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>
54 #include <sys/nvpair.h>
55 #include <sys/zfs_znode.h>
59 * By design including kmem.h overrides the Linux slab interfaces to provide
60 * the Illumos kmem cache interfaces. To override this and gain access to
61 * the Linux interfaces these preprocessor macros must be undefined.
63 #ifdef kmem_cache_destroy
64 #undef kmem_cache_destroy
67 #ifdef kmem_cache_create
68 #undef kmem_cache_create
71 #ifdef kmem_cache_alloc
72 #undef kmem_cache_alloc
75 #ifdef kmem_cache_free
76 #undef kmem_cache_free
79 #define LUSTRE_ROOT_FID_SEQ 0
80 #define DMU_OSD_SVNAME "svname"
81 #define DMU_OSD_OI_NAME_BASE "oi"
83 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
85 /* Statfs space reservation for grant, fragmentation, and unlink space. */
86 #define OSD_STATFS_RESERVED_BLKS (1ULL << (22 - SPA_MAXBLOCKSHIFT)) /* 4MB */
87 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
89 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
90 #define OSD_DNODE_MIN_BLKSHIFT (SPA_MAXBLOCKSHIFT - DNODE_SHIFT) /* 17-9 =8 */
91 #define OSD_DNODE_EST_BLKSHIFT (SPA_MAXBLOCKSHIFT - 12) /* 17-12=5 */
92 #define OSD_DNODE_EST_COUNT 1024
94 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
97 * Iterator's in-memory data structure for quota file.
100 struct osd_object *oiq_obj;
101 /* DMU accounting object id */
104 zap_cursor_t *oiq_zc;
105 /** identifier for current quota record */
107 unsigned oiq_reset:1; /* 1 -- no need to advance */
111 * Iterator's in-memory data structure for ZAPs
113 * ZFS does not store . and .. on a disk, instead they are
114 * generated up on request
115 * to follow this format we do the same
118 zap_cursor_t *ozi_zc;
119 struct osd_object *ozi_obj;
120 struct lustre_capa *ozi_capa;
121 unsigned ozi_reset:1; /* 1 -- no need to advance */
122 /* ozi_pos - position of the cursor:
123 * 0 - before any record
126 * 3 - real records */
129 char ozi_name[MAXNAMELEN]; /* file name for dir */
130 __u64 ozi_key; /* binary key for index files */
133 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
136 * regular ZFS direntry
138 struct zpl_direntry {
139 uint64_t zde_dnode:48,
142 } __attribute__((packed));
145 * lustre direntry adds a fid to regular ZFS direntry
147 struct luz_direntry {
148 struct zpl_direntry lzd_reg;
149 struct lu_fid lzd_fid;
150 } __attribute__((packed));
153 /* cached SA attributes */
167 struct osd_thread_info {
168 const struct lu_env *oti_env;
170 struct lu_fid oti_fid;
172 * XXX temporary: for ->i_op calls.
174 struct timespec oti_time;
176 * XXX temporary: for capa operations.
178 struct lustre_capa_key oti_capa_key;
179 struct lustre_capa oti_capa;
181 struct ost_id oti_ostid;
187 char oti_key[MAXNAMELEN + 1];
188 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
190 struct lustre_mdt_attrs oti_mdt_attrs;
192 struct lu_attr oti_la;
193 struct osa_attr oti_osa;
194 zap_attribute_t oti_za;
195 dmu_object_info_t oti_doi;
196 struct luz_direntry oti_zde;
198 struct lquota_id_info oti_qi;
199 struct lu_seq_range oti_seq_range;
202 extern struct lu_context_key osd_key;
204 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
206 return lu_context_key_get(&env->le_ctx, &osd_key);
210 struct thandle ot_super;
211 struct list_head ot_dcb_list;
212 struct list_head ot_sa_list;
213 struct semaphore ot_sa_lock;
215 struct lquota_trans ot_quota_trans;
216 __u32 ot_write_commit:1,
220 #define OSD_OI_NAME_SIZE 16
223 * Object Index (OI) instance.
226 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
231 uint64_t *os_compat_dirs;
232 int os_subdir_count; /* subdir count for each seq */
233 u64 os_seq; /* seq number */
234 struct list_head os_seq_list; /* list to seq_list */
237 struct osd_seq_list {
238 rwlock_t osl_seq_list_lock; /* lock for seq_list */
239 struct list_head osl_seq_list; /* list head for seq */
240 struct semaphore osl_seq_init_sem;
243 #define OSD_OST_MAP_SIZE 32
250 struct dt_device od_dt_dev;
251 /* information about underlying file system */
252 struct objset *od_os;
253 uint64_t od_rootid; /* id of root znode */
254 /* SA attr mapping->id,
255 * name is the same as in ZFS to use defines SA_ZPL_...*/
256 sa_attr_type_t *z_attr_table;
261 unsigned int od_fl_capa:1;
262 unsigned long od_capa_timeout;
264 struct lustre_capa_key *od_capa_keys;
265 struct hlist_head *od_capa_hash;
267 struct proc_dir_entry *od_proc_entry;
268 struct lprocfs_stats *od_stats;
272 struct osd_oi **od_oi_table;
273 unsigned int od_oi_count;
274 struct osd_seq_list od_seq_list;
276 unsigned int od_rdonly:1,
278 od_quota_iused_est:1,
286 struct lu_site od_site;
288 /* object IDs of the inode accounting indexes */
289 uint64_t od_iusr_oid;
290 uint64_t od_igrp_oid;
292 /* quota slave instance */
293 struct qsd_instance *od_quota_slave;
295 struct brw_stats od_brw_stats;
296 atomic_t od_r_in_flight;
297 atomic_t od_w_in_flight;
299 /* used to debug zerocopy logic: the fields track all
300 * allocated, loaned and referenced buffers in use.
301 * to be removed once the change is tested well. */
302 atomic_t od_zerocopy_alloc;
303 atomic_t od_zerocopy_loan;
304 atomic_t od_zerocopy_pin;
306 arc_prune_t *arc_prune_cb;
308 /* osd seq instance */
309 struct lu_client_seq *od_cl_seq;
313 struct dt_object oo_dt;
315 * Inode for file system object represented by this osd_object. This
316 * inode is pinned for the whole duration of lu_object life.
318 * Not modified concurrently (either setup early during object
319 * creation, or assigned by osd_object_create() under write lock).
322 sa_handle_t *oo_sa_hdl;
323 nvlist_t *oo_sa_xattr;
324 struct list_head oo_sa_linkage;
326 struct rw_semaphore oo_sem;
328 /* cached attributes */
329 rwlock_t oo_attr_lock;
330 struct lu_attr oo_attr;
332 /* protects extended attributes */
333 struct semaphore oo_guard;
336 /* record size for index file */
337 unsigned char oo_keysize;
338 unsigned char oo_recsize;
339 unsigned char oo_recusize; /* unit size */
342 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
343 extern const struct dt_index_operations osd_acct_index_ops;
344 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
345 extern struct lu_device_operations osd_lu_ops;
346 extern struct dt_index_operations osd_dir_ops;
347 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
348 qid_t uid, qid_t gid, long long space,
349 struct osd_thandle *oh, bool is_blk, int *flags,
351 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
357 static inline int lu_device_is_osd(const struct lu_device *d)
359 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
362 static inline struct osd_object *osd_obj(const struct lu_object *o)
364 LASSERT(lu_device_is_osd(o->lo_dev));
365 return container_of0(o, struct osd_object, oo_dt.do_lu);
368 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
370 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
371 return container_of0(d, struct osd_device, od_dt_dev);
374 static inline struct osd_device *osd_dev(const struct lu_device *d)
376 LASSERT(lu_device_is_osd(d));
377 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
380 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
382 return osd_obj(&d->do_lu);
385 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
387 return osd_dev(o->oo_dt.do_lu.lo_dev);
390 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
392 return &osd->od_dt_dev.dd_lu_dev;
395 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
397 return osd_dev(o->do_lu.lo_dev)->od_os;
400 static inline int osd_invariant(const struct osd_object *obj)
405 static inline int osd_object_invariant(const struct lu_object *l)
407 return osd_invariant(osd_obj(l));
410 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
412 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
415 static inline char *osd_name(struct osd_device *osd)
417 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
420 #ifdef CONFIG_PROC_FS
422 LPROC_OSD_READ_BYTES = 0,
423 LPROC_OSD_WRITE_BYTES = 1,
424 LPROC_OSD_GET_PAGE = 2,
425 LPROC_OSD_NO_PAGE = 3,
426 LPROC_OSD_CACHE_ACCESS = 4,
427 LPROC_OSD_CACHE_HIT = 5,
428 LPROC_OSD_CACHE_MISS = 6,
429 LPROC_OSD_COPY_IO = 7,
430 LPROC_OSD_ZEROCOPY_IO = 8,
431 LPROC_OSD_TAIL_IO = 9,
435 extern struct kmem_cache *osd_zapit_cachep;
437 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
439 int osd_procfs_init(struct osd_device *osd, const char *name);
440 int osd_procfs_fini(struct osd_device *osd);
443 extern char *osd_obj_tag;
444 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
445 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
446 uint64_t oid, dmu_buf_t **dbp);
447 struct lu_object *osd_object_alloc(const struct lu_env *env,
448 const struct lu_object_header *hdr,
449 struct lu_device *d);
450 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
451 void *buf, uint32_t buflen, struct osd_thandle *oh);
452 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
453 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
454 uint64_t parent, zap_flags_t flags);
455 int __osd_object_create(const struct lu_env *env, struct osd_object *obj,
456 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
460 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
461 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
462 int osd_fid_lookup(const struct lu_env *env,
463 struct osd_device *, const struct lu_fid *, uint64_t *);
464 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
465 const struct lu_fid *fid, char *buf);
466 int osd_options_init(void);
467 int osd_convert_root_to_new_seq(const struct lu_env *env,
468 struct osd_device *o);
469 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
472 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
473 const struct dt_index_features *feat);
474 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
475 u64 seq, struct lu_seq_range *range);
476 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
477 uint64_t id, uint64_t dirhash);
478 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
479 uint64_t id, uint64_t dirhash);
480 void osd_zap_cursor_fini(zap_cursor_t *zc);
481 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
484 int __osd_xattr_load(struct osd_device *osd, uint64_t dnode,
485 nvlist_t **sa_xattr);
486 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
487 uint64_t xattr, struct lu_buf *buf,
488 const char *name, int *sizep);
489 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
490 struct lu_buf *buf, const char *name,
491 struct lustre_capa *capa);
492 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
493 const struct lu_buf *buf, const char *name,
494 int fl, struct thandle *handle);
495 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
496 const struct lu_buf *buf, const char *name, int fl,
497 struct thandle *handle, struct lustre_capa *capa);
498 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
499 const char *name, struct thandle *handle);
500 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
501 const char *name, struct thandle *handle,
502 struct lustre_capa *capa);
503 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
504 const struct lu_buf *lb, struct lustre_capa *capa);
505 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
506 int vallen, const char *name, struct osd_thandle *oh);
507 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
508 const struct lu_buf *buf, const char *name, int fl,
509 struct osd_thandle *oh);;
510 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
511 const struct lu_buf *buf, const char *name, int fl,
512 struct osd_thandle *oh);
514 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
515 const struct lu_buf *buf, const char *name, int fl,
516 struct osd_thandle *oh, struct lustre_capa *capa)
520 if (osd_obj2dev(obj)->od_xattr_in_sa) {
521 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
523 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
525 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
531 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
533 return (((flags & FS_APPEND_FL) ? ZFS_APPENDONLY : 0) |
534 ((flags & FS_NODUMP_FL) ? ZFS_NODUMP : 0) |
535 ((flags & FS_IMMUTABLE_FL) ? ZFS_IMMUTABLE : 0));
538 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
540 return (((flags & ZFS_APPENDONLY) ? FS_APPEND_FL : 0) |
541 ((flags & ZFS_NODUMP) ? FS_NODUMP_FL : 0) |
542 ((flags & ZFS_IMMUTABLE) ? FS_IMMUTABLE_FL : 0));
547 #ifndef HAVE_DSL_POOL_CONFIG
548 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
552 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
558 #ifdef HAVE_SA_SPILL_ALLOC
560 osd_zio_buf_alloc(size_t size)
562 return sa_spill_alloc(KM_SLEEP);
566 osd_zio_buf_free(void *buf, size_t size)
571 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
572 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
575 #endif /* _OSD_INTERNAL_H */