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, 2015, 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_SIZE (16ULL << 20) /* reserve 16MB minimum */
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 (DNODES_PER_BLOCK_SHIFT)
91 #define OSD_DNODE_EST_BLKSHIFT (DNODES_PER_BLOCK_SHIFT >> 1)
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 unsigned ozi_reset:1; /* 1 -- no need to advance */
121 /* ozi_pos - position of the cursor:
122 * 0 - before any record
125 * 3 - real records */
128 char ozi_name[MAXNAMELEN]; /* file name for dir */
129 __u64 ozi_key; /* binary key for index files */
132 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
135 * regular ZFS direntry
137 struct zpl_direntry {
138 uint64_t zde_dnode:48,
141 } __attribute__((packed));
144 * lustre direntry adds a fid to regular ZFS direntry
146 struct luz_direntry {
147 struct zpl_direntry lzd_reg;
148 struct lu_fid lzd_fid;
149 } __attribute__((packed));
152 /* cached SA attributes */
166 struct osd_thread_info {
167 const struct lu_env *oti_env;
169 struct lu_fid oti_fid;
171 * XXX temporary: for ->i_op calls.
173 struct timespec oti_time;
175 struct ost_id oti_ostid;
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;
196 extern struct lu_context_key osd_key;
198 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
200 return lu_context_key_get(&env->le_ctx, &osd_key);
204 struct thandle ot_super;
205 struct list_head ot_dcb_list;
206 struct list_head ot_stop_dcb_list;
207 struct list_head ot_unlinked_list;
208 struct list_head ot_sa_list;
209 struct semaphore ot_sa_lock;
211 struct lquota_trans ot_quota_trans;
212 __u32 ot_write_commit:1,
216 #define OSD_OI_NAME_SIZE 16
219 * Object Index (OI) instance.
222 char oi_name[OSD_OI_NAME_SIZE]; /* unused */
227 uint64_t *os_compat_dirs;
228 int os_subdir_count; /* subdir count for each seq */
229 u64 os_seq; /* seq number */
230 struct list_head os_seq_list; /* list to seq_list */
233 struct osd_seq_list {
234 rwlock_t osl_seq_list_lock; /* lock for seq_list */
235 struct list_head osl_seq_list; /* list head for seq */
236 struct semaphore osl_seq_init_sem;
239 #define OSD_OST_MAP_SIZE 32
246 struct dt_device od_dt_dev;
247 /* information about underlying file system */
248 struct objset *od_os;
249 uint64_t od_rootid; /* id of root znode */
250 uint64_t od_unlinkedid; /* id of unlinked zapobj */
251 /* SA attr mapping->id,
252 * name is the same as in ZFS to use defines SA_ZPL_...*/
253 sa_attr_type_t *z_attr_table;
255 struct proc_dir_entry *od_proc_entry;
256 struct lprocfs_stats *od_stats;
258 uint64_t od_max_blksz;
261 struct osd_oi **od_oi_table;
262 unsigned int od_oi_count;
263 struct osd_seq_list od_seq_list;
265 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
266 od_prop_rdonly:1, /**< ZFS property readonly */
268 od_quota_iused_est:1,
276 struct lu_site od_site;
278 /* object IDs of the inode accounting indexes */
279 uint64_t od_iusr_oid;
280 uint64_t od_igrp_oid;
282 /* quota slave instance */
283 struct qsd_instance *od_quota_slave;
285 struct brw_stats od_brw_stats;
286 atomic_t od_r_in_flight;
287 atomic_t od_w_in_flight;
289 /* used to debug zerocopy logic: the fields track all
290 * allocated, loaned and referenced buffers in use.
291 * to be removed once the change is tested well. */
292 atomic_t od_zerocopy_alloc;
293 atomic_t od_zerocopy_loan;
294 atomic_t od_zerocopy_pin;
296 arc_prune_t *arc_prune_cb;
298 /* osd seq instance */
299 struct lu_client_seq *od_cl_seq;
302 enum osd_destroy_type {
303 OSD_DESTROY_NONE = 0,
304 OSD_DESTROY_SYNC = 1,
305 OSD_DESTROY_ASYNC = 2,
309 struct dt_object oo_dt;
311 * Inode for file system object represented by this osd_object. This
312 * inode is pinned for the whole duration of lu_object life.
314 * Not modified concurrently (either setup early during object
315 * creation, or assigned by osd_object_create() under write lock).
318 sa_handle_t *oo_sa_hdl;
319 nvlist_t *oo_sa_xattr;
320 struct list_head oo_sa_linkage;
322 /* used to implement osd_object_*_{lock|unlock} */
323 struct rw_semaphore oo_sem;
325 /* to serialize some updates: destroy vs. others,
327 struct rw_semaphore oo_guard;
329 /* protected by oo_guard */
330 struct list_head oo_unlinked_linkage;
332 /* cached attributes */
333 rwlock_t oo_attr_lock;
334 struct lu_attr oo_attr;
336 /* external dnode holding large EAs, protected by oo_guard */
338 enum osd_destroy_type oo_destroy;
340 __u32 oo_destroyed:1;
342 /* the i_flags in LMA */
344 /* record size for index file */
345 unsigned char oo_keysize;
346 unsigned char oo_recsize;
347 unsigned char oo_recusize; /* unit size */
350 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
351 extern const struct dt_index_operations osd_acct_index_ops;
352 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
353 extern struct lu_device_operations osd_lu_ops;
354 extern struct dt_index_operations osd_dir_ops;
355 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
356 qid_t uid, qid_t gid, long long space,
357 struct osd_thandle *oh, bool is_blk, int *flags,
359 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
360 uint64_t nrblocks, uint64_t est_maxblockshift);
365 static inline int lu_device_is_osd(const struct lu_device *d)
367 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
370 static inline struct osd_object *osd_obj(const struct lu_object *o)
372 LASSERT(lu_device_is_osd(o->lo_dev));
373 return container_of0(o, struct osd_object, oo_dt.do_lu);
376 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
378 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
379 return container_of0(d, struct osd_device, od_dt_dev);
382 static inline struct osd_device *osd_dev(const struct lu_device *d)
384 LASSERT(lu_device_is_osd(d));
385 return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
388 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
390 return osd_obj(&d->do_lu);
393 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
395 return osd_dev(o->oo_dt.do_lu.lo_dev);
398 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
400 return &osd->od_dt_dev.dd_lu_dev;
403 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
405 return osd_dev(o->do_lu.lo_dev)->od_os;
408 static inline int osd_invariant(const struct osd_object *obj)
413 static inline int osd_object_invariant(const struct lu_object *l)
415 return osd_invariant(osd_obj(l));
418 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
420 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
423 static inline char *osd_name(struct osd_device *osd)
425 return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
428 #ifdef CONFIG_PROC_FS
430 LPROC_OSD_READ_BYTES = 0,
431 LPROC_OSD_WRITE_BYTES = 1,
432 LPROC_OSD_GET_PAGE = 2,
433 LPROC_OSD_NO_PAGE = 3,
434 LPROC_OSD_CACHE_ACCESS = 4,
435 LPROC_OSD_CACHE_HIT = 5,
436 LPROC_OSD_CACHE_MISS = 6,
437 LPROC_OSD_COPY_IO = 7,
438 LPROC_OSD_ZEROCOPY_IO = 8,
439 LPROC_OSD_TAIL_IO = 9,
443 extern struct kmem_cache *osd_zapit_cachep;
445 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
447 int osd_procfs_init(struct osd_device *osd, const char *name);
448 int osd_procfs_fini(struct osd_device *osd);
451 extern char *osd_obj_tag;
452 void osd_object_sa_dirty_rele(struct osd_thandle *oh);
453 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
454 uint64_t oid, dmu_buf_t **dbp);
455 struct lu_object *osd_object_alloc(const struct lu_env *env,
456 const struct lu_object_header *hdr,
457 struct lu_device *d);
458 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
459 void *buf, uint32_t buflen, struct osd_thandle *oh);
460 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
461 dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
462 uint64_t parent, zap_flags_t flags);
463 int __osd_object_create(const struct lu_env *env, struct osd_object *obj,
464 dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
468 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
469 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
470 int osd_fid_lookup(const struct lu_env *env,
471 struct osd_device *, const struct lu_fid *, uint64_t *);
472 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
473 const struct lu_fid *fid, char *buf);
474 int osd_options_init(void);
475 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
478 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
479 const struct dt_index_features *feat);
480 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
481 u64 seq, struct lu_seq_range *range);
482 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
483 uint64_t id, uint64_t dirhash);
484 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
485 uint64_t id, uint64_t dirhash);
486 void osd_zap_cursor_fini(zap_cursor_t *zc);
487 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
490 int __osd_xattr_load(struct osd_device *osd, uint64_t dnode,
491 nvlist_t **sa_xattr);
492 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
493 uint64_t xattr, struct lu_buf *buf,
494 const char *name, int *sizep);
495 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
496 struct lu_buf *buf, const char *name);
497 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
498 const struct lu_buf *buf, const char *name,
499 int fl, struct thandle *handle);
500 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
501 const struct lu_buf *buf, const char *name, int fl,
502 struct thandle *handle);
503 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
504 const char *name, struct thandle *handle);
505 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
506 const char *name, struct thandle *handle);
507 void osd_declare_xattrs_destroy(const struct lu_env *env,
508 struct osd_object *obj,
509 struct osd_thandle *oh);
510 int osd_xattrs_destroy(const struct lu_env *env,
511 struct osd_object *obj, struct osd_thandle *oh);
512 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
513 const struct lu_buf *lb);
514 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
515 int vallen, const char *name, struct osd_thandle *oh);
516 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
517 const struct lu_buf *buf, const char *name, int fl,
518 struct osd_thandle *oh);;
519 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
520 const struct lu_buf *buf, const char *name, int fl,
521 struct osd_thandle *oh);
523 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
524 const struct lu_buf *buf, const char *name, int fl,
525 struct osd_thandle *oh)
529 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
533 if (osd_obj2dev(obj)->od_xattr_in_sa) {
534 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
536 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
538 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
544 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
546 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
547 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
548 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
551 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
553 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
554 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
555 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
560 #ifndef HAVE_DSL_POOL_CONFIG
561 static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name)
565 static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name)
570 #ifdef HAVE_SPA_MAXBLOCKSIZE
571 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
572 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
574 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
575 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
576 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
579 #ifdef HAVE_SA_SPILL_ALLOC
581 osd_zio_buf_alloc(size_t size)
583 return sa_spill_alloc(KM_SLEEP);
587 osd_zio_buf_free(void *buf, size_t size)
592 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
593 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
596 #endif /* _OSD_INTERNAL_H */