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 /* SPL redefines this but to the same value: ~0UL vs -1, but GCC complains.
53 * fixed in SPL master 52479ecf58fa89190e384edcf838fecccc786af5 */
55 #include <sys/kstat.h>
56 #define kmem_zalloc(a, b) kzalloc(a, b)
57 #define kmem_free(ptr, sz) ((void)(sz), kfree(ptr))
59 #define KM_SLEEP GFP_KERNEL
63 #include <sys/nvpair.h>
64 #include <sys/zfs_znode.h>
67 #define LUSTRE_ROOT_FID_SEQ 0
68 #define DMU_OSD_SVNAME "svname"
69 #define DMU_OSD_OI_NAME_BASE "oi"
71 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
73 /* Statfs space reservation for grant, fragmentation, and unlink space. */
74 #define OSD_STATFS_RESERVED_BLKS (1ULL << (22 - SPA_MAXBLOCKSHIFT)) /* 4MB */
75 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
77 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
78 #define OSD_DNODE_MIN_BLKSHIFT (SPA_MAXBLOCKSHIFT - DNODE_SHIFT) /* 17-9 =8 */
79 #define OSD_DNODE_EST_BLKSHIFT (SPA_MAXBLOCKSHIFT - 12) /* 17-12=5 */
80 #define OSD_DNODE_EST_COUNT 1024
82 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
85 * Iterator's in-memory data structure for quota file.
88 struct osd_object *oiq_obj;
89 /* DMU accounting object id */
93 /** identifier for current quota record */
95 unsigned oiq_reset:1; /* 1 -- no need to advance */
99 * Iterator's in-memory data structure for ZAPs
101 * ZFS does not store . and .. on a disk, instead they are
102 * generated up on request
103 * to follow this format we do the same
106 zap_cursor_t *ozi_zc;
107 struct osd_object *ozi_obj;
108 struct lustre_capa *ozi_capa;
109 unsigned ozi_reset:1; /* 1 -- no need to advance */
110 /* ozi_pos - position of the cursor:
111 * 0 - before any record
114 * 3 - real records */
117 char ozi_name[MAXNAMELEN]; /* file name for dir */
118 __u64 ozi_key; /* binary key for index files */
121 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
124 * regular ZFS direntry
126 struct zpl_direntry {
127 uint64_t zde_dnode:48,
130 } __attribute__((packed));
133 * lustre direntry adds a fid to regular ZFS direntry
135 struct luz_direntry {
136 struct zpl_direntry lzd_reg;
137 struct lu_fid lzd_fid;
138 } __attribute__((packed));
141 /* cached SA attributes */
155 struct osd_thread_info {
156 const struct lu_env *oti_env;
158 struct lu_fid oti_fid;
160 * XXX temporary: for ->i_op calls.
162 struct timespec oti_time;
164 * XXX temporary: for capa operations.
166 struct lustre_capa_key oti_capa_key;
167 struct lustre_capa oti_capa;
169 struct ost_id oti_ostid;
173 /** osd iterator context used for iterator session */
175 struct osd_zap_it oti_it_zap;
176 struct osd_it_quota oti_it_quota;
181 char oti_key[MAXNAMELEN + 1];
182 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
184 struct lustre_mdt_attrs oti_mdt_attrs;
186 struct lu_attr oti_la;
187 struct osa_attr oti_osa;
188 zap_attribute_t oti_za;
189 dmu_object_info_t oti_doi;
190 struct luz_direntry oti_zde;
192 struct lquota_id_info oti_qi;
193 struct lu_seq_range oti_seq_range;
194 unsigned int oti_it_inline:1;
197 extern struct lu_context_key osd_key;
199 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
201 return lu_context_key_get(&env->le_ctx, &osd_key);
205 struct thandle ot_super;
206 struct list_head ot_dcb_list;
207 struct list_head ot_sa_list;
208 struct semaphore ot_sa_lock;
210 struct lquota_trans ot_quota_trans;
211 __u32 ot_write_commit:1,
215 #define OSD_OI_NAME_SIZE 16
218 * Object Index (OI) instance.
221 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
226 uint64_t *os_compat_dirs;
227 int os_subdir_count; /* subdir count for each seq */
228 u64 os_seq; /* seq number */
229 struct list_head os_seq_list; /* list to seq_list */
232 struct osd_seq_list {
233 rwlock_t osl_seq_list_lock; /* lock for seq_list */
234 struct list_head osl_seq_list; /* list head for seq */
235 struct semaphore osl_seq_init_sem;
238 #define OSD_OST_MAP_SIZE 32
245 struct dt_device od_dt_dev;
246 /* information about underlying file system */
247 struct objset *od_os;
248 uint64_t od_rootid; /* id of root znode */
249 /* SA attr mapping->id,
250 * name is the same as in ZFS to use defines SA_ZPL_...*/
251 sa_attr_type_t *z_attr_table;
256 unsigned int od_fl_capa:1;
257 unsigned long od_capa_timeout;
259 struct lustre_capa_key *od_capa_keys;
260 struct hlist_head *od_capa_hash;
262 struct proc_dir_entry *od_proc_entry;
263 struct lprocfs_stats *od_stats;
267 struct osd_oi **od_oi_table;
268 unsigned int od_oi_count;
269 struct osd_seq_list od_seq_list;
271 unsigned int od_rdonly:1,
273 od_quota_iused_est:1,
281 struct lu_site od_site;
283 /* object IDs of the inode accounting indexes */
284 uint64_t od_iusr_oid;
285 uint64_t od_igrp_oid;
287 /* quota slave instance */
288 struct qsd_instance *od_quota_slave;
290 struct brw_stats od_brw_stats;
291 atomic_t od_r_in_flight;
292 atomic_t od_w_in_flight;
294 /* used to debug zerocopy logic: the fields track all
295 * allocated, loaned and referenced buffers in use.
296 * to be removed once the change is tested well. */
297 atomic_t od_zerocopy_alloc;
298 atomic_t od_zerocopy_loan;
299 atomic_t od_zerocopy_pin;
301 arc_prune_t *arc_prune_cb;
303 /* osd seq instance */
304 struct lu_client_seq *od_cl_seq;
308 struct dt_object oo_dt;
310 * Inode for file system object represented by this osd_object. This
311 * inode is pinned for the whole duration of lu_object life.
313 * Not modified concurrently (either setup early during object
314 * creation, or assigned by osd_object_create() under write lock).
317 sa_handle_t *oo_sa_hdl;
318 nvlist_t *oo_sa_xattr;
319 struct list_head oo_sa_linkage;
321 struct rw_semaphore oo_sem;
323 /* cached attributes */
324 rwlock_t oo_attr_lock;
325 struct lu_attr oo_attr;
327 /* protects extended attributes */
328 struct semaphore oo_guard;
331 /* record size for index file */
332 unsigned char oo_keysize;
333 unsigned char oo_recsize;
334 unsigned char oo_recusize; /* unit size */
337 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
338 extern const struct dt_index_operations osd_acct_index_ops;
339 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
340 extern struct lu_device_operations osd_lu_ops;
341 extern struct dt_index_operations osd_dir_ops;
342 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
343 qid_t uid, qid_t gid, long long space,
344 struct osd_thandle *oh, bool is_blk, int *flags,
346 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
352 static inline int lu_device_is_osd(const struct lu_device *d)
354 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
357 static inline struct osd_object *osd_obj(const struct lu_object *o)
359 LASSERT(lu_device_is_osd(o->lo_dev));
360 return container_of0(o, struct osd_object, oo_dt.do_lu);
363 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
365 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
366 return container_of0(d, struct osd_device, od_dt_dev);
369 static inline struct osd_device *osd_dev(const struct lu_device *d)
371 LASSERT(lu_device_is_osd(d));
372 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
375 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
377 return osd_obj(&d->do_lu);
380 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
382 return osd_dev(o->oo_dt.do_lu.lo_dev);
385 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
387 return &osd->od_dt_dev.dd_lu_dev;
390 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
392 return osd_dev(o->do_lu.lo_dev)->od_os;
395 static inline int osd_invariant(const struct osd_object *obj)
400 static inline int osd_object_invariant(const struct lu_object *l)
402 return osd_invariant(osd_obj(l));
405 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
407 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
410 static inline char *osd_name(struct osd_device *osd)
412 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
417 LPROC_OSD_READ_BYTES = 0,
418 LPROC_OSD_WRITE_BYTES = 1,
419 LPROC_OSD_GET_PAGE = 2,
420 LPROC_OSD_NO_PAGE = 3,
421 LPROC_OSD_CACHE_ACCESS = 4,
422 LPROC_OSD_CACHE_HIT = 5,
423 LPROC_OSD_CACHE_MISS = 6,
424 LPROC_OSD_COPY_IO = 7,
425 LPROC_OSD_ZEROCOPY_IO = 8,
426 LPROC_OSD_TAIL_IO = 9,
431 extern struct lprocfs_seq_vars lprocfs_osd_obd_vars[];
433 int osd_procfs_init(struct osd_device *osd, const char *name);
434 int osd_procfs_fini(struct osd_device *osd);
437 extern char *osd_obj_tag;
438 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
439 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
440 uint64_t oid, dmu_buf_t **dbp);
441 struct lu_object *osd_object_alloc(const struct lu_env *env,
442 const struct lu_object_header *hdr,
443 struct lu_device *d);
444 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
445 void *buf, uint32_t buflen, struct osd_thandle *oh);
446 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
447 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
448 uint64_t parent, zap_flags_t flags);
449 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
450 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
454 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
455 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
456 int osd_fid_lookup(const struct lu_env *env,
457 struct osd_device *, const struct lu_fid *, uint64_t *);
458 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
459 const struct lu_fid *fid, char *buf);
460 int osd_options_init(void);
461 int osd_convert_root_to_new_seq(const struct lu_env *env,
462 struct osd_device *o);
463 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
466 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
467 const struct dt_index_features *feat);
468 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
469 u64 seq, struct lu_seq_range *range);
470 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
471 uint64_t id, uint64_t dirhash);
472 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
473 uint64_t id, uint64_t dirhash);
474 void osd_zap_cursor_fini(zap_cursor_t *zc);
475 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
478 int __osd_xattr_load(struct osd_device *osd, uint64_t dnode,
479 nvlist_t **sa_xattr);
480 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
481 uint64_t xattr, struct lu_buf *buf,
482 const char *name, int *sizep);
483 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
484 struct lu_buf *buf, const char *name,
485 struct lustre_capa *capa);
486 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
487 const struct lu_buf *buf, const char *name,
488 int fl, struct thandle *handle);
489 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
490 const struct lu_buf *buf, const char *name, int fl,
491 struct thandle *handle, struct lustre_capa *capa);
492 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
493 const char *name, struct thandle *handle);
494 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
495 const char *name, struct thandle *handle,
496 struct lustre_capa *capa);
497 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
498 const struct lu_buf *lb, struct lustre_capa *capa);
499 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
500 int vallen, const char *name, struct osd_thandle *oh);
501 int __osd_sa_xattr_set(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);;
504 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
505 const struct lu_buf *buf, const char *name, int fl,
506 struct osd_thandle *oh);
508 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
509 const struct lu_buf *buf, const char *name, int fl,
510 struct osd_thandle *oh, struct lustre_capa *capa)
514 if (osd_obj2dev(obj)->od_xattr_in_sa) {
515 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
517 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
519 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
525 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
527 return (((flags & FS_APPEND_FL) ? ZFS_APPENDONLY : 0) |
528 ((flags & FS_NODUMP_FL) ? ZFS_NODUMP : 0) |
529 ((flags & FS_IMMUTABLE_FL) ? ZFS_IMMUTABLE : 0));
532 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
534 return (((flags & ZFS_APPENDONLY) ? FS_APPEND_FL : 0) |
535 ((flags & ZFS_NODUMP) ? FS_NODUMP_FL : 0) |
536 ((flags & ZFS_IMMUTABLE) ? FS_IMMUTABLE_FL : 0));
541 #ifndef HAVE_DSL_POOL_CONFIG
542 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
546 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
552 #endif /* _OSD_INTERNAL_H */