4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
31 * Copyright (c) 2012, 2013, Intel Corporation.
32 * Use is subject to license terms.
35 * This file is part of Lustre, http://www.lustre.org/
36 * Lustre is a trademark of Sun Microsystems, Inc.
38 * lustre/osd-zfs/osd_internal.h
39 * Shared definitions and declarations for zfs/dmu osd
41 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
42 * Author: Mike Pershin <tappro@whamcloud.com>
43 * Author: Johann Lombardi <johann@whamcloud.com>
46 #ifndef _OSD_INTERNAL_H
47 #define _OSD_INTERNAL_H
49 #include <dt_object.h>
50 #include <lustre_quota.h>
53 #include <sys/nvpair.h>
55 #include <sys/zfs_znode.h>
58 #define LUSTRE_ROOT_FID_SEQ 0
59 #define DMU_OSD_SVNAME "svname"
60 #define DMU_OSD_OI_NAME_BASE "oi"
62 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
65 * Iterator's in-memory data structure for quota file.
68 struct osd_object *oiq_obj;
69 /* DMU accounting object id */
73 /** identifier for current quota record */
75 unsigned oiq_reset:1; /* 1 -- no need to advance */
79 * Iterator's in-memory data structure for ZAPs
81 * ZFS does not store . and .. on a disk, instead they are
82 * generated up on request
83 * to follow this format we do the same
87 struct osd_object *ozi_obj;
88 struct lustre_capa *ozi_capa;
89 unsigned ozi_reset:1; /* 1 -- no need to advance */
90 /* ozi_pos - position of the cursor:
91 * 0 - before any record
97 char ozi_name[NAME_MAX + 1]; /* file name for dir */
98 __u64 ozi_key; /* binary key for index files */
101 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
104 * regular ZFS direntry
106 struct zpl_direntry {
107 uint64_t zde_dnode:48,
110 } __attribute__((packed));
113 * lustre direntry adds a fid to regular ZFS direntry
115 struct luz_direntry {
116 struct zpl_direntry lzd_reg;
117 struct lu_fid lzd_fid;
118 } __attribute__((packed));
121 /* cached SA attributes */
135 struct osd_thread_info {
136 const struct lu_env *oti_env;
138 struct lu_fid oti_fid;
140 * XXX temporary: for ->i_op calls.
142 struct timespec oti_time;
144 * XXX temporary: for capa operations.
146 struct lustre_capa_key oti_capa_key;
147 struct lustre_capa oti_capa;
149 struct ost_id oti_ostid;
153 /** osd iterator context used for iterator session */
155 struct osd_zap_it oti_it_zap;
156 struct osd_it_quota oti_it_quota;
160 char oti_key[MAXNAMELEN + 1];
161 struct lustre_mdt_attrs oti_mdt_attrs;
163 struct lu_attr oti_la;
164 struct osa_attr oti_osa;
165 zap_attribute_t oti_za;
166 dmu_object_info_t oti_doi;
167 struct luz_direntry oti_zde;
169 struct lquota_id_info oti_qi;
170 struct lu_seq_range oti_seq_range;
173 extern struct lu_context_key osd_key;
175 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
177 return lu_context_key_get(&env->le_ctx, &osd_key);
181 struct thandle ot_super;
182 cfs_list_t ot_dcb_list;
183 cfs_list_t ot_sa_list;
184 struct semaphore ot_sa_lock;
186 struct lquota_trans ot_quota_trans;
187 __u32 ot_write_commit:1,
191 #define OSD_OI_NAME_SIZE 16
194 * Object Index (OI) instance.
197 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
202 uint64_t *os_compat_dirs;
203 int os_subdir_count; /* subdir count for each seq */
204 obd_seq os_seq; /* seq number */
205 cfs_list_t os_seq_list; /* list to seq_list */
208 struct osd_seq_list {
209 rwlock_t osl_seq_list_lock; /* lock for seq_list */
210 cfs_list_t osl_seq_list; /* list head for seq */
211 struct semaphore osl_seq_init_sem;
214 #define OSD_OST_MAP_SIZE 32
221 struct dt_device od_dt_dev;
222 /* information about underlying file system */
223 udmu_objset_t od_objset;
228 unsigned int od_fl_capa:1;
229 unsigned long od_capa_timeout;
231 struct lustre_capa_key *od_capa_keys;
232 cfs_hlist_head_t *od_capa_hash;
234 cfs_proc_dir_entry_t *od_proc_entry;
235 struct lprocfs_stats *od_stats;
238 struct osd_oi **od_oi_table;
239 unsigned int od_oi_count;
240 struct osd_seq_list od_seq_list;
242 unsigned int od_rdonly:1,
244 od_quota_iused_est:1;
249 struct lu_site od_site;
251 /* object IDs of the inode accounting indexes */
252 uint64_t od_iusr_oid;
253 uint64_t od_igrp_oid;
255 /* quota slave instance */
256 struct qsd_instance *od_quota_slave;
258 /* used to debug zerocopy logic: the fields track all
259 * allocated, loaned and referenced buffers in use.
260 * to be removed once the change is tested well. */
261 cfs_atomic_t od_zerocopy_alloc;
262 cfs_atomic_t od_zerocopy_loan;
263 cfs_atomic_t od_zerocopy_pin;
265 arc_prune_t *arc_prune_cb;
269 struct dt_object oo_dt;
271 * Inode for file system object represented by this osd_object. This
272 * inode is pinned for the whole duration of lu_object life.
274 * Not modified concurrently (either setup early during object
275 * creation, or assigned by osd_object_create() under write lock).
278 sa_handle_t *oo_sa_hdl;
279 nvlist_t *oo_sa_xattr;
280 cfs_list_t oo_sa_linkage;
282 struct rw_semaphore oo_sem;
284 /* cached attributes */
285 rwlock_t oo_attr_lock;
286 struct lu_attr oo_attr;
288 /* protects extended attributes */
289 struct semaphore oo_guard;
292 /* record size for index file */
296 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
297 extern const struct dt_index_operations osd_acct_index_ops;
298 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
299 extern struct lu_device_operations osd_lu_ops;
300 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
301 qid_t uid, qid_t gid, long long space,
302 struct osd_thandle *oh, bool is_blk, int *flags,
308 static inline int lu_device_is_osd(const struct lu_device *d)
310 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
313 static inline struct osd_object *osd_obj(const struct lu_object *o)
315 LASSERT(lu_device_is_osd(o->lo_dev));
316 return container_of0(o, struct osd_object, oo_dt.do_lu);
319 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
321 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
322 return container_of0(d, struct osd_device, od_dt_dev);
325 static inline struct osd_device *osd_dev(const struct lu_device *d)
327 LASSERT(lu_device_is_osd(d));
328 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
331 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
333 return osd_obj(&d->do_lu);
336 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
338 return osd_dev(o->oo_dt.do_lu.lo_dev);
341 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
343 return &osd->od_dt_dev.dd_lu_dev;
346 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
348 return osd_dev(o->do_lu.lo_dev)->od_objset.os;
351 static inline int osd_invariant(const struct osd_object *obj)
356 static inline int osd_object_invariant(const struct lu_object *l)
358 return osd_invariant(osd_obj(l));
361 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
363 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
366 static inline char *osd_name(struct osd_device *osd)
368 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
373 LPROC_OSD_READ_BYTES = 0,
374 LPROC_OSD_WRITE_BYTES = 1,
375 LPROC_OSD_GET_PAGE = 2,
376 LPROC_OSD_NO_PAGE = 3,
377 LPROC_OSD_CACHE_ACCESS = 4,
378 LPROC_OSD_CACHE_HIT = 5,
379 LPROC_OSD_CACHE_MISS = 6,
380 LPROC_OSD_COPY_IO = 7,
381 LPROC_OSD_ZEROCOPY_IO = 8,
382 LPROC_OSD_TAIL_IO = 9,
387 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
388 extern struct lprocfs_vars lprocfs_osd_module_vars[];
390 int osd_procfs_init(struct osd_device *osd, const char *name);
391 int osd_procfs_fini(struct osd_device *osd);
393 int udmu_zap_cursor_retrieve_key(const struct lu_env *env,
394 zap_cursor_t *zc, char *key, int max);
395 int udmu_zap_cursor_retrieve_value(const struct lu_env *env,
396 zap_cursor_t *zc, char *buf,
397 int buf_size, int *bytes_read);
400 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
401 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
402 uint64_t oid, dmu_buf_t **dbp, void *tag);
403 struct lu_object *osd_object_alloc(const struct lu_env *env,
404 const struct lu_object_header *hdr,
405 struct lu_device *d);
406 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
407 void *buf, uint32_t buflen, struct osd_thandle *oh);
408 int __osd_zap_create(const struct lu_env *env, udmu_objset_t *uos,
409 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
410 uint64_t parent, void *tag, zap_flags_t flags);
411 int __osd_object_create(const struct lu_env *env, udmu_objset_t *uos,
412 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
413 uint64_t parent, void *tag);
414 int __osd_object_free(udmu_objset_t *uos, uint64_t oid, dmu_tx_t *tx);
417 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
418 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
419 int osd_fid_lookup(const struct lu_env *env,
420 struct osd_device *, const struct lu_fid *, uint64_t *);
421 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
422 const struct lu_fid *fid, char *buf);
423 int osd_options_init(void);
424 int osd_convert_root_to_new_seq(const struct lu_env *env,
425 struct osd_device *o);
428 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
429 const struct dt_index_features *feat);
430 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
431 const struct lu_fid *fid, struct lu_seq_range *range);
434 int __osd_xattr_load(udmu_objset_t *uos, uint64_t dnode, nvlist_t **sa_xattr);
435 int __osd_xattr_get_large(const struct lu_env *env, udmu_objset_t *uos,
436 uint64_t xattr, struct lu_buf *buf,
437 const char *name, int *sizep);
438 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
439 struct lu_buf *buf, const char *name,
440 struct lustre_capa *capa);
441 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
442 const struct lu_buf *buf, const char *name,
443 int fl, struct thandle *handle);
444 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
445 const struct lu_buf *buf, const char *name, int fl,
446 struct thandle *handle, struct lustre_capa *capa);
447 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
448 const char *name, struct thandle *handle);
449 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
450 const char *name, struct thandle *handle,
451 struct lustre_capa *capa);
452 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
453 struct lu_buf *lb, struct lustre_capa *capa);
454 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
455 int vallen, const char *name, struct osd_thandle *oh);
456 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
457 const struct lu_buf *buf, const char *name, int fl,
458 struct osd_thandle *oh);;
459 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
460 const struct lu_buf *buf, const char *name, int fl,
461 struct osd_thandle *oh);
463 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
464 const struct lu_buf *buf, const char *name, int fl,
465 struct osd_thandle *oh, struct lustre_capa *capa)
469 if (osd_obj2dev(obj)->od_xattr_in_sa) {
470 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
472 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
474 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
480 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
482 return (((flags & FS_APPEND_FL) ? ZFS_APPENDONLY : 0) |
483 ((flags & FS_NODUMP_FL) ? ZFS_NODUMP : 0) |
484 ((flags & FS_IMMUTABLE_FL) ? ZFS_IMMUTABLE : 0));
487 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
489 return (((flags & ZFS_APPENDONLY) ? FS_APPEND_FL : 0) |
490 ((flags & ZFS_NODUMP) ? FS_NODUMP_FL : 0) |
491 ((flags & ZFS_IMMUTABLE) ? FS_IMMUTABLE_FL : 0));
495 #endif /* _OSD_INTERNAL_H */