• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * GPL HEADER START
3  *
4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5  *
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.
9  *
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).
15  *
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
19  *
20  * GPL HEADER END
21  */
22 /*
23  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Use is subject to license terms.
25  *
26  * Copyright (c) 2010, 2015, Intel Corporation.
27  */
28 /*
29  * This file is part of Lustre, http://www.lustre.org/
30  * Lustre is a trademark of Sun Microsystems, Inc.
31  *
32  * lustre/include/lustre/lustre_user.h
33  *
34  * Lustre public user-space interface definitions.
35  */
36 
37 #ifndef _LUSTRE_USER_H
38 #define _LUSTRE_USER_H
39 
40 /** \defgroup lustreuser lustreuser
41  *
42  * @{
43  */
44 
45 #ifdef __KERNEL__
46 # include <linux/quota.h>
47 # include <linux/string.h> /* snprintf() */
48 # include <linux/version.h>
49 #else /* !__KERNEL__ */
50 # define NEED_QUOTA_DEFS
51 # include <stdio.h> /* snprintf() */
52 # include <string.h>
53 # include <sys/quota.h>
54 # include <sys/stat.h>
55 #endif /* __KERNEL__ */
56 #include "ll_fiemap.h"
57 
58 /*
59  * We need to always use 64bit version because the structure
60  * is shared across entire cluster where 32bit and 64bit machines
61  * are co-existing.
62  */
63 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
64 typedef struct stat64   lstat_t;
65 #define lstat_f  lstat64
66 #else
67 typedef struct stat     lstat_t;
68 #define lstat_f  lstat
69 #endif
70 
71 #define HAVE_LOV_USER_MDS_DATA
72 
73 #define LUSTRE_EOF 0xffffffffffffffffULL
74 
75 /* for statfs() */
76 #define LL_SUPER_MAGIC 0x0BD00BD0
77 
78 #ifndef FSFILT_IOC_GETFLAGS
79 #define FSFILT_IOC_GETFLAGS	       _IOR('f', 1, long)
80 #define FSFILT_IOC_SETFLAGS	       _IOW('f', 2, long)
81 #define FSFILT_IOC_GETVERSION	     _IOR('f', 3, long)
82 #define FSFILT_IOC_SETVERSION	     _IOW('f', 4, long)
83 #define FSFILT_IOC_GETVERSION_OLD	 _IOR('v', 1, long)
84 #define FSFILT_IOC_SETVERSION_OLD	 _IOW('v', 2, long)
85 #define FSFILT_IOC_FIEMAP		 _IOWR('f', 11, struct ll_user_fiemap)
86 #endif
87 
88 /* FIEMAP flags supported by Lustre */
89 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
90 
91 enum obd_statfs_state {
92 	OS_STATE_DEGRADED       = 0x00000001, /**< RAID degraded/rebuilding */
93 	OS_STATE_READONLY       = 0x00000002, /**< filesystem is read-only */
94 	OS_STATE_RDONLY_1       = 0x00000004, /**< obsolete 1.6, was EROFS=30 */
95 	OS_STATE_RDONLY_2       = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
96 	OS_STATE_RDONLY_3       = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
97 };
98 
99 struct obd_statfs {
100 	__u64	   os_type;
101 	__u64	   os_blocks;
102 	__u64	   os_bfree;
103 	__u64	   os_bavail;
104 	__u64	   os_files;
105 	__u64	   os_ffree;
106 	__u8	    os_fsid[40];
107 	__u32	   os_bsize;
108 	__u32	   os_namelen;
109 	__u64	   os_maxbytes;
110 	__u32	   os_state;       /**< obd_statfs_state OS_STATE_* flag */
111 	__u32	   os_fprecreated; /* objs available now to the caller */
112 				   /* used in QoS code to find preferred OSTs */
113 	__u32	   os_spare2;
114 	__u32	   os_spare3;
115 	__u32	   os_spare4;
116 	__u32	   os_spare5;
117 	__u32	   os_spare6;
118 	__u32	   os_spare7;
119 	__u32	   os_spare8;
120 	__u32	   os_spare9;
121 };
122 
123 /**
124  * File IDentifier.
125  *
126  * FID is a cluster-wide unique identifier of a file or an object (stripe).
127  * FIDs are never reused.
128  **/
129 struct lu_fid {
130        /**
131 	* FID sequence. Sequence is a unit of migration: all files (objects)
132 	* with FIDs from a given sequence are stored on the same server.
133 	* Lustre should support 2^64 objects, so even if each sequence
134 	* has only a single object we can still enumerate 2^64 objects.
135 	**/
136 	__u64 f_seq;
137 	/* FID number within sequence. */
138 	__u32 f_oid;
139 	/**
140 	 * FID version, used to distinguish different versions (in the sense
141 	 * of snapshots, etc.) of the same file system object. Not currently
142 	 * used.
143 	 **/
144 	__u32 f_ver;
145 };
146 
fid_is_zero(const struct lu_fid * fid)147 static inline bool fid_is_zero(const struct lu_fid *fid)
148 {
149 	return !fid->f_seq && !fid->f_oid;
150 }
151 
152 struct filter_fid {
153 	struct lu_fid	ff_parent;  /* ff_parent.f_ver == file stripe number */
154 };
155 
156 /* keep this one for compatibility */
157 struct filter_fid_old {
158 	struct lu_fid	ff_parent;
159 	__u64		ff_objid;
160 	__u64		ff_seq;
161 };
162 
163 /* Userspace should treat lu_fid as opaque, and only use the following methods
164  * to print or parse them.  Other functions (e.g. compare, swab) could be moved
165  * here from lustre_idl.h if needed.
166  */
167 struct lu_fid;
168 
169 /**
170  * Following struct for object attributes, that will be kept inode's EA.
171  * Introduced in 2.0 release (please see b15993, for details)
172  * Added to all objects since Lustre 2.4 as contains self FID
173  */
174 struct lustre_mdt_attrs {
175 	/**
176 	 * Bitfield for supported data in this structure. From enum lma_compat.
177 	 * lma_self_fid and lma_flags are always available.
178 	 */
179 	__u32   lma_compat;
180 	/**
181 	 * Per-file incompat feature list. Lustre version should support all
182 	 * flags set in this field. The supported feature mask is available in
183 	 * LMA_INCOMPAT_SUPP.
184 	 */
185 	__u32   lma_incompat;
186 	/** FID of this inode */
187 	struct lu_fid  lma_self_fid;
188 };
189 
190 /**
191  * Prior to 2.4, the LMA structure also included SOM attributes which has since
192  * been moved to a dedicated xattr
193  * lma_flags was also removed because of lma_compat/incompat fields.
194  */
195 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
196 
197 /**
198  * OST object IDentifier.
199  */
200 struct ost_id {
201 	union {
202 		struct {
203 			__u64	oi_id;
204 			__u64	oi_seq;
205 		} oi;
206 		struct lu_fid oi_fid;
207 	};
208 };
209 
210 #define DOSTID "%#llx:%llu"
211 #define POSTID(oi) ostid_seq(oi), ostid_id(oi)
212 
213 /*
214  * The ioctl naming rules:
215  * LL_*     - works on the currently opened filehandle instead of parent dir
216  * *_OBD_*  - gets data for both OSC or MDC (LOV, LMV indirectly)
217  * *_MDC_*  - gets/sets data related to MDC
218  * *_LOV_*  - gets/sets data related to OSC/LOV
219  * *FILE*   - called on parent dir and passes in a filename
220  * *STRIPE* - set/get lov_user_md
221  * *INFO    - set/get lov_user_mds_data
222  */
223 /*	lustre_ioctl.h			101-150 */
224 #define LL_IOC_GETFLAGS		 _IOR('f', 151, long)
225 #define LL_IOC_SETFLAGS		 _IOW('f', 152, long)
226 #define LL_IOC_CLRFLAGS		 _IOW('f', 153, long)
227 #define LL_IOC_LOV_SETSTRIPE	    _IOW('f', 154, long)
228 #define LL_IOC_LOV_GETSTRIPE	    _IOW('f', 155, long)
229 #define LL_IOC_LOV_SETEA		_IOW('f', 156, long)
230 /*	LL_IOC_RECREATE_OBJ		157 obsolete */
231 /*	LL_IOC_RECREATE_FID		158 obsolete */
232 #define LL_IOC_GROUP_LOCK	       _IOW('f', 158, long)
233 #define LL_IOC_GROUP_UNLOCK	     _IOW('f', 159, long)
234 /* #define LL_IOC_QUOTACHECK		160 OBD_IOC_QUOTACHECK */
235 /* #define LL_IOC_POLL_QUOTACHECK	161 OBD_IOC_POLL_QUOTACHECK */
236 /* #define LL_IOC_QUOTACTL		162 OBD_IOC_QUOTACTL */
237 #define IOC_OBD_STATFS		  _IOWR('f', 164, struct obd_statfs *)
238 #define IOC_LOV_GETINFO		 _IOWR('f', 165, struct lov_user_mds_data *)
239 #define LL_IOC_FLUSHCTX		 _IOW('f', 166, long)
240 /* LL_IOC_RMTACL			167 obsolete */
241 #define LL_IOC_GETOBDCOUNT	      _IOR('f', 168, long)
242 #define LL_IOC_LLOOP_ATTACH	     _IOWR('f', 169, long)
243 #define LL_IOC_LLOOP_DETACH	     _IOWR('f', 170, long)
244 #define LL_IOC_LLOOP_INFO	       _IOWR('f', 171, struct lu_fid)
245 #define LL_IOC_LLOOP_DETACH_BYDEV       _IOWR('f', 172, long)
246 #define LL_IOC_PATH2FID		 _IOR('f', 173, long)
247 #define LL_IOC_GET_CONNECT_FLAGS	_IOWR('f', 174, __u64 *)
248 #define LL_IOC_GET_MDTIDX	       _IOR('f', 175, int)
249 
250 /*	lustre_ioctl.h			177-210 */
251 #define LL_IOC_HSM_STATE_GET		_IOR('f', 211, struct hsm_user_state)
252 #define LL_IOC_HSM_STATE_SET		_IOW('f', 212, struct hsm_state_set)
253 #define LL_IOC_HSM_CT_START		_IOW('f', 213, struct lustre_kernelcomm)
254 #define LL_IOC_HSM_COPY_START		_IOW('f', 214, struct hsm_copy *)
255 #define LL_IOC_HSM_COPY_END		_IOW('f', 215, struct hsm_copy *)
256 #define LL_IOC_HSM_PROGRESS		_IOW('f', 216, struct hsm_user_request)
257 #define LL_IOC_HSM_REQUEST		_IOW('f', 217, struct hsm_user_request)
258 #define LL_IOC_DATA_VERSION		_IOR('f', 218, struct ioc_data_version)
259 #define LL_IOC_LOV_SWAP_LAYOUTS		_IOW('f', 219, \
260 						struct lustre_swap_layouts)
261 #define LL_IOC_HSM_ACTION		_IOR('f', 220, \
262 						struct hsm_current_action)
263 /* see <lustre_lib.h> for ioctl numbers 221-232 */
264 
265 #define LL_IOC_LMV_SETSTRIPE	    _IOWR('f', 240, struct lmv_user_md)
266 #define LL_IOC_LMV_GETSTRIPE	    _IOWR('f', 241, struct lmv_user_md)
267 #define LL_IOC_SET_LEASE		_IOWR('f', 243, long)
268 #define LL_IOC_GET_LEASE		_IO('f', 244)
269 #define LL_IOC_HSM_IMPORT		_IOWR('f', 245, struct hsm_user_import)
270 #define LL_IOC_LMV_SET_DEFAULT_STRIPE	_IOWR('f', 246, struct lmv_user_md)
271 #define LL_IOC_MIGRATE			_IOR('f', 247, int)
272 #define LL_IOC_FID2MDTIDX		_IOWR('f', 248, struct lu_fid)
273 #define LL_IOC_GETPARENT		_IOWR('f', 249, struct getparent)
274 
275 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
276 enum ll_lease_type {
277 	LL_LEASE_RDLCK	= 0x1,
278 	LL_LEASE_WRLCK	= 0x2,
279 	LL_LEASE_UNLCK	= 0x4,
280 };
281 
282 #define LL_STATFS_LMV	   1
283 #define LL_STATFS_LOV	   2
284 #define LL_STATFS_NODELAY	4
285 
286 #define IOC_MDC_TYPE	    'i'
287 #define IOC_MDC_LOOKUP	  _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
288 #define IOC_MDC_GETFILESTRIPE   _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
289 #define IOC_MDC_GETFILEINFO     _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
290 #define LL_IOC_MDC_GETINFO      _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
291 
292 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
293 
294 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
295  * files, but are unlikely to be used in practice and are not harmful if
296  * used incorrectly.  O_NOCTTY and FASYNC are only meaningful for character
297  * devices and are safe for use on new files (See LU-812, LU-4209).
298  */
299 #define O_LOV_DELAY_CREATE	(O_NOCTTY | FASYNC)
300 
301 #define LL_FILE_IGNORE_LOCK     0x00000001
302 #define LL_FILE_GROUP_LOCKED    0x00000002
303 #define LL_FILE_READAHEA	0x00000004
304 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
305 #define LL_FILE_LOCKLESS_IO     0x00000010 /* server-side locks with cio */
306 #define LL_FILE_RMTACL	  0x00000020
307 
308 #define LOV_USER_MAGIC_V1	0x0BD10BD0
309 #define LOV_USER_MAGIC		LOV_USER_MAGIC_V1
310 #define LOV_USER_MAGIC_JOIN_V1	0x0BD20BD0
311 #define LOV_USER_MAGIC_V3	0x0BD30BD0
312 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
313 #define LOV_USER_MAGIC_SPECIFIC	0x0BD50BD0	/* for specific OSTs */
314 
315 #define LMV_USER_MAGIC    0x0CD30CD0    /*default lmv magic*/
316 
317 #define LOV_PATTERN_RAID0	0x001
318 #define LOV_PATTERN_RAID1	0x002
319 #define LOV_PATTERN_FIRST	0x100
320 #define LOV_PATTERN_CMOBD	0x200
321 
322 #define LOV_PATTERN_F_MASK	0xffff0000
323 #define LOV_PATTERN_F_HOLE	0x40000000 /* there is hole in LOV EA */
324 #define LOV_PATTERN_F_RELEASED	0x80000000 /* HSM released file */
325 
326 #define LOV_MAXPOOLNAME 15
327 #define LOV_POOLNAMEF "%.15s"
328 
329 #define LOV_MIN_STRIPE_BITS 16   /* maximum PAGE_SIZE (ia64), power of 2 */
330 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
331 #define LOV_MAX_STRIPE_COUNT_OLD 160
332 /* This calculation is crafted so that input of 4096 will result in 160
333  * which in turn is equal to old maximal stripe count.
334  * XXX: In fact this is too simplified for now, what it also need is to get
335  * ea_type argument to clearly know how much space each stripe consumes.
336  *
337  * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
338  * allocation that is sufficient for the current generation of systems.
339  *
340  * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1)
341  */
342 #define LOV_MAX_STRIPE_COUNT 2000  /* ((12 * 4096 - 256) / 24) */
343 #define LOV_ALL_STRIPES       0xffff /* only valid for directories */
344 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
345 
346 #define lov_user_ost_data lov_user_ost_data_v1
347 struct lov_user_ost_data_v1 {     /* per-stripe data structure */
348 	struct ost_id l_ost_oi;	  /* OST object ID */
349 	__u32 l_ost_gen;	  /* generation of this OST index */
350 	__u32 l_ost_idx;	  /* OST index in LOV */
351 } __packed;
352 
353 #define lov_user_md lov_user_md_v1
354 struct lov_user_md_v1 {	   /* LOV EA user data (host-endian) */
355 	__u32 lmm_magic;	  /* magic number = LOV_USER_MAGIC_V1 */
356 	__u32 lmm_pattern;	/* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
357 	struct ost_id lmm_oi;	  /* LOV object ID */
358 	__u32 lmm_stripe_size;    /* size of stripe in bytes */
359 	__u16 lmm_stripe_count;   /* num stripes in use for this object */
360 	union {
361 		__u16 lmm_stripe_offset;  /* starting stripe offset in
362 					   * lmm_objects, use when writing
363 					   */
364 		__u16 lmm_layout_gen;     /* layout generation number
365 					   * used when reading
366 					   */
367 	};
368 	struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
369 } __attribute__((packed,  __may_alias__));
370 
371 struct lov_user_md_v3 {	   /* LOV EA user data (host-endian) */
372 	__u32 lmm_magic;	  /* magic number = LOV_USER_MAGIC_V3 */
373 	__u32 lmm_pattern;	/* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
374 	struct ost_id lmm_oi;	  /* LOV object ID */
375 	__u32 lmm_stripe_size;    /* size of stripe in bytes */
376 	__u16 lmm_stripe_count;   /* num stripes in use for this object */
377 	union {
378 		__u16 lmm_stripe_offset;  /* starting stripe offset in
379 					   * lmm_objects, use when writing
380 					   */
381 		__u16 lmm_layout_gen;     /* layout generation number
382 					   * used when reading
383 					   */
384 	};
385 	char  lmm_pool_name[LOV_MAXPOOLNAME + 1];   /* pool name */
386 	struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
387 } __packed;
388 
lov_user_md_size(__u16 stripes,__u32 lmm_magic)389 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
390 {
391 	if (lmm_magic == LOV_USER_MAGIC_V1)
392 		return sizeof(struct lov_user_md_v1) +
393 				stripes * sizeof(struct lov_user_ost_data_v1);
394 	return sizeof(struct lov_user_md_v3) +
395 	       stripes * sizeof(struct lov_user_ost_data_v1);
396 }
397 
398 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
399  * use this.  It is unsafe to #define those values in this header as it
400  * is possible the application has already #included <sys/stat.h>. */
401 #ifdef HAVE_LOV_USER_MDS_DATA
402 #define lov_user_mds_data lov_user_mds_data_v1
403 struct lov_user_mds_data_v1 {
404 	lstat_t lmd_st;		 /* MDS stat struct */
405 	struct lov_user_md_v1 lmd_lmm;  /* LOV EA V1 user data */
406 } __packed;
407 
408 struct lov_user_mds_data_v3 {
409 	lstat_t lmd_st;		 /* MDS stat struct */
410 	struct lov_user_md_v3 lmd_lmm;  /* LOV EA V3 user data */
411 } __packed;
412 #endif
413 
414 struct lmv_user_mds_data {
415 	struct lu_fid	lum_fid;
416 	__u32		lum_padding;
417 	__u32		lum_mds;
418 };
419 
420 enum lmv_hash_type {
421 	LMV_HASH_TYPE_UNKNOWN	= 0,	/* 0 is reserved for testing purpose */
422 	LMV_HASH_TYPE_ALL_CHARS = 1,
423 	LMV_HASH_TYPE_FNV_1A_64 = 2,
424 };
425 
426 #define LMV_HASH_NAME_ALL_CHARS		"all_char"
427 #define LMV_HASH_NAME_FNV_1A_64		"fnv_1a_64"
428 
429 /*
430  * Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
431  * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data)
432  */
433 #define LMV_MAX_STRIPE_COUNT 2000  /* ((12 * 4096 - 256) / 24) */
434 #define lmv_user_md lmv_user_md_v1
435 struct lmv_user_md_v1 {
436 	__u32	lum_magic;	 /* must be the first field */
437 	__u32	lum_stripe_count;  /* dirstripe count */
438 	__u32	lum_stripe_offset; /* MDT idx for default dirstripe */
439 	__u32	lum_hash_type;     /* Dir stripe policy */
440 	__u32	lum_type;	  /* LMV type: default or normal */
441 	__u32	lum_padding1;
442 	__u32	lum_padding2;
443 	__u32	lum_padding3;
444 	char	lum_pool_name[LOV_MAXPOOLNAME + 1];
445 	struct	lmv_user_mds_data  lum_objects[0];
446 } __packed;
447 
lmv_user_md_size(int stripes,int lmm_magic)448 static inline int lmv_user_md_size(int stripes, int lmm_magic)
449 {
450 	return sizeof(struct lmv_user_md) +
451 		      stripes * sizeof(struct lmv_user_mds_data);
452 }
453 
454 void lustre_swab_lmv_user_md(struct lmv_user_md *lum);
455 
456 struct ll_recreate_obj {
457 	__u64 lrc_id;
458 	__u32 lrc_ost_idx;
459 };
460 
461 struct ll_fid {
462 	__u64 id;	 /* holds object id */
463 	__u32 generation; /* holds object generation */
464 	__u32 f_type;     /* holds object type or stripe idx when passing it to
465 			   * OST for saving into EA. */
466 };
467 
468 #define UUID_MAX	40
469 struct obd_uuid {
470 	char uuid[UUID_MAX];
471 };
472 
obd_uuid_equals(const struct obd_uuid * u1,const struct obd_uuid * u2)473 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
474 				   const struct obd_uuid *u2)
475 {
476 	return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
477 }
478 
obd_uuid_empty(struct obd_uuid * uuid)479 static inline int obd_uuid_empty(struct obd_uuid *uuid)
480 {
481 	return uuid->uuid[0] == '\0';
482 }
483 
obd_str2uuid(struct obd_uuid * uuid,const char * tmp)484 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
485 {
486 	strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
487 	uuid->uuid[sizeof(*uuid) - 1] = '\0';
488 }
489 
490 /* For printf's only, make sure uuid is terminated */
obd_uuid2str(const struct obd_uuid * uuid)491 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
492 {
493 	if (!uuid)
494 		return NULL;
495 
496 	if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
497 		/* Obviously not safe, but for printfs, no real harm done...
498 		 * we're always null-terminated, even in a race.
499 		 */
500 		static char temp[sizeof(*uuid)];
501 
502 		memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
503 		temp[sizeof(*uuid) - 1] = '\0';
504 		return temp;
505 	}
506 	return (char *)(uuid->uuid);
507 }
508 
509 /* Extract fsname from uuid (or target name) of a target
510  * e.g. (myfs-OST0007_UUID -> myfs)
511  * see also deuuidify.
512  */
obd_uuid2fsname(char * buf,char * uuid,int buflen)513 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
514 {
515 	char *p;
516 
517 	strncpy(buf, uuid, buflen - 1);
518 	buf[buflen - 1] = '\0';
519 	p = strrchr(buf, '-');
520 	if (p)
521 		*p = '\0';
522 }
523 
524 /* printf display format
525  * e.g. printf("file FID is "DFID"\n", PFID(fid));
526  */
527 #define FID_NOBRACE_LEN 40
528 #define FID_LEN (FID_NOBRACE_LEN + 2)
529 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
530 #define DFID "["DFID_NOBRACE"]"
531 #define PFID(fid)     \
532 	(fid)->f_seq, \
533 	(fid)->f_oid, \
534 	(fid)->f_ver
535 
536 /* scanf input parse format -- strip '[' first.
537  * e.g. sscanf(fidstr, SFID, RFID(&fid));
538  */
539 #define SFID "0x%llx:0x%x:0x%x"
540 #define RFID(fid)     \
541 	&((fid)->f_seq), \
542 	&((fid)->f_oid), \
543 	&((fid)->f_ver)
544 
545 /********* Quotas **********/
546 
547 #define Q_QUOTACHECK   0x800100 /* deprecated as of 2.4 */
548 #define Q_INITQUOTA    0x800101 /* deprecated as of 2.4  */
549 #define Q_GETOINFO     0x800102 /* get obd quota info */
550 #define Q_GETOQUOTA    0x800103 /* get obd quotas */
551 #define Q_FINVALIDATE  0x800104 /* deprecated as of 2.4 */
552 
553 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
554 #define LUSTRE_Q_QUOTAON    0x800002     /* turn quotas on */
555 #define LUSTRE_Q_QUOTAOFF   0x800003     /* turn quotas off */
556 #define LUSTRE_Q_GETINFO    0x800005     /* get information about quota files */
557 #define LUSTRE_Q_SETINFO    0x800006     /* set information about quota files */
558 #define LUSTRE_Q_GETQUOTA   0x800007     /* get user quota structure */
559 #define LUSTRE_Q_SETQUOTA   0x800008     /* set user quota structure */
560 /* lustre-specific control commands */
561 #define LUSTRE_Q_INVALIDATE  0x80000b     /* invalidate quota data */
562 #define LUSTRE_Q_FINVALIDATE 0x80000c     /* invalidate filter quota data */
563 
564 #define UGQUOTA 2       /* set both USRQUOTA and GRPQUOTA */
565 
566 struct if_quotacheck {
567 	char		    obd_type[16];
568 	struct obd_uuid	 obd_uuid;
569 };
570 
571 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
572 
573 /* permission */
574 #define N_PERMS_MAX      64
575 
576 struct perm_downcall_data {
577 	__u64 pdd_nid;
578 	__u32 pdd_perm;
579 	__u32 pdd_padding;
580 };
581 
582 struct identity_downcall_data {
583 	__u32			    idd_magic;
584 	__u32			    idd_err;
585 	__u32			    idd_uid;
586 	__u32			    idd_gid;
587 	__u32			    idd_nperms;
588 	__u32			    idd_ngroups;
589 	struct perm_downcall_data idd_perms[N_PERMS_MAX];
590 	__u32			    idd_groups[0];
591 };
592 
593 /* lustre volatile file support
594  * file name header: .^L^S^T^R:volatile"
595  */
596 #define LUSTRE_VOLATILE_HDR	".\x0c\x13\x14\x12:VOLATILE"
597 #define LUSTRE_VOLATILE_HDR_LEN	14
598 /* hdr + MDT index */
599 #define LUSTRE_VOLATILE_IDX	LUSTRE_VOLATILE_HDR":%.4X:"
600 
601 enum lustre_quota_version {
602 	LUSTRE_QUOTA_V2 = 1
603 };
604 
605 /* XXX: same as if_dqinfo struct in kernel */
606 struct obd_dqinfo {
607 	__u64 dqi_bgrace;
608 	__u64 dqi_igrace;
609 	__u32 dqi_flags;
610 	__u32 dqi_valid;
611 };
612 
613 /* XXX: same as if_dqblk struct in kernel, plus one padding */
614 struct obd_dqblk {
615 	__u64 dqb_bhardlimit;
616 	__u64 dqb_bsoftlimit;
617 	__u64 dqb_curspace;
618 	__u64 dqb_ihardlimit;
619 	__u64 dqb_isoftlimit;
620 	__u64 dqb_curinodes;
621 	__u64 dqb_btime;
622 	__u64 dqb_itime;
623 	__u32 dqb_valid;
624 	__u32 dqb_padding;
625 };
626 
627 enum {
628 	QC_GENERAL      = 0,
629 	QC_MDTIDX       = 1,
630 	QC_OSTIDX       = 2,
631 	QC_UUID	 = 3
632 };
633 
634 struct if_quotactl {
635 	__u32		   qc_cmd;
636 	__u32		   qc_type;
637 	__u32		   qc_id;
638 	__u32		   qc_stat;
639 	__u32		   qc_valid;
640 	__u32		   qc_idx;
641 	struct obd_dqinfo       qc_dqinfo;
642 	struct obd_dqblk	qc_dqblk;
643 	char		    obd_type[16];
644 	struct obd_uuid	 obd_uuid;
645 };
646 
647 /* swap layout flags */
648 #define SWAP_LAYOUTS_CHECK_DV1		(1 << 0)
649 #define SWAP_LAYOUTS_CHECK_DV2		(1 << 1)
650 #define SWAP_LAYOUTS_KEEP_MTIME		(1 << 2)
651 #define SWAP_LAYOUTS_KEEP_ATIME		(1 << 3)
652 
653 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
654 #define SWAP_LAYOUTS_MDS_HSM		(1 << 31)
655 struct lustre_swap_layouts {
656 	__u64	sl_flags;
657 	__u32	sl_fd;
658 	__u32	sl_gid;
659 	__u64	sl_dv1;
660 	__u64	sl_dv2;
661 };
662 
663 /********* Changelogs **********/
664 /** Changelog record types */
665 enum changelog_rec_type {
666 	CL_MARK     = 0,
667 	CL_CREATE   = 1,  /* namespace */
668 	CL_MKDIR    = 2,  /* namespace */
669 	CL_HARDLINK = 3,  /* namespace */
670 	CL_SOFTLINK = 4,  /* namespace */
671 	CL_MKNOD    = 5,  /* namespace */
672 	CL_UNLINK   = 6,  /* namespace */
673 	CL_RMDIR    = 7,  /* namespace */
674 	CL_RENAME   = 8,  /* namespace */
675 	CL_EXT      = 9,  /* namespace extended record (2nd half of rename) */
676 	CL_OPEN     = 10, /* not currently used */
677 	CL_CLOSE    = 11, /* may be written to log only with mtime change */
678 	CL_LAYOUT   = 12, /* file layout/striping modified */
679 	CL_TRUNC    = 13,
680 	CL_SETATTR  = 14,
681 	CL_XATTR    = 15,
682 	CL_HSM      = 16, /* HSM specific events, see flags */
683 	CL_MTIME    = 17, /* Precedence: setattr > mtime > ctime > atime */
684 	CL_CTIME    = 18,
685 	CL_ATIME    = 19,
686 	CL_LAST
687 };
688 
changelog_type2str(int type)689 static inline const char *changelog_type2str(int type)
690 {
691 	static const char *changelog_str[] = {
692 		"MARK",  "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
693 		"RMDIR", "RENME", "RNMTO", "OPEN",  "CLOSE", "LYOUT", "TRUNC",
694 		"SATTR", "XATTR", "HSM",   "MTIME", "CTIME", "ATIME",
695 	};
696 
697 	if (type >= 0 && type < CL_LAST)
698 		return changelog_str[type];
699 	return NULL;
700 }
701 
702 /* per-record flags */
703 #define CLF_FLAGSHIFT   12
704 #define CLF_FLAGMASK    ((1U << CLF_FLAGSHIFT) - 1)
705 #define CLF_VERMASK     (~CLF_FLAGMASK)
706 enum changelog_rec_flags {
707 	CLF_VERSION	= 0x1000,
708 	CLF_RENAME	= 0x2000,
709 	CLF_JOBID	= 0x4000,
710 	CLF_SUPPORTED	= CLF_VERSION | CLF_RENAME | CLF_JOBID
711 };
712 
713 /* Anything under the flagmask may be per-type (if desired) */
714 /* Flags for unlink */
715 #define CLF_UNLINK_LAST       0x0001 /* Unlink of last hardlink */
716 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
717 				     /* HSM cleaning needed */
718 /* Flags for rename */
719 #define CLF_RENAME_LAST		0x0001	/* rename unlink last hardlink of
720 					 * target
721 					 */
722 #define CLF_RENAME_LAST_EXISTS	0x0002	/* rename unlink last hardlink of target
723 					 * has an archive in backend
724 					 */
725 
726 /* Flags for HSM */
727 /* 12b used (from high weight to low weight):
728  * 2b for flags
729  * 3b for event
730  * 7b for error code
731  */
732 #define CLF_HSM_ERR_L	0 /* HSM return code, 7 bits */
733 #define CLF_HSM_ERR_H	6
734 #define CLF_HSM_EVENT_L      7 /* HSM event, 3 bits, see enum hsm_event */
735 #define CLF_HSM_EVENT_H      9
736 #define CLF_HSM_FLAG_L      10 /* HSM flags, 2 bits, 1 used, 1 spare */
737 #define CLF_HSM_FLAG_H      11
738 #define CLF_HSM_SPARE_L     12 /* 4 spare bits */
739 #define CLF_HSM_SPARE_H     15
740 #define CLF_HSM_LAST	15
741 
742 /* Remove bits higher than _h, then extract the value
743  * between _h and _l by shifting lower weigth to bit 0.
744  */
745 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
746 				   >> (CLF_HSM_LAST - _h + _l))
747 
748 #define CLF_HSM_SUCCESS      0x00
749 #define CLF_HSM_MAXERROR     0x7E
750 #define CLF_HSM_ERROVERFLOW  0x7F
751 
752 #define CLF_HSM_DIRTY	1 /* file is dirty after HSM request end */
753 
754 /* 3 bits field => 8 values allowed */
755 enum hsm_event {
756 	HE_ARCHIVE      = 0,
757 	HE_RESTORE      = 1,
758 	HE_CANCEL       = 2,
759 	HE_RELEASE      = 3,
760 	HE_REMOVE       = 4,
761 	HE_STATE	= 5,
762 	HE_SPARE1       = 6,
763 	HE_SPARE2       = 7,
764 };
765 
hsm_get_cl_event(__u16 flags)766 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
767 {
768 	return CLF_GET_BITS(flags, CLF_HSM_EVENT_H, CLF_HSM_EVENT_L);
769 }
770 
hsm_set_cl_event(int * flags,enum hsm_event he)771 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
772 {
773 	*flags |= (he << CLF_HSM_EVENT_L);
774 }
775 
hsm_get_cl_flags(int flags)776 static inline __u16 hsm_get_cl_flags(int flags)
777 {
778 	return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
779 }
780 
hsm_set_cl_flags(int * flags,int bits)781 static inline void hsm_set_cl_flags(int *flags, int bits)
782 {
783 	*flags |= (bits << CLF_HSM_FLAG_L);
784 }
785 
hsm_get_cl_error(int flags)786 static inline int hsm_get_cl_error(int flags)
787 {
788 	return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
789 }
790 
hsm_set_cl_error(int * flags,int error)791 static inline void hsm_set_cl_error(int *flags, int error)
792 {
793 	*flags |= (error << CLF_HSM_ERR_L);
794 }
795 
796 enum changelog_send_flag {
797 	/* Not yet implemented */
798 	CHANGELOG_FLAG_FOLLOW	= BIT(0),
799 	/*
800 	 * Blocking IO makes sense in case of slow user parsing of the records,
801 	 * but it also prevents us from cleaning up if the records are not
802 	 * consumed.
803 	 */
804 	CHANGELOG_FLAG_BLOCK	= BIT(1),
805 	/* Pack jobid into the changelog records if available. */
806 	CHANGELOG_FLAG_JOBID	= BIT(2),
807 };
808 
809 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
810 				  changelog_rec_offset(CLF_SUPPORTED))
811 
812 /* 31 usable bytes string + null terminator. */
813 #define LUSTRE_JOBID_SIZE	32
814 
815 /*
816  * This is the minimal changelog record. It can contain extensions
817  * such as rename fields or process jobid. Its exact content is described
818  * by the cr_flags.
819  *
820  * Extensions are packed in the same order as their corresponding flags.
821  */
822 struct changelog_rec {
823 	__u16		 cr_namelen;
824 	__u16		 cr_flags; /**< \a changelog_rec_flags */
825 	__u32		 cr_type;  /**< \a changelog_rec_type */
826 	__u64		 cr_index; /**< changelog record number */
827 	__u64		 cr_prev;  /**< last index for this target fid */
828 	__u64		 cr_time;
829 	union {
830 		struct lu_fid    cr_tfid;	/**< target fid */
831 		__u32	 cr_markerflags; /**< CL_MARK flags */
832 	};
833 	struct lu_fid	    cr_pfid;	/**< parent fid */
834 } __packed;
835 
836 /* Changelog extension for RENAME. */
837 struct changelog_ext_rename {
838 	struct lu_fid	cr_sfid;	/**< source fid, or zero */
839 	struct lu_fid	cr_spfid;	/**< source parent fid, or zero */
840 };
841 
842 /* Changelog extension to include JOBID. */
843 struct changelog_ext_jobid {
844 	char	cr_jobid[LUSTRE_JOBID_SIZE];	/**< zero-terminated string. */
845 };
846 
changelog_rec_offset(enum changelog_rec_flags crf)847 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf)
848 {
849 	size_t size = sizeof(struct changelog_rec);
850 
851 	if (crf & CLF_RENAME)
852 		size += sizeof(struct changelog_ext_rename);
853 
854 	if (crf & CLF_JOBID)
855 		size += sizeof(struct changelog_ext_jobid);
856 
857 	return size;
858 }
859 
changelog_rec_size(struct changelog_rec * rec)860 static inline size_t changelog_rec_size(struct changelog_rec *rec)
861 {
862 	return changelog_rec_offset(rec->cr_flags);
863 }
864 
changelog_rec_varsize(struct changelog_rec * rec)865 static inline size_t changelog_rec_varsize(struct changelog_rec *rec)
866 {
867 	return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
868 }
869 
870 static inline
changelog_rec_rename(struct changelog_rec * rec)871 struct changelog_ext_rename *changelog_rec_rename(struct changelog_rec *rec)
872 {
873 	enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
874 
875 	return (struct changelog_ext_rename *)((char *)rec +
876 					       changelog_rec_offset(crf));
877 }
878 
879 /* The jobid follows the rename extension, if present */
880 static inline
changelog_rec_jobid(struct changelog_rec * rec)881 struct changelog_ext_jobid *changelog_rec_jobid(struct changelog_rec *rec)
882 {
883 	enum changelog_rec_flags crf = rec->cr_flags &
884 				       (CLF_VERSION | CLF_RENAME);
885 
886 	return (struct changelog_ext_jobid *)((char *)rec +
887 					      changelog_rec_offset(crf));
888 }
889 
890 /* The name follows the rename and jobid extensions, if present */
changelog_rec_name(struct changelog_rec * rec)891 static inline char *changelog_rec_name(struct changelog_rec *rec)
892 {
893 	return (char *)rec + changelog_rec_offset(rec->cr_flags &
894 						  CLF_SUPPORTED);
895 }
896 
changelog_rec_snamelen(struct changelog_rec * rec)897 static inline size_t changelog_rec_snamelen(struct changelog_rec *rec)
898 {
899 	return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
900 }
901 
changelog_rec_sname(struct changelog_rec * rec)902 static inline char *changelog_rec_sname(struct changelog_rec *rec)
903 {
904 	char *cr_name = changelog_rec_name(rec);
905 
906 	return cr_name + strlen(cr_name) + 1;
907 }
908 
909 /**
910  * Remap a record to the desired format as specified by the crf flags.
911  * The record must be big enough to contain the final remapped version.
912  * Superfluous extension fields are removed and missing ones are added
913  * and zeroed. The flags of the record are updated accordingly.
914  *
915  * The jobid and rename extensions can be added to a record, to match the
916  * format an application expects, typically. In this case, the newly added
917  * fields will be zeroed.
918  * The Jobid field can be removed, to guarantee compatibility with older
919  * clients that don't expect this field in the records they process.
920  *
921  * The following assumptions are being made:
922  *	- CLF_RENAME will not be removed
923  *	- CLF_JOBID will not be added without CLF_RENAME being added too
924  *
925  * @param[in,out]  rec		The record to remap.
926  * @param[in]	   crf_wanted	Flags describing the desired extensions.
927  */
changelog_remap_rec(struct changelog_rec * rec,enum changelog_rec_flags crf_wanted)928 static inline void changelog_remap_rec(struct changelog_rec *rec,
929 				       enum changelog_rec_flags crf_wanted)
930 {
931 	char *jid_mov, *rnm_mov;
932 
933 	crf_wanted &= CLF_SUPPORTED;
934 
935 	if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted)
936 		return;
937 
938 	/* First move the variable-length name field */
939 	memmove((char *)rec + changelog_rec_offset(crf_wanted),
940 		changelog_rec_name(rec), rec->cr_namelen);
941 
942 	/* Locations of jobid and rename extensions in the remapped record */
943 	jid_mov = (char *)rec +
944 		  changelog_rec_offset(crf_wanted & ~CLF_JOBID);
945 	rnm_mov = (char *)rec +
946 		  changelog_rec_offset(crf_wanted & ~(CLF_JOBID | CLF_RENAME));
947 
948 	/* Move the extension fields to the desired positions */
949 	if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
950 		memmove(jid_mov, changelog_rec_jobid(rec),
951 			sizeof(struct changelog_ext_jobid));
952 
953 	if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
954 		memmove(rnm_mov, changelog_rec_rename(rec),
955 			sizeof(struct changelog_ext_rename));
956 
957 	/* Clear newly added fields */
958 	if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
959 		memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
960 
961 	if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
962 		memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
963 
964 	/* Update the record's flags accordingly */
965 	rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
966 }
967 
968 struct ioc_changelog {
969 	__u64 icc_recno;
970 	__u32 icc_mdtindex;
971 	__u32 icc_id;
972 	__u32 icc_flags;
973 };
974 
975 enum changelog_message_type {
976 	CL_RECORD = 10, /* message is a changelog_rec */
977 	CL_EOF    = 11, /* at end of current changelog */
978 };
979 
980 /********* Misc **********/
981 
982 struct ioc_data_version {
983 	__u64 idv_version;
984 	__u64 idv_flags;     /* See LL_DV_xxx */
985 };
986 
987 #define LL_DV_RD_FLUSH	BIT(0)	/* Flush dirty pages from clients */
988 #define LL_DV_WR_FLUSH	BIT(1)	/* Flush all caching pages from clients */
989 
990 #ifndef offsetof
991 # define offsetof(typ, memb)     ((unsigned long)((char *)&(((typ *)0)->memb)))
992 #endif
993 
994 #define dot_lustre_name ".lustre"
995 
996 /********* HSM **********/
997 
998 /** HSM per-file state
999  * See HSM_FLAGS below.
1000  */
1001 enum hsm_states {
1002 	HS_EXISTS	= 0x00000001,
1003 	HS_DIRTY	= 0x00000002,
1004 	HS_RELEASED	= 0x00000004,
1005 	HS_ARCHIVED	= 0x00000008,
1006 	HS_NORELEASE	= 0x00000010,
1007 	HS_NOARCHIVE	= 0x00000020,
1008 	HS_LOST		= 0x00000040,
1009 };
1010 
1011 /* HSM user-setable flags. */
1012 #define HSM_USER_MASK   (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1013 
1014 /* Other HSM flags. */
1015 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1016 
1017 /*
1018  * All HSM-related possible flags that could be applied to a file.
1019  * This should be kept in sync with hsm_states.
1020  */
1021 #define HSM_FLAGS_MASK  (HSM_USER_MASK | HSM_STATUS_MASK)
1022 
1023 /**
1024  * HSM request progress state
1025  */
1026 enum hsm_progress_states {
1027 	HPS_WAITING	= 1,
1028 	HPS_RUNNING	= 2,
1029 	HPS_DONE	= 3,
1030 };
1031 
1032 #define HPS_NONE	0
1033 
hsm_progress_state2name(enum hsm_progress_states s)1034 static inline char *hsm_progress_state2name(enum hsm_progress_states s)
1035 {
1036 	switch  (s) {
1037 	case HPS_WAITING:	return "waiting";
1038 	case HPS_RUNNING:	return "running";
1039 	case HPS_DONE:		return "done";
1040 	default:		return "unknown";
1041 	}
1042 }
1043 
1044 struct hsm_extent {
1045 	__u64 offset;
1046 	__u64 length;
1047 } __packed;
1048 
1049 /**
1050  * Current HSM states of a Lustre file.
1051  *
1052  * This structure purpose is to be sent to user-space mainly. It describes the
1053  * current HSM flags and in-progress action.
1054  */
1055 struct hsm_user_state {
1056 	/** Current HSM states, from enum hsm_states. */
1057 	__u32			hus_states;
1058 	__u32			hus_archive_id;
1059 	/**  The current undergoing action, if there is one */
1060 	__u32			hus_in_progress_state;
1061 	__u32			hus_in_progress_action;
1062 	struct hsm_extent	hus_in_progress_location;
1063 	char			hus_extended_info[];
1064 };
1065 
1066 struct hsm_state_set_ioc {
1067 	struct lu_fid	hssi_fid;
1068 	__u64		hssi_setmask;
1069 	__u64		hssi_clearmask;
1070 };
1071 
1072 /*
1073  * This structure describes the current in-progress action for a file.
1074  * it is returned to user space and send over the wire
1075  */
1076 struct hsm_current_action {
1077 	/**  The current undergoing action, if there is one */
1078 	/* state is one of hsm_progress_states */
1079 	__u32			hca_state;
1080 	/* action is one of hsm_user_action */
1081 	__u32			hca_action;
1082 	struct hsm_extent	hca_location;
1083 };
1084 
1085 /***** HSM user requests ******/
1086 /* User-generated (lfs/ioctl) request types */
1087 enum hsm_user_action {
1088 	HUA_NONE    =  1, /* no action (noop) */
1089 	HUA_ARCHIVE = 10, /* copy to hsm */
1090 	HUA_RESTORE = 11, /* prestage */
1091 	HUA_RELEASE = 12, /* drop ost objects */
1092 	HUA_REMOVE  = 13, /* remove from archive */
1093 	HUA_CANCEL  = 14  /* cancel a request */
1094 };
1095 
hsm_user_action2name(enum hsm_user_action a)1096 static inline char *hsm_user_action2name(enum hsm_user_action  a)
1097 {
1098 	switch  (a) {
1099 	case HUA_NONE:    return "NOOP";
1100 	case HUA_ARCHIVE: return "ARCHIVE";
1101 	case HUA_RESTORE: return "RESTORE";
1102 	case HUA_RELEASE: return "RELEASE";
1103 	case HUA_REMOVE:  return "REMOVE";
1104 	case HUA_CANCEL:  return "CANCEL";
1105 	default:	  return "UNKNOWN";
1106 	}
1107 }
1108 
1109 /*
1110  * List of hr_flags (bit field)
1111  */
1112 #define HSM_FORCE_ACTION 0x0001
1113 /* used by CT, connot be set by user */
1114 #define HSM_GHOST_COPY   0x0002
1115 
1116 /**
1117  * Contains all the fixed part of struct hsm_user_request.
1118  *
1119  */
1120 struct hsm_request {
1121 	__u32 hr_action;	/* enum hsm_user_action */
1122 	__u32 hr_archive_id;	/* archive id, used only with HUA_ARCHIVE */
1123 	__u64 hr_flags;		/* request flags */
1124 	__u32 hr_itemcount;	/* item count in hur_user_item vector */
1125 	__u32 hr_data_len;
1126 };
1127 
1128 struct hsm_user_item {
1129 	struct lu_fid	hui_fid;
1130 	struct hsm_extent hui_extent;
1131 } __packed;
1132 
1133 struct hsm_user_request {
1134 	struct hsm_request	hur_request;
1135 	struct hsm_user_item	hur_user_item[0];
1136 	/* extra data blob at end of struct (after all
1137 	 * hur_user_items), only use helpers to access it
1138 	 */
1139 } __packed;
1140 
1141 /** Return pointer to data field in a hsm user request */
hur_data(struct hsm_user_request * hur)1142 static inline void *hur_data(struct hsm_user_request *hur)
1143 {
1144 	return &hur->hur_user_item[hur->hur_request.hr_itemcount];
1145 }
1146 
1147 /**
1148  * Compute the current length of the provided hsm_user_request.  This returns -1
1149  * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1150  *
1151  * return -1 on bounds check error.
1152  */
hur_len(struct hsm_user_request * hur)1153 static inline ssize_t hur_len(struct hsm_user_request *hur)
1154 {
1155 	__u64	size;
1156 
1157 	/* can't overflow a __u64 since hr_itemcount is only __u32 */
1158 	size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1159 		(__u64)hur->hur_request.hr_itemcount *
1160 		sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1161 
1162 	if (size != (ssize_t)size)
1163 		return -1;
1164 
1165 	return size;
1166 }
1167 
1168 /****** HSM RPCs to copytool *****/
1169 /* Message types the copytool may receive */
1170 enum hsm_message_type {
1171 	HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1172 };
1173 
1174 /* Actions the copytool may be instructed to take for a given action_item */
1175 enum hsm_copytool_action {
1176 	HSMA_NONE    = 10, /* no action */
1177 	HSMA_ARCHIVE = 20, /* arbitrary offset */
1178 	HSMA_RESTORE = 21,
1179 	HSMA_REMOVE  = 22,
1180 	HSMA_CANCEL  = 23
1181 };
1182 
hsm_copytool_action2name(enum hsm_copytool_action a)1183 static inline char *hsm_copytool_action2name(enum hsm_copytool_action  a)
1184 {
1185 	switch  (a) {
1186 	case HSMA_NONE:    return "NOOP";
1187 	case HSMA_ARCHIVE: return "ARCHIVE";
1188 	case HSMA_RESTORE: return "RESTORE";
1189 	case HSMA_REMOVE:  return "REMOVE";
1190 	case HSMA_CANCEL:  return "CANCEL";
1191 	default:	   return "UNKNOWN";
1192 	}
1193 }
1194 
1195 /* Copytool item action description */
1196 struct hsm_action_item {
1197 	__u32      hai_len;     /* valid size of this struct */
1198 	__u32      hai_action;  /* hsm_copytool_action, but use known size */
1199 	struct lu_fid hai_fid;     /* Lustre FID to operated on */
1200 	struct lu_fid hai_dfid;    /* fid used for data access */
1201 	struct hsm_extent hai_extent;  /* byte range to operate on */
1202 	__u64      hai_cookie;  /* action cookie from coordinator */
1203 	__u64      hai_gid;     /* grouplock id */
1204 	char       hai_data[0]; /* variable length */
1205 } __packed;
1206 
1207 /*
1208  * helper function which print in hexa the first bytes of
1209  * hai opaque field
1210  * \param hai [IN] record to print
1211  * \param buffer [OUT] output buffer
1212  * \param len [IN] max buffer len
1213  * \retval buffer
1214  */
hai_dump_data_field(struct hsm_action_item * hai,char * buffer,size_t len)1215 static inline char *hai_dump_data_field(struct hsm_action_item *hai,
1216 					char *buffer, size_t len)
1217 {
1218 	int i, data_len;
1219 	char *ptr;
1220 
1221 	ptr = buffer;
1222 	data_len = hai->hai_len - sizeof(*hai);
1223 	for (i = 0; (i < data_len) && (len > 2); i++) {
1224 		snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1225 		ptr += 2;
1226 		len -= 2;
1227 	}
1228 
1229 	*ptr = '\0';
1230 
1231 	return buffer;
1232 }
1233 
1234 /* Copytool action list */
1235 #define HAL_VERSION 1
1236 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1237 struct hsm_action_list {
1238 	__u32 hal_version;
1239 	__u32 hal_count;       /* number of hai's to follow */
1240 	__u64 hal_compound_id; /* returned by coordinator */
1241 	__u64 hal_flags;
1242 	__u32 hal_archive_id; /* which archive backend */
1243 	__u32 padding1;
1244 	char  hal_fsname[0];   /* null-terminated */
1245 	/* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1246 	 * boundaries. See hai_first
1247 	 */
1248 } __packed;
1249 
1250 #ifndef HAVE_CFS_SIZE_ROUND
cfs_size_round(int val)1251 static inline int cfs_size_round(int val)
1252 {
1253 	return (val + 7) & (~0x7);
1254 }
1255 
1256 #define HAVE_CFS_SIZE_ROUND
1257 #endif
1258 
1259 /* Return pointer to first hai in action list */
hai_first(struct hsm_action_list * hal)1260 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1261 {
1262 	return (struct hsm_action_item *)(hal->hal_fsname +
1263 					  cfs_size_round(strlen(hal-> \
1264 								hal_fsname)
1265 							 + 1));
1266 }
1267 
1268 /* Return pointer to next hai */
hai_next(struct hsm_action_item * hai)1269 static inline struct hsm_action_item *hai_next(struct hsm_action_item *hai)
1270 {
1271 	return (struct hsm_action_item *)((char *)hai +
1272 					  cfs_size_round(hai->hai_len));
1273 }
1274 
1275 /* Return size of an hsm_action_list */
hal_size(struct hsm_action_list * hal)1276 static inline int hal_size(struct hsm_action_list *hal)
1277 {
1278 	int i, sz;
1279 	struct hsm_action_item *hai;
1280 
1281 	sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1282 	hai = hai_first(hal);
1283 	for (i = 0; i < hal->hal_count; i++, hai = hai_next(hai))
1284 		sz += cfs_size_round(hai->hai_len);
1285 
1286 	return sz;
1287 }
1288 
1289 /* HSM file import
1290  * describe the attributes to be set on imported file
1291  */
1292 struct hsm_user_import {
1293 	__u64		hui_size;
1294 	__u64		hui_atime;
1295 	__u64		hui_mtime;
1296 	__u32		hui_atime_ns;
1297 	__u32		hui_mtime_ns;
1298 	__u32		hui_uid;
1299 	__u32		hui_gid;
1300 	__u32		hui_mode;
1301 	__u32		hui_archive_id;
1302 };
1303 
1304 /* Copytool progress reporting */
1305 #define HP_FLAG_COMPLETED 0x01
1306 #define HP_FLAG_RETRY     0x02
1307 
1308 struct hsm_progress {
1309 	struct lu_fid		hp_fid;
1310 	__u64			hp_cookie;
1311 	struct hsm_extent	hp_extent;
1312 	__u16			hp_flags;
1313 	__u16			hp_errval; /* positive val */
1314 	__u32			padding;
1315 };
1316 
1317 struct hsm_copy {
1318 	__u64			hc_data_version;
1319 	__u16			hc_flags;
1320 	__u16			hc_errval; /* positive val */
1321 	__u32			padding;
1322 	struct hsm_action_item	hc_hai;
1323 };
1324 
1325 /** @} lustreuser */
1326 
1327 #endif /* _LUSTRE_USER_H */
1328