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,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License version 2 for more details. A copy is
14 * included in the COPYING file that accompanied this code.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 * Copyright 2009 Sun Microsystems, Inc. All rights reserved
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2013, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/lod/lod_internal.h
34 * Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
35 * Author: Mikhail Pershin <mike.pershin@intel.com>
38 #ifndef _LOD_INTERNAL_H
39 #define _LOD_INTERNAL_H
41 #include <libcfs/libcfs.h>
43 #include <dt_object.h>
45 #define LOV_USES_ASSIGNED_STRIPE 0
46 #define LOV_USES_DEFAULT_STRIPE 1
49 __u32 lqr_start_idx; /* start index of new inode */
50 __u32 lqr_offset_idx; /* aliasing for start_idx */
51 int lqr_start_count; /* reseed counter */
52 struct ost_pool lqr_pool; /* round-robin optimized list */
53 unsigned long lqr_dirty:1; /* recalc round-robin list */
57 char pool_name[LOV_MAXPOOLNAME + 1];
58 struct ost_pool pool_obds; /* pool members */
59 atomic_t pool_refcount;
60 struct lod_qos_rr pool_rr;
61 struct hlist_node pool_hash; /* access by poolname */
62 struct list_head pool_list;
63 struct proc_dir_entry *pool_proc_entry;
64 struct obd_device *pool_lobd; /* owner */
67 #define pool_tgt_size(p) ((p)->pool_obds.op_size)
68 #define pool_tgt_count(p) ((p)->pool_obds.op_count)
69 #define pool_tgt_array(p) ((p)->pool_obds.op_array)
70 #define pool_tgt_rw_sem(p) ((p)->pool_obds.op_rw_sem)
73 struct list_head lq_oss_list;
74 struct rw_semaphore lq_rw_sem;
75 __u32 lq_active_oss_count;
76 unsigned int lq_prio_free; /* priority for free space */
77 unsigned int lq_threshold_rr;/* priority for rr */
78 struct lod_qos_rr lq_rr; /* round robin qos data */
79 bool lq_dirty:1, /* recalc qos data */
80 lq_same_space:1,/* the ost's all have approx.
81 the same space avail */
82 lq_reset:1; /* zero current penalties */
86 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
87 struct list_head lqo_oss_list; /* link to lov_qos */
88 __u64 lqo_bavail; /* total bytes avail on OSS */
89 __u64 lqo_penalty; /* current penalty */
90 __u64 lqo_penalty_per_obj; /* penalty decrease
92 time_t lqo_used; /* last used time, seconds */
93 __u32 lqo_ost_count; /* number of osts on this oss */
97 struct lod_qos_oss *ltq_oss; /* oss info */
98 __u64 ltq_penalty; /* current penalty */
99 __u64 ltq_penalty_per_obj; /* penalty decrease
101 __u64 ltq_weight; /* net weighting */
102 time_t ltq_used; /* last used time, seconds */
103 bool ltq_usable:1; /* usable for striping */
106 struct lod_tgt_desc {
107 struct dt_device *ltd_tgt;
108 struct list_head ltd_kill;
109 struct obd_export *ltd_exp;
110 struct obd_uuid ltd_uuid;
113 struct ltd_qos ltd_qos; /* qos info per target */
114 struct obd_statfs ltd_statfs;
115 unsigned long ltd_active:1,/* is this target up for requests */
116 ltd_activate:1,/* should target be activated */
117 ltd_reap:1; /* should this target be deleted */
120 #define TGT_PTRS 256 /* number of pointers at 1st level */
121 #define TGT_PTRS_PER_BLOCK 256 /* number of pointers at 2nd level */
123 struct lod_tgt_desc_idx {
124 struct lod_tgt_desc *ldi_tgt[TGT_PTRS_PER_BLOCK];
127 #define LTD_TGT(ltd, index) \
128 ((ltd)->ltd_tgt_idx[(index) / \
129 TGT_PTRS_PER_BLOCK]->ldi_tgt[(index) % TGT_PTRS_PER_BLOCK])
131 #define OST_TGT(lod, index) LTD_TGT(&lod->lod_ost_descs, index)
132 #define MDT_TGT(lod, index) LTD_TGT(&lod->lod_mdt_descs, index)
133 struct lod_tgt_descs {
134 /* list of known TGTs */
135 struct lod_tgt_desc_idx *ltd_tgt_idx[TGT_PTRS];
136 /* Size of the lod_tgts array, granted to be a power of 2 */
138 /* number of registered TGTs */
140 /* bitmap of TGTs available */
141 cfs_bitmap_t *ltd_tgt_bitmap;
142 /* TGTs scheduled to be deleted */
144 /* Table refcount used for delayed deletion */
146 /* mutex to serialize concurrent updates to the tgt table */
147 struct mutex ltd_mutex;
148 /* read/write semaphore used for array relocation */
149 struct rw_semaphore ltd_rw_sem;
153 struct dt_device lod_dt_dev;
154 struct obd_export *lod_child_exp;
155 struct dt_device *lod_child;
156 cfs_proc_dir_entry_t *lod_proc_entry;
157 struct lprocfs_stats *lod_stats;
158 spinlock_t lod_connects_lock;
160 unsigned int lod_recovery_completed:1,
164 /* lov settings descriptor storing static information */
165 struct lov_desc lod_desc;
167 /* use to protect ld_active_tgt_count and all ltd_active */
168 spinlock_t lod_desc_lock;
170 /* Description of OST */
171 struct lod_tgt_descs lod_ost_descs;
172 /* Description of MDT */
173 struct lod_tgt_descs lod_mdt_descs;
175 /* maximum EA size underlied OSD may have */
176 unsigned int lod_osd_max_easize;
178 /*FIXME: When QOS and pool is implemented for MDT, probably these
179 * structure should be moved to lod_tgt_descs as well.
181 /* QoS info per LOD */
182 struct lod_qos lod_qos; /* qos info per lod */
185 struct ost_pool lod_pool_info; /* all OSTs in a packed array */
187 cfs_hash_t *lod_pools_hash_body; /* used for key access */
188 struct list_head lod_pool_list; /* used for sequential access */
189 struct proc_dir_entry *lod_pool_proc_entry;
191 enum lustre_sec_part lod_sp_me;
193 struct proc_dir_entry *lod_symlink;
196 #define lod_osts lod_ost_descs.ltd_tgts
197 #define lod_ost_bitmap lod_ost_descs.ltd_tgt_bitmap
198 #define lod_ostnr lod_ost_descs.ltd_tgtnr
199 #define lod_osts_size lod_ost_descs.ltd_tgts_size
200 #define ltd_ost ltd_tgt
201 #define lod_ost_desc lod_tgt_desc
203 #define lod_mdts lod_mdt_descs.ltd_tgts
204 #define lod_mdt_bitmap lod_mdt_descs.ltd_tgt_bitmap
205 #define lod_remote_mdt_count lod_mdt_descs.ltd_tgtnr
206 #define lod_mdts_size lod_mdt_descs.ltd_tgts_size
207 #define ltd_mdt ltd_tgt
208 #define lod_mdt_desc lod_tgt_desc
210 struct lod_dir_stripe_info {
211 __u32 ldsi_stripe_offset;
212 __u32 ldsi_def_stripenr;
213 __u32 ldsi_def_stripe_offset;
214 __u32 ldsi_def_hash_type;
215 __u32 ldsi_hash_type;
217 unsigned int ldsi_striping_cached:1,
218 ldsi_def_striping_set:1,
223 * XXX: shrink this structure, currently it's 72bytes on 32bit arch,
224 * so, slab will be allocating 128bytes
227 struct dt_object ldo_obj;
229 /* if object is striped, then the next fields describe stripes */
230 /* For striped directory, ldo_stripenr == slave stripe count */
232 __u16 ldo_layout_gen;
233 __u32 ldo_stripe_size;
235 __u16 ldo_released_stripenr;
237 struct dt_object **ldo_stripe;
238 /* to know how much memory to free, ldo_stripenr can be less */
239 /* default striping for directory represented by this object
240 * is cached in stripenr/stripe_size */
241 unsigned int ldo_stripes_allocated:16,
242 ldo_striping_cached:1,
243 ldo_def_striping_set:1,
244 /* ldo_dir_slave_stripe indicate this is a slave stripe of
246 ldo_dir_slave_stripe:1;
247 __u32 ldo_def_stripe_size;
248 __u16 ldo_def_stripenr;
249 __u16 ldo_def_stripe_offset;
250 struct lod_dir_stripe_info *ldo_dir_stripe;
253 #define ldo_dir_stripe_offset ldo_dir_stripe->ldsi_stripe_offset
254 #define ldo_dir_def_stripenr ldo_dir_stripe->ldsi_def_stripenr
255 #define ldo_dir_hash_type ldo_dir_stripe->ldsi_hash_type
256 #define ldo_dir_def_hash_type ldo_dir_stripe->ldsi_def_hash_type
257 #define ldo_dir_striping_cached ldo_dir_stripe->ldsi_striping_cached
258 #define ldo_dir_striped ldo_dir_stripe->ldsi_striped
259 #define ldo_dir_def_striping_set ldo_dir_stripe->ldsi_def_striping_set
260 #define ldo_dir_def_stripe_offset ldo_dir_stripe->ldsi_def_stripe_offset
263 struct dt_object *lit_obj; /* object from the layer below */
264 /* stripe offset of iteration */
265 __u32 lit_stripe_index;
267 struct dt_it *lit_it; /* iterator from the layer below */
270 struct lod_thread_info {
271 /* per-thread buffer for LOV EA */
273 __u32 lti_ea_store_size;
274 /* per-thread buffer for LMV EA */
275 struct lu_buf lti_buf;
276 struct ost_id lti_ostid;
277 struct lu_fid lti_fid;
278 struct obd_statfs lti_osfs;
279 struct lu_attr lti_attr;
280 struct lod_it lti_it;
281 struct ldlm_res_id lti_res_id;
282 /* used to hold lu_dirent, sizeof(struct lu_dirent) + NAME_MAX */
283 char lti_key[sizeof(struct lu_dirent) + NAME_MAX];
284 struct dt_object_format lti_format;
285 struct lu_name lti_name;
286 struct lu_buf lti_linkea_buf;
287 struct dt_insert_rec lti_dt_rec;
290 extern const struct lu_device_operations lod_lu_ops;
292 static inline int lu_device_is_lod(struct lu_device *d)
294 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &lod_lu_ops);
297 static inline struct lod_device* lu2lod_dev(struct lu_device *d)
299 LASSERT(lu_device_is_lod(d));
300 return container_of0(d, struct lod_device, lod_dt_dev.dd_lu_dev);
303 static inline struct lu_device *lod2lu_dev(struct lod_device *d)
305 return &d->lod_dt_dev.dd_lu_dev;
308 static inline struct obd_device *lod2obd(struct lod_device *d)
310 return d->lod_dt_dev.dd_lu_dev.ld_obd;
313 static inline struct lod_device *dt2lod_dev(struct dt_device *d)
315 LASSERT(lu_device_is_lod(&d->dd_lu_dev));
316 return container_of0(d, struct lod_device, lod_dt_dev);
319 static inline struct lod_object *lu2lod_obj(struct lu_object *o)
321 LASSERT(ergo(o != NULL, lu_device_is_lod(o->lo_dev)));
322 return container_of0(o, struct lod_object, ldo_obj.do_lu);
325 static inline struct lu_object *lod2lu_obj(struct lod_object *obj)
327 return &obj->ldo_obj.do_lu;
330 static inline struct lod_object *lod_obj(const struct lu_object *o)
332 LASSERT(lu_device_is_lod(o->lo_dev));
333 return container_of0(o, struct lod_object, ldo_obj.do_lu);
336 static inline struct lod_object *lod_dt_obj(const struct dt_object *d)
338 return lod_obj(&d->do_lu);
341 static inline struct dt_object* lod_object_child(struct lod_object *o)
343 return container_of0(lu_object_next(lod2lu_obj(o)),
344 struct dt_object, do_lu);
347 static inline struct dt_object *dt_object_child(struct dt_object *o)
349 return container_of0(lu_object_next(&(o)->do_lu),
350 struct dt_object, do_lu);
353 extern struct lu_context_key lod_thread_key;
355 static inline struct lod_thread_info *lod_env_info(const struct lu_env *env)
357 struct lod_thread_info *info;
358 info = lu_context_key_get(&env->le_ctx, &lod_thread_key);
363 static inline struct lu_name *
364 lod_name_get(const struct lu_env *env, const void *area, int len)
366 struct lu_name *lname;
368 lname = &lod_env_info(env)->lti_name;
369 lname->ln_name = area;
370 lname->ln_namelen = len;
374 #define lod_foreach_ost(__dev, index) \
375 if ((__dev)->lod_osts_size > 0) \
376 cfs_foreach_bit((__dev)->lod_ost_bitmap, (index))
379 int lod_fld_lookup(const struct lu_env *env, struct lod_device *lod,
380 const struct lu_fid *fid, __u32 *tgt, int *flags);
382 void lod_getref(struct lod_tgt_descs *ltd);
383 void lod_putref(struct lod_device *lod, struct lod_tgt_descs *ltd);
384 int lod_add_device(const struct lu_env *env, struct lod_device *lod,
385 char *osp, unsigned index, unsigned gen, int mdt_index,
386 char *type, int active);
387 int lod_del_device(const struct lu_env *env, struct lod_device *lod,
388 struct lod_tgt_descs *ltd, char *osp, unsigned idx,
389 unsigned gen, bool for_ost);
390 int lod_fini_tgt(const struct lu_env *env, struct lod_device *lod,
391 struct lod_tgt_descs *ltd, bool for_ost);
392 int lod_load_striping_locked(const struct lu_env *env, struct lod_object *lo);
393 int lod_load_striping(const struct lu_env *env, struct lod_object *lo);
395 int lod_get_ea(const struct lu_env *env, struct lod_object *lo,
398 lod_get_lov_ea(const struct lu_env *env, struct lod_object *lo)
400 return lod_get_ea(env, lo, XATTR_NAME_LOV);
404 lod_get_lmv_ea(const struct lu_env *env, struct lod_object *lo)
406 return lod_get_ea(env, lo, XATTR_NAME_LMV);
410 lod_get_default_lmv_ea(const struct lu_env *env, struct lod_object *lo)
412 return lod_get_ea(env, lo, XATTR_NAME_DEFAULT_LMV);
415 void lod_fix_desc(struct lov_desc *desc);
416 void lod_fix_desc_qos_maxage(__u32 *val);
417 void lod_fix_desc_pattern(__u32 *val);
418 void lod_fix_desc_stripe_count(__u32 *val);
419 void lod_fix_desc_stripe_size(__u64 *val);
420 int lod_pools_init(struct lod_device *m, struct lustre_cfg *cfg);
421 int lod_pools_fini(struct lod_device *m);
422 int lod_parse_striping(const struct lu_env *env, struct lod_object *mo,
423 const struct lu_buf *buf);
424 int lod_parse_dir_striping(const struct lu_env *env, struct lod_object *lo,
425 const struct lu_buf *buf);
426 int lod_initialize_objects(const struct lu_env *env, struct lod_object *mo,
427 struct lov_ost_data_v1 *objs);
428 int lod_store_def_striping(const struct lu_env *env, struct dt_object *dt,
430 int lod_verify_striping(struct lod_device *d, const struct lu_buf *buf,
432 int lod_generate_and_set_lovea(const struct lu_env *env,
433 struct lod_object *mo, struct thandle *th);
434 int lod_ea_store_resize(struct lod_thread_info *info, size_t size);
436 int lod_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count);
437 int lod_ost_pool_remove(struct ost_pool *op, __u32 idx);
438 int lod_ost_pool_extend(struct ost_pool *op, unsigned int min_count);
439 struct pool_desc *lod_find_pool(struct lod_device *lod, char *poolname);
440 void lod_pool_putref(struct pool_desc *pool);
441 int lod_ost_pool_free(struct ost_pool *op);
442 int lod_pool_del(struct obd_device *obd, char *poolname);
443 int lod_ost_pool_init(struct ost_pool *op, unsigned int count);
444 extern cfs_hash_ops_t pool_hash_operations;
445 int lod_check_index_in_pool(__u32 idx, struct pool_desc *pool);
446 int lod_pool_new(struct obd_device *obd, char *poolname);
447 int lod_pool_add(struct obd_device *obd, char *poolname, char *ostname);
448 int lod_pool_remove(struct obd_device *obd, char *poolname, char *ostname);
451 int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo,
452 struct lu_attr *attr, const struct lu_buf *buf,
454 int qos_add_tgt(struct lod_device*, struct lod_tgt_desc *);
455 int qos_del_tgt(struct lod_device *, struct lod_tgt_desc *);
458 int lod_procfs_init(struct lod_device *lod);
459 void lod_procfs_fini(struct lod_device *lod);
462 int lod_load_lmv_shards(const struct lu_env *env, struct lod_object *lo,
463 struct lu_buf *buf, bool resize);
464 int lod_object_set_pool(struct lod_object *o, char *pool);
465 int lod_declare_striped_object(const struct lu_env *env, struct dt_object *dt,
466 struct lu_attr *attr,
467 const struct lu_buf *lovea, struct thandle *th);
468 int lod_striping_create(const struct lu_env *env, struct dt_object *dt,
469 struct lu_attr *attr, struct dt_object_format *dof,
471 void lod_object_free_striping(const struct lu_env *env, struct lod_object *lo);