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.
30 * Copyright (c) 2012, 2014, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/osd-zfs/osd_internal.h
37 * Shared definitions and declarations for zfs/dmu osd
39 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
40 * Author: Mike Pershin <tappro@whamcloud.com>
41 * Author: Johann Lombardi <johann@whamcloud.com>
44 #ifndef _OSD_INTERNAL_H
45 #define _OSD_INTERNAL_H
47 #include <dt_object.h>
48 #include <md_object.h>
49 #include <lustre_quota.h>
52 #include <sys/kstat.h>
53 #define kmem_zalloc(a, b) kzalloc(a, b)
54 #define kmem_free(ptr, sz) ((void)(sz), kfree(ptr))
56 #define KM_SLEEP GFP_KERNEL
60 #include <sys/nvpair.h>
61 #include <sys/zfs_znode.h>
64 #define LUSTRE_ROOT_FID_SEQ 0
65 #define DMU_OSD_SVNAME "svname"
66 #define DMU_OSD_OI_NAME_BASE "oi"
68 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
70 /* Statfs space reservation for grant, fragmentation, and unlink space. */
71 #define OSD_STATFS_RESERVED_BLKS (1ULL << (22 - SPA_MAXBLOCKSHIFT)) /* 4MB */
72 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
74 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
75 #define OSD_DNODE_MIN_BLKSHIFT (SPA_MAXBLOCKSHIFT - DNODE_SHIFT) /* 17-9 =8 */
76 #define OSD_DNODE_EST_BLKSHIFT (SPA_MAXBLOCKSHIFT - 12) /* 17-12=5 */
77 #define OSD_DNODE_EST_COUNT 1024
79 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
82 * Iterator's in-memory data structure for quota file.
85 struct osd_object *oiq_obj;
86 /* DMU accounting object id */
90 /** identifier for current quota record */
92 unsigned oiq_reset:1; /* 1 -- no need to advance */
96 * Iterator's in-memory data structure for ZAPs
98 * ZFS does not store . and .. on a disk, instead they are
99 * generated up on request
100 * to follow this format we do the same
103 zap_cursor_t *ozi_zc;
104 struct osd_object *ozi_obj;
105 struct lustre_capa *ozi_capa;
106 unsigned ozi_reset:1; /* 1 -- no need to advance */
107 /* ozi_pos - position of the cursor:
108 * 0 - before any record
111 * 3 - real records */
114 char ozi_name[MAXNAMELEN]; /* file name for dir */
115 __u64 ozi_key; /* binary key for index files */
118 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
121 * regular ZFS direntry
123 struct zpl_direntry {
124 uint64_t zde_dnode:48,
127 } __attribute__((packed));
130 * lustre direntry adds a fid to regular ZFS direntry
132 struct luz_direntry {
133 struct zpl_direntry lzd_reg;
134 struct lu_fid lzd_fid;
135 } __attribute__((packed));
138 /* cached SA attributes */
152 struct osd_thread_info {
153 const struct lu_env *oti_env;
155 struct lu_fid oti_fid;
157 * XXX temporary: for ->i_op calls.
159 struct timespec oti_time;
161 * XXX temporary: for capa operations.
163 struct lustre_capa_key oti_capa_key;
164 struct lustre_capa oti_capa;
166 struct ost_id oti_ostid;
170 /** osd iterator context used for iterator session */
172 struct osd_zap_it oti_it_zap;
173 struct osd_it_quota oti_it_quota;
178 char oti_key[MAXNAMELEN + 1];
179 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
181 struct lustre_mdt_attrs oti_mdt_attrs;
183 struct lu_attr oti_la;
184 struct osa_attr oti_osa;
185 zap_attribute_t oti_za;
186 dmu_object_info_t oti_doi;
187 struct luz_direntry oti_zde;
189 struct lquota_id_info oti_qi;
190 struct lu_seq_range oti_seq_range;
191 unsigned int oti_it_inline:1;
194 extern struct lu_context_key osd_key;
196 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
198 return lu_context_key_get(&env->le_ctx, &osd_key);
202 struct thandle ot_super;
203 struct list_head ot_dcb_list;
204 struct list_head ot_sa_list;
205 struct semaphore ot_sa_lock;
207 struct lquota_trans ot_quota_trans;
208 __u32 ot_write_commit:1,
212 #define OSD_OI_NAME_SIZE 16
215 * Object Index (OI) instance.
218 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
223 uint64_t *os_compat_dirs;
224 int os_subdir_count; /* subdir count for each seq */
225 u64 os_seq; /* seq number */
226 struct list_head os_seq_list; /* list to seq_list */
229 struct osd_seq_list {
230 rwlock_t osl_seq_list_lock; /* lock for seq_list */
231 struct list_head osl_seq_list; /* list head for seq */
232 struct semaphore osl_seq_init_sem;
235 #define OSD_OST_MAP_SIZE 32
242 struct dt_device od_dt_dev;
243 /* information about underlying file system */
244 struct objset *od_os;
245 uint64_t od_rootid; /* id of root znode */
246 /* SA attr mapping->id,
247 * name is the same as in ZFS to use defines SA_ZPL_...*/
248 sa_attr_type_t *z_attr_table;
253 unsigned int od_fl_capa:1;
254 unsigned long od_capa_timeout;
256 struct lustre_capa_key *od_capa_keys;
257 struct hlist_head *od_capa_hash;
259 struct proc_dir_entry *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,
278 struct lu_site od_site;
280 /* object IDs of the inode accounting indexes */
281 uint64_t od_iusr_oid;
282 uint64_t od_igrp_oid;
284 /* quota slave instance */
285 struct qsd_instance *od_quota_slave;
287 /* used to debug zerocopy logic: the fields track all
288 * allocated, loaned and referenced buffers in use.
289 * to be removed once the change is tested well. */
290 atomic_t od_zerocopy_alloc;
291 atomic_t od_zerocopy_loan;
292 atomic_t od_zerocopy_pin;
294 arc_prune_t *arc_prune_cb;
296 /* osd seq instance */
297 struct lu_client_seq *od_cl_seq;
301 struct dt_object oo_dt;
303 * Inode for file system object represented by this osd_object. This
304 * inode is pinned for the whole duration of lu_object life.
306 * Not modified concurrently (either setup early during object
307 * creation, or assigned by osd_object_create() under write lock).
310 sa_handle_t *oo_sa_hdl;
311 nvlist_t *oo_sa_xattr;
312 struct list_head oo_sa_linkage;
314 struct rw_semaphore oo_sem;
316 /* cached attributes */
317 rwlock_t oo_attr_lock;
318 struct lu_attr oo_attr;
320 /* protects extended attributes */
321 struct semaphore oo_guard;
324 /* record size for index file */
325 unsigned char oo_keysize;
326 unsigned char oo_recsize;
327 unsigned char oo_recusize; /* unit size */
330 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
331 extern const struct dt_index_operations osd_acct_index_ops;
332 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
333 extern struct lu_device_operations osd_lu_ops;
334 extern struct dt_index_operations osd_dir_ops;
335 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
336 qid_t uid, qid_t gid, long long space,
337 struct osd_thandle *oh, bool is_blk, int *flags,
339 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
345 static inline int lu_device_is_osd(const struct lu_device *d)
347 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
350 static inline struct osd_object *osd_obj(const struct lu_object *o)
352 LASSERT(lu_device_is_osd(o->lo_dev));
353 return container_of0(o, struct osd_object, oo_dt.do_lu);
356 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
358 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
359 return container_of0(d, struct osd_device, od_dt_dev);
362 static inline struct osd_device *osd_dev(const struct lu_device *d)
364 LASSERT(lu_device_is_osd(d));
365 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
368 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
370 return osd_obj(&d->do_lu);
373 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
375 return osd_dev(o->oo_dt.do_lu.lo_dev);
378 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
380 return &osd->od_dt_dev.dd_lu_dev;
383 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
385 return osd_dev(o->do_lu.lo_dev)->od_os;
388 static inline int osd_invariant(const struct osd_object *obj)
393 static inline int osd_object_invariant(const struct lu_object *l)
395 return osd_invariant(osd_obj(l));
398 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
400 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
403 static inline char *osd_name(struct osd_device *osd)
405 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
410 LPROC_OSD_READ_BYTES = 0,
411 LPROC_OSD_WRITE_BYTES = 1,
412 LPROC_OSD_GET_PAGE = 2,
413 LPROC_OSD_NO_PAGE = 3,
414 LPROC_OSD_CACHE_ACCESS = 4,
415 LPROC_OSD_CACHE_HIT = 5,
416 LPROC_OSD_CACHE_MISS = 6,
417 LPROC_OSD_COPY_IO = 7,
418 LPROC_OSD_ZEROCOPY_IO = 8,
419 LPROC_OSD_TAIL_IO = 9,
424 extern struct lprocfs_seq_vars lprocfs_osd_obd_vars[];
426 int osd_procfs_init(struct osd_device *osd, const char *name);
427 int osd_procfs_fini(struct osd_device *osd);
430 extern char *osd_obj_tag;
431 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
432 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
433 uint64_t oid, dmu_buf_t **dbp);
434 struct lu_object *osd_object_alloc(const struct lu_env *env,
435 const struct lu_object_header *hdr,
436 struct lu_device *d);
437 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
438 void *buf, uint32_t buflen, struct osd_thandle *oh);
439 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
440 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
441 uint64_t parent, zap_flags_t flags);
442 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
443 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
447 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
448 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
449 int osd_fid_lookup(const struct lu_env *env,
450 struct osd_device *, const struct lu_fid *, uint64_t *);
451 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
452 const struct lu_fid *fid, char *buf);
453 int osd_options_init(void);
454 int osd_convert_root_to_new_seq(const struct lu_env *env,
455 struct osd_device *o);
456 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
459 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
460 const struct dt_index_features *feat);
461 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
462 u64 seq, struct lu_seq_range *range);
463 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
464 uint64_t id, uint64_t dirhash);
465 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
466 uint64_t id, uint64_t dirhash);
467 void osd_zap_cursor_fini(zap_cursor_t *zc);
468 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
471 int __osd_xattr_load(struct osd_device *osd, uint64_t dnode,
472 nvlist_t **sa_xattr);
473 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
474 uint64_t xattr, struct lu_buf *buf,
475 const char *name, int *sizep);
476 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
477 struct lu_buf *buf, const char *name,
478 struct lustre_capa *capa);
479 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
480 const struct lu_buf *buf, const char *name,
481 int fl, struct thandle *handle);
482 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
483 const struct lu_buf *buf, const char *name, int fl,
484 struct thandle *handle, struct lustre_capa *capa);
485 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
486 const char *name, struct thandle *handle);
487 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
488 const char *name, struct thandle *handle,
489 struct lustre_capa *capa);
490 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
491 const struct lu_buf *lb, struct lustre_capa *capa);
492 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
493 int vallen, const char *name, struct osd_thandle *oh);
494 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
495 const struct lu_buf *buf, const char *name, int fl,
496 struct osd_thandle *oh);;
497 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
498 const struct lu_buf *buf, const char *name, int fl,
499 struct osd_thandle *oh);
501 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
502 const struct lu_buf *buf, const char *name, int fl,
503 struct osd_thandle *oh, struct lustre_capa *capa)
507 if (osd_obj2dev(obj)->od_xattr_in_sa) {
508 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
510 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
512 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
518 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
520 return (((flags & FS_APPEND_FL) ? ZFS_APPENDONLY : 0) |
521 ((flags & FS_NODUMP_FL) ? ZFS_NODUMP : 0) |
522 ((flags & FS_IMMUTABLE_FL) ? ZFS_IMMUTABLE : 0));
525 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
527 return (((flags & ZFS_APPENDONLY) ? FS_APPEND_FL : 0) |
528 ((flags & ZFS_NODUMP) ? FS_NODUMP_FL : 0) |
529 ((flags & ZFS_IMMUTABLE) ? FS_IMMUTABLE_FL : 0));
534 #ifndef HAVE_DSL_POOL_CONFIG
535 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
539 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
545 #endif /* _OSD_INTERNAL_H */