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 /* the hash where object accounting is cached */
87 unsigned oiq_reset:1; /* 1 -- no need to advance */
91 * Iterator's in-memory data structure for ZAPs
93 * ZFS does not store . and .. on a disk, instead they are
94 * generated up on request
95 * to follow this format we do the same
99 struct osd_object *ozi_obj;
100 struct lustre_capa *ozi_capa;
101 unsigned ozi_reset:1; /* 1 -- no need to advance */
102 /* ozi_pos - position of the cursor:
103 * 0 - before any record
106 * 3 - real records */
109 char ozi_name[MAXNAMELEN]; /* file name for dir */
110 __u64 ozi_key; /* binary key for index files */
113 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
116 * regular ZFS direntry
118 struct zpl_direntry {
119 uint64_t zde_dnode:48,
122 } __attribute__((packed));
125 * lustre direntry adds a fid to regular ZFS direntry
127 struct luz_direntry {
128 struct zpl_direntry lzd_reg;
129 struct lu_fid lzd_fid;
130 } __attribute__((packed));
133 /* cached SA attributes */
147 struct osd_thread_info {
148 const struct lu_env *oti_env;
150 struct lu_fid oti_fid;
152 * XXX temporary: for ->i_op calls.
154 struct timespec oti_time;
156 * XXX temporary: for capa operations.
158 struct lustre_capa_key oti_capa_key;
159 struct lustre_capa oti_capa;
161 struct ost_id oti_ostid;
165 /** osd iterator context used for iterator session */
167 struct osd_zap_it oti_it_zap;
168 struct osd_it_quota oti_it_quota;
173 char oti_key[MAXNAMELEN + 1];
174 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
176 struct lustre_mdt_attrs oti_mdt_attrs;
178 struct lu_attr oti_la;
179 struct osa_attr oti_osa;
180 zap_attribute_t oti_za;
181 dmu_object_info_t oti_doi;
182 struct luz_direntry oti_zde;
184 struct lquota_id_info oti_qi;
185 struct lu_seq_range oti_seq_range;
188 extern struct lu_context_key osd_key;
190 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
192 return lu_context_key_get(&env->le_ctx, &osd_key);
196 struct thandle ot_super;
197 cfs_list_t ot_dcb_list;
198 cfs_list_t ot_sa_list;
199 struct semaphore ot_sa_lock;
201 struct lquota_trans ot_quota_trans;
202 __u32 ot_write_commit:1,
206 #define OSD_OI_NAME_SIZE 16
209 * Object Index (OI) instance.
212 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
217 uint64_t *os_compat_dirs;
218 int os_subdir_count; /* subdir count for each seq */
219 obd_seq os_seq; /* seq number */
220 cfs_list_t os_seq_list; /* list to seq_list */
223 struct osd_seq_list {
224 rwlock_t osl_seq_list_lock; /* lock for seq_list */
225 cfs_list_t osl_seq_list; /* list head for seq */
226 struct semaphore osl_seq_init_sem;
229 #define OSD_OST_MAP_SIZE 32
232 * this structure tracks changes made to the accounting within specific TXG
234 struct osd_zfs_acct_txg {
238 struct osd_device *zat_osd;
246 struct dt_device od_dt_dev;
247 /* information about underlying file system */
248 udmu_objset_t od_objset;
253 unsigned int od_fl_capa:1;
254 unsigned long od_capa_timeout;
256 struct lustre_capa_key *od_capa_keys;
257 cfs_hlist_head_t *od_capa_hash;
259 cfs_proc_dir_entry_t *od_proc_entry;
260 struct lprocfs_stats *od_stats;
264 struct osd_oi **od_oi_table;
265 unsigned int od_oi_count;
266 struct osd_seq_list od_seq_list;
268 unsigned int od_rdonly:1,
270 od_quota_iused_est:1,
277 struct lu_site od_site;
279 /* object IDs of the inode accounting indexes */
280 uint64_t od_iusr_oid;
281 uint64_t od_igrp_oid;
283 /* quota slave instance */
284 struct qsd_instance *od_quota_slave;
286 /* used to debug zerocopy logic: the fields track all
287 * allocated, loaned and referenced buffers in use.
288 * to be removed once the change is tested well. */
289 atomic_t od_zerocopy_alloc;
290 atomic_t od_zerocopy_loan;
291 atomic_t od_zerocopy_pin;
293 arc_prune_t *arc_prune_cb;
295 /* quota: object accounting */
296 spinlock_t od_known_txg_lock;
297 uint64_t od_known_txg;
298 struct osd_zfs_acct_txg *od_acct_delta;
299 cfs_hash_t *od_acct_usr;
300 cfs_hash_t *od_acct_grp;
302 /* osd seq instance */
303 struct lu_client_seq *od_cl_seq;
307 struct dt_object oo_dt;
309 * Inode for file system object represented by this osd_object. This
310 * inode is pinned for the whole duration of lu_object life.
312 * Not modified concurrently (either setup early during object
313 * creation, or assigned by osd_object_create() under write lock).
316 sa_handle_t *oo_sa_hdl;
317 nvlist_t *oo_sa_xattr;
318 cfs_list_t oo_sa_linkage;
320 struct rw_semaphore oo_sem;
322 /* cached attributes */
323 rwlock_t oo_attr_lock;
324 struct lu_attr oo_attr;
326 /* protects extended attributes */
327 struct semaphore oo_guard;
330 /* record size for index file */
331 unsigned char oo_keysize;
332 unsigned char oo_recsize;
333 unsigned char oo_recusize; /* unit size */
336 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
337 extern const struct dt_index_operations osd_acct_index_ops;
338 extern struct lu_device_operations osd_lu_ops;
339 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
340 qid_t uid, qid_t gid, long long space,
341 struct osd_thandle *oh, bool is_blk, int *flags,
347 static inline int lu_device_is_osd(const struct lu_device *d)
349 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
352 static inline struct osd_object *osd_obj(const struct lu_object *o)
354 LASSERT(lu_device_is_osd(o->lo_dev));
355 return container_of0(o, struct osd_object, oo_dt.do_lu);
358 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
360 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
361 return container_of0(d, struct osd_device, od_dt_dev);
364 static inline struct osd_device *osd_dev(const struct lu_device *d)
366 LASSERT(lu_device_is_osd(d));
367 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
370 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
372 return osd_obj(&d->do_lu);
375 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
377 return osd_dev(o->oo_dt.do_lu.lo_dev);
380 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
382 return &osd->od_dt_dev.dd_lu_dev;
385 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
387 return osd_dev(o->do_lu.lo_dev)->od_objset.os;
390 static inline int osd_invariant(const struct osd_object *obj)
395 static inline int osd_object_invariant(const struct lu_object *l)
397 return osd_invariant(osd_obj(l));
400 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
402 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
405 static inline char *osd_name(struct osd_device *osd)
407 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
412 LPROC_OSD_READ_BYTES = 0,
413 LPROC_OSD_WRITE_BYTES = 1,
414 LPROC_OSD_GET_PAGE = 2,
415 LPROC_OSD_NO_PAGE = 3,
416 LPROC_OSD_CACHE_ACCESS = 4,
417 LPROC_OSD_CACHE_HIT = 5,
418 LPROC_OSD_CACHE_MISS = 6,
419 LPROC_OSD_COPY_IO = 7,
420 LPROC_OSD_ZEROCOPY_IO = 8,
421 LPROC_OSD_TAIL_IO = 9,
426 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
427 extern struct lprocfs_vars lprocfs_osd_module_vars[];
429 int osd_procfs_init(struct osd_device *osd, const char *name);
430 int osd_procfs_fini(struct osd_device *osd);
432 int udmu_zap_cursor_retrieve_key(const struct lu_env *env,
433 zap_cursor_t *zc, char *key, int max);
434 int udmu_zap_cursor_retrieve_value(const struct lu_env *env,
435 zap_cursor_t *zc, char *buf,
436 int buf_size, int *bytes_read);
439 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
440 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
441 uint64_t oid, dmu_buf_t **dbp, void *tag);
442 struct lu_object *osd_object_alloc(const struct lu_env *env,
443 const struct lu_object_header *hdr,
444 struct lu_device *d);
445 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
446 void *buf, uint32_t buflen, struct osd_thandle *oh);
447 int __osd_zap_create(const struct lu_env *env, udmu_objset_t *uos,
448 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
449 uint64_t parent, void *tag, zap_flags_t flags);
450 int __osd_object_create(const struct lu_env *env, udmu_objset_t *uos,
451 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
452 uint64_t parent, void *tag);
453 int __osd_object_free(udmu_objset_t *uos, uint64_t oid, dmu_tx_t *tx);
456 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
457 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
458 int osd_fid_lookup(const struct lu_env *env,
459 struct osd_device *, const struct lu_fid *, uint64_t *);
460 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
461 const struct lu_fid *fid, char *buf);
462 int osd_options_init(void);
463 int osd_convert_root_to_new_seq(const struct lu_env *env,
464 struct osd_device *o);
465 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
468 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
469 const struct dt_index_features *feat);
470 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
471 obd_seq seq, struct lu_seq_range *range);
474 int __osd_xattr_load(udmu_objset_t *uos, uint64_t dnode, nvlist_t **sa_xattr);
475 int __osd_xattr_get_large(const struct lu_env *env, udmu_objset_t *uos,
476 uint64_t xattr, struct lu_buf *buf,
477 const char *name, int *sizep);
478 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
479 struct lu_buf *buf, const char *name,
480 struct lustre_capa *capa);
481 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
482 const struct lu_buf *buf, const char *name,
483 int fl, struct thandle *handle);
484 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
485 const struct lu_buf *buf, const char *name, int fl,
486 struct thandle *handle, struct lustre_capa *capa);
487 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
488 const char *name, struct thandle *handle);
489 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
490 const char *name, struct thandle *handle,
491 struct lustre_capa *capa);
492 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
493 struct lu_buf *lb, struct lustre_capa *capa);
494 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
495 int vallen, const char *name, struct osd_thandle *oh);
496 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
497 const struct lu_buf *buf, const char *name, int fl,
498 struct osd_thandle *oh);;
499 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
500 const struct lu_buf *buf, const char *name, int fl,
501 struct osd_thandle *oh);
503 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
504 const struct lu_buf *buf, const char *name, int fl,
505 struct osd_thandle *oh, struct lustre_capa *capa)
509 if (osd_obj2dev(obj)->od_xattr_in_sa) {
510 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
512 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
514 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
520 void osd_zfs_acct_uid(const struct lu_env *env, struct osd_device *osd,
521 __u64 uid, int delta, struct osd_thandle *oh);
522 void osd_zfs_acct_gid(const struct lu_env *env, struct osd_device *osd,
523 __u64 gid, int delta, struct osd_thandle *oh);
524 int osd_zfs_acct_init(const struct lu_env *env, struct osd_device *osd);
525 void osd_zfs_acct_fini(const struct lu_env *env, struct osd_device *osd);
527 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
529 return (((flags & FS_APPEND_FL) ? ZFS_APPENDONLY : 0) |
530 ((flags & FS_NODUMP_FL) ? ZFS_NODUMP : 0) |
531 ((flags & FS_IMMUTABLE_FL) ? ZFS_IMMUTABLE : 0));
534 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
536 return (((flags & ZFS_APPENDONLY) ? FS_APPEND_FL : 0) |
537 ((flags & ZFS_NODUMP) ? FS_NODUMP_FL : 0) |
538 ((flags & ZFS_IMMUTABLE) ? FS_IMMUTABLE_FL : 0));
543 #ifndef HAVE_DSL_POOL_CONFIG
544 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
548 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
554 #endif /* _OSD_INTERNAL_H */