obd_uuid;
859 /* swap layout flags */
860 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
861 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
862 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
863 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
864 #define SWAP_LAYOUTS_CLOSE (1 << 4)
866 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
867 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
868 struct lustre_swap_layouts {
877 /********* Changelogs **********/
878 /** Changelog record types */
879 enum changelog_rec_type {
881 CL_CREATE = 1, /* namespace */
882 CL_MKDIR = 2, /* namespace */
883 CL_HARDLINK = 3, /* namespace */
884 CL_SOFTLINK = 4, /* namespace */
885 CL_MKNOD = 5, /* namespace */
886 CL_UNLINK = 6, /* namespace */
887 CL_RMDIR = 7, /* namespace */
888 CL_RENAME = 8, /* namespace */
889 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
890 CL_OPEN = 10, /* not currently used */
891 CL_CLOSE = 11, /* may be written to log only with mtime change */
892 CL_LAYOUT = 12, /* file layout/striping modified */
896 CL_HSM = 16, /* HSM specific events, see flags */
897 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
904 static inline const char *changelog_type2str(int type) {
905 static const char *changelog_str[] = {
906 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
907 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
908 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT"
911 if (type >= 0 && type < CL_LAST)
912 return changelog_str[type];
916 /* per-record flags */
917 #define CLF_FLAGSHIFT 12
918 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
919 #define CLF_VERMASK (~CLF_FLAGMASK)
920 enum changelog_rec_flags {
921 CLF_VERSION = 0x1000,
924 CLF_EXTRA_FLAGS = 0x8000,
925 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS
929 /* Anything under the flagmask may be per-type (if desired) */
930 /* Flags for unlink */
931 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
932 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
933 /* HSM cleaning needed */
934 /* Flags for rename */
935 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
937 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
938 * has an archive in backend */
941 /* 12b used (from high weight to low weight):
946 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
947 #define CLF_HSM_ERR_H 6
948 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
949 #define CLF_HSM_EVENT_H 9
950 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
951 #define CLF_HSM_FLAG_H 11
952 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
953 #define CLF_HSM_SPARE_H 15
954 #define CLF_HSM_LAST 15
956 /* Remove bits higher than _h, then extract the value
957 * between _h and _l by shifting lower weigth to bit 0. */
958 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
959 >> (CLF_HSM_LAST - _h + _l))
961 #define CLF_HSM_SUCCESS 0x00
962 #define CLF_HSM_MAXERROR 0x7E
963 #define CLF_HSM_ERROVERFLOW 0x7F
965 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
967 /* 3 bits field => 8 values allowed */
979 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
981 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
985 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
987 *flags |= (he << CLF_HSM_EVENT_L);
990 static inline __u16 hsm_get_cl_flags(int flags)
992 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
995 static inline void hsm_set_cl_flags(int *flags, int bits)
997 *flags |= (bits << CLF_HSM_FLAG_L);
1000 static inline int hsm_get_cl_error(int flags)
1002 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1005 static inline void hsm_set_cl_error(int *flags, int error)
1007 *flags |= (error << CLF_HSM_ERR_L);
1010 enum changelog_rec_extra_flags {
1011 CLFE_INVALID = 0, /* No additional flags currently implemented */
1012 CLFE_SUPPORTED = CLFE_INVALID
1015 enum changelog_send_flag {
1016 /* Not yet implemented */
1017 CHANGELOG_FLAG_FOLLOW = 0x01,
1018 /* Blocking IO makes sense in case of slow user parsing of the records,
1019 * but it also prevents us from cleaning up if the records are not
1021 CHANGELOG_FLAG_BLOCK = 0x02,
1022 /* Pack jobid into the changelog records if available. */
1023 CHANGELOG_FLAG_JOBID = 0x04,
1024 /* Pack additional flag bits into the changelog record */
1025 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1028 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
1029 changelog_rec_offset(CLF_SUPPORTED, \
1032 /* 31 usable bytes string + null terminator. */
1033 #define LUSTRE_JOBID_SIZE 32
1035 /* This is the minimal changelog record. It can contain extensions
1036 * such as rename fields or process jobid. Its exact content is described
1037 * by the cr_flags and cr_extra_flags.
1039 * Extensions are packed in the same order as their corresponding flags,
1040 * then in the same order as their corresponding extra flags.
1042 struct changelog_rec {
1044 __u16 cr_flags; /**< \a changelog_rec_flags */
1045 __u32 cr_type; /**< \a changelog_rec_type */
1046 __u64 cr_index; /**< changelog record number */
1047 __u64 cr_prev; /**< last index for this target fid */
1050 struct lu_fid cr_tfid; /**< target fid */
1051 __u32 cr_markerflags; /**< CL_MARK flags */
1053 struct lu_fid cr_pfid; /**< parent fid */
1056 /* Changelog extension for RENAME. */
1057 struct changelog_ext_rename {
1058 struct lu_fid cr_sfid; /**< source fid, or zero */
1059 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1062 /* Changelog extension to include JOBID. */
1063 struct changelog_ext_jobid {
1064 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1067 /* Changelog extension to include additional flags. */
1068 struct changelog_ext_extra_flags {
1069 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1072 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1073 const struct changelog_rec *rec);
1075 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1076 enum changelog_rec_extra_flags cref)
1078 size_t size = sizeof(struct changelog_rec);
1080 if (crf & CLF_RENAME)
1081 size += sizeof(struct changelog_ext_rename);
1083 if (crf & CLF_JOBID)
1084 size += sizeof(struct changelog_ext_jobid);
1086 if (crf & CLF_EXTRA_FLAGS) {
1087 size += sizeof(struct changelog_ext_extra_flags);
1093 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1095 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1097 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1098 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1100 return changelog_rec_offset(rec->cr_flags, cref);
1103 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1105 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1109 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1111 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1113 return (struct changelog_ext_rename *)((char *)rec +
1114 changelog_rec_offset(crf,
1118 /* The jobid follows the rename extension, if present */
1120 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1122 enum changelog_rec_flags crf = rec->cr_flags &
1123 (CLF_VERSION | CLF_RENAME);
1125 return (struct changelog_ext_jobid *)((char *)rec +
1126 changelog_rec_offset(crf,
1130 /* The additional flags follow the rename and jobid extensions, if present */
1132 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1133 const struct changelog_rec *rec)
1135 enum changelog_rec_flags crf = rec->cr_flags &
1136 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1138 return (struct changelog_ext_extra_flags *)((char *)rec +
1139 changelog_rec_offset(crf,
1143 /* The name follows the rename, jobid and extra flags extns, if present */
1144 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1146 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1148 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1149 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1151 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1152 cref & CLFE_SUPPORTED);
1155 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1157 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1160 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1162 char *cr_name = changelog_rec_name(rec);
1164 return cr_name + strlen(cr_name) + 1;
1168 * Remap a record to the desired format as specified by the crf flags.
1169 * The record must be big enough to contain the final remapped version.
1170 * Superfluous extension fields are removed and missing ones are added
1171 * and zeroed. The flags of the record are updated accordingly.
1173 * The jobid and rename extensions can be added to a record, to match the
1174 * format an application expects, typically. In this case, the newly added
1175 * fields will be zeroed.
1176 * The Jobid field can be removed, to guarantee compatibility with older
1177 * clients that don't expect this field in the records they process.
1179 * The following assumptions are being made:
1180 * - CLF_RENAME will not be removed
1181 * - CLF_JOBID will not be added without CLF_RENAME being added too
1182 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1184 * @param[in,out] rec The record to remap.
1185 * @param[in] crf_wanted Flags describing the desired extensions.
1186 * @param[in] cref_want Flags describing the desired extra extensions.
1188 static inline void changelog_remap_rec(struct changelog_rec *rec,
1189 enum changelog_rec_flags crf_wanted,
1190 enum changelog_rec_extra_flags cref_want)
1196 crf_wanted &= CLF_SUPPORTED;
1197 cref_want &= CLFE_SUPPORTED;
1199 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1200 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1201 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1202 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1208 /* First move the variable-length name field */
1209 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1210 changelog_rec_name(rec), rec->cr_namelen);
1212 /* Locations of extensions in the remapped record */
1213 ef_mov = (char *)rec +
1214 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1216 jid_mov = (char *)rec +
1217 changelog_rec_offset(crf_wanted &
1218 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1220 rnm_mov = (char *)rec +
1221 changelog_rec_offset(crf_wanted &
1227 /* Move the extension fields to the desired positions */
1228 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1229 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1230 memmove(ef_mov, changelog_rec_extra_flags(rec),
1231 sizeof(struct changelog_ext_extra_flags));
1234 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1235 memmove(jid_mov, changelog_rec_jobid(rec),
1236 sizeof(struct changelog_ext_jobid));
1238 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1239 memmove(rnm_mov, changelog_rec_rename(rec),
1240 sizeof(struct changelog_ext_rename));
1242 /* Clear newly added fields */
1243 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1244 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1245 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1247 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1248 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1250 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1251 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1253 /* Update the record's flags accordingly */
1254 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1255 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1256 changelog_rec_extra_flags(rec)->cr_extra_flags =
1257 changelog_rec_extra_flags(rec)->cr_extra_flags |
1261 enum changelog_message_type {
1262 CL_RECORD = 10, /* message is a changelog_rec */
1263 CL_EOF = 11, /* at end of current changelog */
1266 /********* Misc **********/
1268 struct ioc_data_version {
1270 __u64 idv_flags; /* See LL_DV_xxx */
1272 #define LL_DV_RD_FLUSH (1 << 0) /* Flush dirty pages from clients */
1273 #define LL_DV_WR_FLUSH (1 << 1) /* Flush all caching pages from clients */
1276 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1279 #define dot_lustre_name ".lustre"
1282 /********* HSM **********/
1284 /** HSM per-file state
1285 * See HSM_FLAGS below.
1288 HS_NONE = 0x00000000,
1289 HS_EXISTS = 0x00000001,
1290 HS_DIRTY = 0x00000002,
1291 HS_RELEASED = 0x00000004,
1292 HS_ARCHIVED = 0x00000008,
1293 HS_NORELEASE = 0x00000010,
1294 HS_NOARCHIVE = 0x00000020,
1295 HS_LOST = 0x00000040,
1298 /* HSM user-setable flags. */
1299 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1301 /* Other HSM flags. */
1302 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1305 * All HSM-related possible flags that could be applied to a file.
1306 * This should be kept in sync with hsm_states.
1308 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1311 * HSMÂ request progress state
1313 enum hsm_progress_states {
1320 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1323 case HPS_WAITING: return "waiting";
1324 case HPS_RUNNING: return "running";
1325 case HPS_DONE: return "done";
1326 default: return "unknown";
1333 } __attribute__((packed));
1336 * Current HSM states of a Lustre file.
1338 * This structure purpose is to be sent to user-space mainly. It describes the
1339 * current HSM flags and in-progress action.
1341 struct hsm_user_state {
1342 /** Current HSM states, from enum hsm_states. */
1344 __u32 hus_archive_id;
1345 /** The current undergoing action, if there is one */
1346 __u32 hus_in_progress_state;
1347 __u32 hus_in_progress_action;
1348 struct hsm_extent hus_in_progress_location;
1349 char hus_extended_info[];
1352 struct hsm_state_set_ioc {
1353 struct lu_fid hssi_fid;
1355 __u64 hssi_clearmask;
1359 * This structure describes the current in-progress action for a file.
1360 * it is retuned to user space and send over the wire
1362 struct hsm_current_action {
1363 /** The current undergoing action, if there is one */
1364 /* state is one of hsm_progress_states */
1366 /* action is one of hsm_user_action */
1368 struct hsm_extent hca_location;
1371 /***** HSM user requests ******/
1372 /* User-generated (lfs/ioctl) request types */
1373 enum hsm_user_action {
1374 HUA_NONE = 1, /* no action (noop) */
1375 HUA_ARCHIVE = 10, /* copy to hsm */
1376 HUA_RESTORE = 11, /* prestage */
1377 HUA_RELEASE = 12, /* drop ost objects */
1378 HUA_REMOVE = 13, /* remove from archive */
1379 HUA_CANCEL = 14 /* cancel a request */
1382 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1385 case HUA_NONE: return "NOOP";
1386 case HUA_ARCHIVE: return "ARCHIVE";
1387 case HUA_RESTORE: return "RESTORE";
1388 case HUA_RELEASE: return "RELEASE";
1389 case HUA_REMOVE: return "REMOVE";
1390 case HUA_CANCEL: return "CANCEL";
1391 default: return "UNKNOWN";
1396 * List of hr_flags (bit field)
1398 #define HSM_FORCE_ACTION 0x0001
1399 /* used by CT, cannot be set by user */
1400 #define HSM_GHOST_COPY 0x0002
1403 * Contains all the fixed part of struct hsm_user_request.
1406 struct hsm_request {
1407 __u32 hr_action; /* enum hsm_user_action */
1408 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1409 __u64 hr_flags; /* request flags */
1410 __u32 hr_itemcount; /* item count in hur_user_item vector */
1414 struct hsm_user_item {
1415 struct lu_fid hui_fid;
1416 struct hsm_extent hui_extent;
1417 } __attribute__((packed));
1419 struct hsm_user_request {
1420 struct hsm_request hur_request;
1421 struct hsm_user_item hur_user_item[0];
1422 /* extra data blob at end of struct (after all
1423 * hur_user_items), only use helpers to access it
1425 } __attribute__((packed));
1427 /** Return pointer to data field in a hsm user request */
1428 static inline void *hur_data(struct hsm_user_request *hur)
1430 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1434 * Compute the current length of the provided hsm_user_request. This returns -1
1435 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1437 * return -1 on bounds check error.
1439 static inline ssize_t hur_len(struct hsm_user_request *hur)
1443 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1444 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1445 (__u64)hur->hur_request.hr_itemcount *
1446 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1448 if (size != (ssize_t)size)
1454 /****** HSM RPCs to copytool *****/
1455 /* Message types the copytool may receive */
1456 enum hsm_message_type {
1457 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1460 /* Actions the copytool may be instructed to take for a given action_item */
1461 enum hsm_copytool_action {
1462 HSMA_NONE = 10, /* no action */
1463 HSMA_ARCHIVE = 20, /* arbitrary offset */
1469 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1472 case HSMA_NONE: return "NOOP";
1473 case HSMA_ARCHIVE: return "ARCHIVE";
1474 case HSMA_RESTORE: return "RESTORE";
1475 case HSMA_REMOVE: return "REMOVE";
1476 case HSMA_CANCEL: return "CANCEL";
1477 default: return "UNKNOWN";
1481 /* Copytool item action description */
1482 struct hsm_action_item {
1483 __u32 hai_len; /* valid size of this struct */
1484 __u32 hai_action; /* hsm_copytool_action, but use known size */
1485 struct lu_fid hai_fid; /* Lustre FID to operate on */
1486 struct lu_fid hai_dfid; /* fid used for data access */
1487 struct hsm_extent hai_extent; /* byte range to operate on */
1488 __u64 hai_cookie; /* action cookie from coordinator */
1489 __u64 hai_gid; /* grouplock id */
1490 char hai_data[0]; /* variable length */
1491 } __attribute__((packed));
1494 * helper function which print in hexa the first bytes of
1497 * \param hai [IN] record to print
1498 * \param buffer [IN,OUT] buffer to write the hex string to
1499 * \param len [IN] max buffer length
1503 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1504 char *buffer, size_t len)
1511 data_len = hai->hai_len - sizeof(*hai);
1512 for (i = 0; (i < data_len) && (len > 2); i++) {
1513 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1523 /* Copytool action list */
1524 #define HAL_VERSION 1
1525 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1526 struct hsm_action_list {
1528 __u32 hal_count; /* number of hai's to follow */
1529 __u64 hal_compound_id; /* returned by coordinator */
1531 __u32 hal_archive_id; /* which archive backend */
1533 char hal_fsname[0]; /* null-terminated */
1534 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1535 boundaries. See hai_zero */
1536 } __attribute__((packed));
1538 #ifndef HAVE_CFS_SIZE_ROUND
1539 static inline int cfs_size_round (int val)
1541 return (val + 7) & (~0x7);
1543 #define HAVE_CFS_SIZE_ROUND
1546 /* Return pointer to first hai in action list */
1547 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1549 return (struct hsm_action_item *)(hal->hal_fsname +
1550 cfs_size_round(strlen(hal-> \
1554 /* Return pointer to next hai */
1555 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1557 return (struct hsm_action_item *)((char *)hai +
1558 cfs_size_round(hai->hai_len));
1561 /* Return size of an hsm_action_list */
1562 static inline size_t hal_size(struct hsm_action_list *hal)
1566 struct hsm_action_item *hai;
1568 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1569 hai = hai_first(hal);
1570 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1571 sz += cfs_size_round(hai->hai_len);
1577 * describe the attributes to be set on imported file
1579 struct hsm_user_import {
1588 __u32 hui_archive_id;
1591 /* Copytool progress reporting */
1592 #define HP_FLAG_COMPLETED 0x01
1593 #define HP_FLAG_RETRY 0x02
1595 struct hsm_progress {
1596 struct lu_fid hp_fid;
1598 struct hsm_extent hp_extent;
1600 __u16 hp_errval; /* positive val */
1605 __u64 hc_data_version;
1607 __u16 hc_errval; /* positive val */
1609 struct hsm_action_item hc_hai;
1613 enum llapi_json_types {
1614 LLAPI_JSON_INTEGER = 1,
1620 struct llapi_json_item {
1629 struct llapi_json_item *lji_next;
1632 struct llapi_json_item_list {
1633 int ljil_item_count;
1634 struct llapi_json_item *ljil_items;
1637 enum lu_ladvise_type {
1638 LU_LADVISE_INVALID = 0,
1639 LU_LADVISE_WILLREAD = 1,
1640 LU_LADVISE_DONTNEED = 2,
1641 LU_LADVISE_LOCKNOEXPAND = 3,
1642 LU_LADVISE_LOCKAHEAD = 4,
1646 #define LU_LADVISE_NAMES { \
1647 [LU_LADVISE_WILLREAD] = "willread", \
1648 [LU_LADVISE_DONTNEED] = "dontneed", \
1649 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
1650 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
1653 /* This is the userspace argument for ladvise. It is currently the same as
1654 * what goes on the wire (struct lu_ladvise), but is defined separately as we
1655 * may need info which is only used locally. */
1656 struct llapi_lu_ladvise {
1657 __u16 lla_advice; /* advice type */
1658 __u16 lla_value1; /* values for different advice types */
1660 __u64 lla_start; /* first byte of extent for advice */
1661 __u64 lla_end; /* last byte of extent for advice */
1667 LF_ASYNC = 0x00000001,
1668 LF_UNSET = 0x00000002,
1671 #define LADVISE_MAGIC 0x1ADF1CE0
1672 /* Masks of valid flags for each advice */
1673 #define LF_LOCKNOEXPAND_MASK LF_UNSET
1674 /* Flags valid for all advices not explicitly specified */
1675 #define LF_DEFAULT_MASK LF_ASYNC
1677 #define LF_MASK (LF_ASYNC | LF_UNSET)
1679 #define lla_lockahead_mode lla_value1
1680 #define lla_peradvice_flags lla_value2
1681 #define lla_lockahead_result lla_value3
1683 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
1684 * is used on the wire. It is defined separately as we may need info which is
1685 * only used locally. */
1686 struct llapi_ladvise_hdr {
1687 __u32 lah_magic; /* LADVISE_MAGIC */
1688 __u32 lah_count; /* number of advices */
1689 __u64 lah_flags; /* from enum ladvise_flag */
1690 __u32 lah_value1; /* unused */
1691 __u32 lah_value2; /* unused */
1692 __u64 lah_value3; /* unused */
1693 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
1696 #define LAH_COUNT_MAX (1024)
1700 SK_CRYPT_INVALID = -1,
1702 SK_CRYPT_AES256_CTR = 1,
1707 SK_HMAC_INVALID = -1,
1714 struct sk_crypt_type {
1719 struct sk_hmac_type {
1724 enum lock_mode_user {
1730 #define LOCK_MODE_NAMES { \
1731 [MODE_READ_USER] = "READ",\
1732 [MODE_WRITE_USER] = "WRITE"\
1735 enum lockahead_results {
1736 LLA_RESULT_SENT = 0,
1737 LLA_RESULT_DIFFERENT,
1741 #if defined(__cplusplus)
1745 /** @} lustreuser */
1747 #endif /* _LUSTRE_USER_H */