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 <md_object.h>
51 #include <lustre_quota.h>
54 #include <sys/kstat.h>
55 #define kmem_zalloc(a, b) kzalloc(a, b)
56 #define kmem_free(ptr, sz) ((void)(sz), kfree(ptr))
58 #define KM_SLEEP GFP_KERNEL
63 #include <sys/nvpair.h>
65 #include <sys/zfs_znode.h>
68 #define LUSTRE_ROOT_FID_SEQ 0
69 #define DMU_OSD_SVNAME "svname"
70 #define DMU_OSD_OI_NAME_BASE "oi"
72 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
75 * Iterator's in-memory data structure for quota file.
78 struct osd_object *oiq_obj;
79 /* DMU accounting object id */
83 /** identifier for current quota record */
85 unsigned oiq_reset:1; /* 1 -- no need to advance */
89 * Iterator's in-memory data structure for ZAPs
91 * ZFS does not store . and .. on a disk, instead they are
92 * generated up on request
93 * to follow this format we do the same
97 struct osd_object *ozi_obj;
98 struct lustre_capa *ozi_capa;
99 unsigned ozi_reset:1; /* 1 -- no need to advance */
100 /* ozi_pos - position of the cursor:
101 * 0 - before any record
104 * 3 - real records */
107 char ozi_name[MAXNAMELEN]; /* file name for dir */
108 __u64 ozi_key; /* binary key for index files */
111 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
114 * regular ZFS direntry
116 struct zpl_direntry {
117 uint64_t zde_dnode:48,
120 } __attribute__((packed));
123 * lustre direntry adds a fid to regular ZFS direntry
125 struct luz_direntry {
126 struct zpl_direntry lzd_reg;
127 struct lu_fid lzd_fid;
128 } __attribute__((packed));
131 /* cached SA attributes */
145 struct osd_thread_info {
146 const struct lu_env *oti_env;
148 struct lu_fid oti_fid;
150 * XXX temporary: for ->i_op calls.
152 struct timespec oti_time;
154 * XXX temporary: for capa operations.
156 struct lustre_capa_key oti_capa_key;
157 struct lustre_capa oti_capa;
159 struct ost_id oti_ostid;
163 /** osd iterator context used for iterator session */
165 struct osd_zap_it oti_it_zap;
166 struct osd_it_quota oti_it_quota;
171 char oti_key[MAXNAMELEN + 1];
172 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
174 struct lustre_mdt_attrs oti_mdt_attrs;
176 struct lu_attr oti_la;
177 struct osa_attr oti_osa;
178 zap_attribute_t oti_za;
179 dmu_object_info_t oti_doi;
180 struct luz_direntry oti_zde;
182 struct lquota_id_info oti_qi;
183 struct lu_seq_range oti_seq_range;
186 extern struct lu_context_key osd_key;
188 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
190 return lu_context_key_get(&env->le_ctx, &osd_key);
194 struct thandle ot_super;
195 struct list_head ot_dcb_list;
196 struct list_head ot_sa_list;
197 struct semaphore ot_sa_lock;
199 struct lquota_trans ot_quota_trans;
200 __u32 ot_write_commit:1,
204 #define OSD_OI_NAME_SIZE 16
207 * Object Index (OI) instance.
210 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
215 uint64_t *os_compat_dirs;
216 int os_subdir_count; /* subdir count for each seq */
217 obd_seq os_seq; /* seq number */
218 struct list_head os_seq_list; /* list to seq_list */
221 struct osd_seq_list {
222 rwlock_t osl_seq_list_lock; /* lock for seq_list */
223 struct list_head osl_seq_list; /* list head for seq */
224 struct semaphore osl_seq_init_sem;
227 #define OSD_OST_MAP_SIZE 32
234 struct dt_device od_dt_dev;
235 /* information about underlying file system */
236 udmu_objset_t od_objset;
241 unsigned int od_fl_capa:1;
242 unsigned long od_capa_timeout;
244 struct lustre_capa_key *od_capa_keys;
245 struct hlist_head *od_capa_hash;
247 cfs_proc_dir_entry_t *od_proc_entry;
248 struct lprocfs_stats *od_stats;
252 struct osd_oi **od_oi_table;
253 unsigned int od_oi_count;
254 struct osd_seq_list od_seq_list;
256 unsigned int od_rdonly:1,
258 od_quota_iused_est:1,
265 struct lu_site od_site;
267 /* object IDs of the inode accounting indexes */
268 uint64_t od_iusr_oid;
269 uint64_t od_igrp_oid;
271 /* quota slave instance */
272 struct qsd_instance *od_quota_slave;
274 /* used to debug zerocopy logic: the fields track all
275 * allocated, loaned and referenced buffers in use.
276 * to be removed once the change is tested well. */
277 atomic_t od_zerocopy_alloc;
278 atomic_t od_zerocopy_loan;
279 atomic_t od_zerocopy_pin;
281 arc_prune_t *arc_prune_cb;
283 /* osd seq instance */
284 struct lu_client_seq *od_cl_seq;
288 struct dt_object oo_dt;
290 * Inode for file system object represented by this osd_object. This
291 * inode is pinned for the whole duration of lu_object life.
293 * Not modified concurrently (either setup early during object
294 * creation, or assigned by osd_object_create() under write lock).
297 sa_handle_t *oo_sa_hdl;
298 nvlist_t *oo_sa_xattr;
299 struct list_head oo_sa_linkage;
301 struct rw_semaphore oo_sem;
303 /* cached attributes */
304 rwlock_t oo_attr_lock;
305 struct lu_attr oo_attr;
307 /* protects extended attributes */
308 struct semaphore oo_guard;
311 /* record size for index file */
312 unsigned char oo_keysize;
313 unsigned char oo_recsize;
314 unsigned char oo_recusize; /* unit size */
317 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
318 extern const struct dt_index_operations osd_acct_index_ops;
319 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
320 extern struct lu_device_operations osd_lu_ops;
321 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
322 qid_t uid, qid_t gid, long long space,
323 struct osd_thandle *oh, bool is_blk, int *flags,
329 static inline int lu_device_is_osd(const struct lu_device *d)
331 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
334 static inline struct osd_object *osd_obj(const struct lu_object *o)
336 LASSERT(lu_device_is_osd(o->lo_dev));
337 return container_of0(o, struct osd_object, oo_dt.do_lu);
340 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
342 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
343 return container_of0(d, struct osd_device, od_dt_dev);
346 static inline struct osd_device *osd_dev(const struct lu_device *d)
348 LASSERT(lu_device_is_osd(d));
349 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
352 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
354 return osd_obj(&d->do_lu);
357 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
359 return osd_dev(o->oo_dt.do_lu.lo_dev);
362 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
364 return &osd->od_dt_dev.dd_lu_dev;
367 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
369 return osd_dev(o->do_lu.lo_dev)->od_objset.os;
372 static inline int osd_invariant(const struct osd_object *obj)
377 static inline int osd_object_invariant(const struct lu_object *l)
379 return osd_invariant(osd_obj(l));
382 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
384 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
387 static inline char *osd_name(struct osd_device *osd)
389 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
394 LPROC_OSD_READ_BYTES = 0,
395 LPROC_OSD_WRITE_BYTES = 1,
396 LPROC_OSD_GET_PAGE = 2,
397 LPROC_OSD_NO_PAGE = 3,
398 LPROC_OSD_CACHE_ACCESS = 4,
399 LPROC_OSD_CACHE_HIT = 5,
400 LPROC_OSD_CACHE_MISS = 6,
401 LPROC_OSD_COPY_IO = 7,
402 LPROC_OSD_ZEROCOPY_IO = 8,
403 LPROC_OSD_TAIL_IO = 9,
408 extern struct lprocfs_seq_vars lprocfs_osd_obd_vars[];
410 int osd_procfs_init(struct osd_device *osd, const char *name);
411 int osd_procfs_fini(struct osd_device *osd);
413 int udmu_zap_cursor_retrieve_key(const struct lu_env *env,
414 zap_cursor_t *zc, char *key, int max);
415 int udmu_zap_cursor_retrieve_value(const struct lu_env *env,
416 zap_cursor_t *zc, char *buf,
417 int buf_size, int *bytes_read);
420 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
421 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
422 uint64_t oid, dmu_buf_t **dbp, void *tag);
423 struct lu_object *osd_object_alloc(const struct lu_env *env,
424 const struct lu_object_header *hdr,
425 struct lu_device *d);
426 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
427 void *buf, uint32_t buflen, struct osd_thandle *oh);
428 int __osd_zap_create(const struct lu_env *env, udmu_objset_t *uos,
429 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
430 uint64_t parent, void *tag, zap_flags_t flags);
431 int __osd_object_create(const struct lu_env *env, udmu_objset_t *uos,
432 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
433 uint64_t parent, void *tag);
434 int __osd_object_free(udmu_objset_t *uos, uint64_t oid, dmu_tx_t *tx);
437 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
438 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
439 int osd_fid_lookup(const struct lu_env *env,
440 struct osd_device *, const struct lu_fid *, uint64_t *);
441 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
442 const struct lu_fid *fid, char *buf);
443 int osd_options_init(void);
444 int osd_convert_root_to_new_seq(const struct lu_env *env,
445 struct osd_device *o);
446 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
449 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
450 const struct dt_index_features *feat);
451 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
452 obd_seq seq, struct lu_seq_range *range);
455 int __osd_xattr_load(udmu_objset_t *uos, uint64_t dnode, nvlist_t **sa_xattr);
456 int __osd_xattr_get_large(const struct lu_env *env, udmu_objset_t *uos,
457 uint64_t xattr, struct lu_buf *buf,
458 const char *name, int *sizep);
459 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
460 struct lu_buf *buf, const char *name,
461 struct lustre_capa *capa);
462 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
463 const struct lu_buf *buf, const char *name,
464 int fl, struct thandle *handle);
465 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
466 const struct lu_buf *buf, const char *name, int fl,
467 struct thandle *handle, struct lustre_capa *capa);
468 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
469 const char *name, struct thandle *handle);
470 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
471 const char *name, struct thandle *handle,
472 struct lustre_capa *capa);
473 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
474 struct lu_buf *lb, struct lustre_capa *capa);
475 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
476 int vallen, const char *name, struct osd_thandle *oh);
477 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
478 const struct lu_buf *buf, const char *name, int fl,
479 struct osd_thandle *oh);;
480 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
481 const struct lu_buf *buf, const char *name, int fl,
482 struct osd_thandle *oh);
484 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
485 const struct lu_buf *buf, const char *name, int fl,
486 struct osd_thandle *oh, struct lustre_capa *capa)
490 if (osd_obj2dev(obj)->od_xattr_in_sa) {
491 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
493 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
495 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
501 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
503 return (((flags & FS_APPEND_FL) ? ZFS_APPENDONLY : 0) |
504 ((flags & FS_NODUMP_FL) ? ZFS_NODUMP : 0) |
505 ((flags & FS_IMMUTABLE_FL) ? ZFS_IMMUTABLE : 0));
508 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
510 return (((flags & ZFS_APPENDONLY) ? FS_APPEND_FL : 0) |
511 ((flags & ZFS_NODUMP) ? FS_NODUMP_FL : 0) |
512 ((flags & ZFS_IMMUTABLE) ? FS_IMMUTABLE_FL : 0));
517 #ifndef HAVE_DSL_POOL_CONFIG
518 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
522 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
528 #endif /* _OSD_INTERNAL_H */