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, 2014, 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
48 /* Special values to remove LOV EA from disk */
49 #define LOVEA_DELETE_VALUES(size, count, offset, pool) \
50 ((size) == 0 && (count) == 0 && \
51 (offset) == (typeof(offset))(-1) && (pool) == NULL)
53 #define LMVEA_DELETE_VALUES(count, offset) \
54 ((count) == 0 && (offset) == (typeof(offset))(-1))
56 #define LOV_OFFSET_DEFAULT ((__u16)-1)
59 __u32 lqr_start_idx; /* start index of new inode */
60 __u32 lqr_offset_idx; /* aliasing for start_idx */
61 int lqr_start_count; /* reseed counter */
62 struct ost_pool lqr_pool; /* round-robin optimized list */
63 unsigned long lqr_dirty:1; /* recalc round-robin list */
67 char pool_name[LOV_MAXPOOLNAME + 1];
68 struct ost_pool pool_obds; /* pool members */
69 atomic_t pool_refcount;
70 struct lod_qos_rr pool_rr;
71 struct hlist_node pool_hash; /* access by poolname */
72 struct list_head pool_list;
73 struct proc_dir_entry *pool_proc_entry;
74 struct obd_device *pool_lobd; /* owner */
77 #define pool_tgt_size(p) ((p)->pool_obds.op_size)
78 #define pool_tgt_count(p) ((p)->pool_obds.op_count)
79 #define pool_tgt_array(p) ((p)->pool_obds.op_array)
80 #define pool_tgt_rw_sem(p) ((p)->pool_obds.op_rw_sem)
83 struct list_head lq_oss_list;
84 struct rw_semaphore lq_rw_sem;
85 __u32 lq_active_oss_count;
86 unsigned int lq_prio_free; /* priority for free space */
87 unsigned int lq_threshold_rr;/* priority for rr */
88 struct lod_qos_rr lq_rr; /* round robin qos data */
89 bool lq_dirty:1, /* recalc qos data */
90 lq_same_space:1,/* the ost's all have approx.
91 the same space avail */
92 lq_reset:1; /* zero current penalties */
96 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
97 struct list_head lqo_oss_list; /* link to lov_qos */
98 __u64 lqo_bavail; /* total bytes avail on OSS */
99 __u64 lqo_penalty; /* current penalty */
100 __u64 lqo_penalty_per_obj; /* penalty decrease
102 time_t lqo_used; /* last used time, seconds */
103 __u32 lqo_ost_count; /* number of osts on this oss */
107 struct lod_qos_oss *ltq_oss; /* oss info */
108 __u64 ltq_penalty; /* current penalty */
109 __u64 ltq_penalty_per_obj; /* penalty decrease
111 __u64 ltq_weight; /* net weighting */
112 time_t ltq_used; /* last used time, seconds */
113 bool ltq_usable:1; /* usable for striping */
116 struct lod_tgt_desc {
117 struct dt_device *ltd_tgt;
118 struct list_head ltd_kill;
119 struct obd_export *ltd_exp;
120 struct obd_uuid ltd_uuid;
123 struct ltd_qos ltd_qos; /* qos info per target */
124 struct obd_statfs ltd_statfs;
125 unsigned long ltd_active:1,/* is this target up for requests */
126 ltd_activate:1,/* should target be activated */
127 ltd_reap:1; /* should this target be deleted */
130 #define TGT_PTRS 256 /* number of pointers at 1st level */
131 #define TGT_PTRS_PER_BLOCK 256 /* number of pointers at 2nd level */
133 struct lod_tgt_desc_idx {
134 struct lod_tgt_desc *ldi_tgt[TGT_PTRS_PER_BLOCK];
137 #define LTD_TGT(ltd, index) \
138 ((ltd)->ltd_tgt_idx[(index) / \
139 TGT_PTRS_PER_BLOCK]->ldi_tgt[(index) % TGT_PTRS_PER_BLOCK])
141 #define OST_TGT(lod, index) LTD_TGT(&lod->lod_ost_descs, index)
142 #define MDT_TGT(lod, index) LTD_TGT(&lod->lod_mdt_descs, index)
143 struct lod_tgt_descs {
144 /* list of known TGTs */
145 struct lod_tgt_desc_idx *ltd_tgt_idx[TGT_PTRS];
146 /* Size of the lod_tgts array, granted to be a power of 2 */
148 /* number of registered TGTs */
150 /* bitmap of TGTs available */
151 cfs_bitmap_t *ltd_tgt_bitmap;
152 /* TGTs scheduled to be deleted */
154 /* Table refcount used for delayed deletion */
156 /* mutex to serialize concurrent updates to the tgt table */
157 struct mutex ltd_mutex;
158 /* read/write semaphore used for array relocation */
159 struct rw_semaphore ltd_rw_sem;
163 struct dt_device lod_dt_dev;
164 struct obd_export *lod_child_exp;
165 struct dt_device *lod_child;
166 struct proc_dir_entry *lod_proc_entry;
167 struct lprocfs_stats *lod_stats;
168 spinlock_t lod_connects_lock;
170 unsigned int lod_recovery_completed:1,
174 /* lov settings descriptor storing static information */
175 struct lov_desc lod_desc;
177 /* use to protect ld_active_tgt_count and all ltd_active */
178 spinlock_t lod_desc_lock;
180 /* Description of OST */
181 struct lod_tgt_descs lod_ost_descs;
182 /* Description of MDT */
183 struct lod_tgt_descs lod_mdt_descs;
185 /* maximum EA size underlied OSD may have */
186 unsigned int lod_osd_max_easize;
188 /*FIXME: When QOS and pool is implemented for MDT, probably these
189 * structure should be moved to lod_tgt_descs as well.
191 /* QoS info per LOD */
192 struct lod_qos lod_qos; /* qos info per lod */
195 struct ost_pool lod_pool_info; /* all OSTs in a packed array */
197 cfs_hash_t *lod_pools_hash_body; /* used for key access */
198 struct list_head lod_pool_list; /* used for sequential access */
199 struct proc_dir_entry *lod_pool_proc_entry;
201 enum lustre_sec_part lod_sp_me;
203 struct proc_dir_entry *lod_symlink;
206 #define lod_osts lod_ost_descs.ltd_tgts
207 #define lod_ost_bitmap lod_ost_descs.ltd_tgt_bitmap
208 #define lod_ostnr lod_ost_descs.ltd_tgtnr
209 #define lod_osts_size lod_ost_descs.ltd_tgts_size
210 #define ltd_ost ltd_tgt
211 #define lod_ost_desc lod_tgt_desc
213 #define lod_mdts lod_mdt_descs.ltd_tgts
214 #define lod_mdt_bitmap lod_mdt_descs.ltd_tgt_bitmap
215 #define lod_remote_mdt_count lod_mdt_descs.ltd_tgtnr
216 #define lod_mdts_size lod_mdt_descs.ltd_tgts_size
217 #define ltd_mdt ltd_tgt
218 #define lod_mdt_desc lod_tgt_desc
220 struct lod_dir_stripe_info {
221 __u32 ldsi_stripe_offset;
222 __u32 ldsi_def_stripenr;
223 __u32 ldsi_def_stripe_offset;
224 __u32 ldsi_def_hash_type;
225 __u32 ldsi_hash_type;
227 unsigned int ldsi_striping_cached:1,
228 ldsi_def_striping_set:1,
233 * XXX: shrink this structure, currently it's 72bytes on 32bit arch,
234 * so, slab will be allocating 128bytes
237 struct dt_object ldo_obj;
239 /* if object is striped, then the next fields describe stripes */
240 /* For striped directory, ldo_stripenr == slave stripe count */
242 __u16 ldo_layout_gen;
243 __u32 ldo_stripe_size;
245 __u16 ldo_released_stripenr;
247 struct dt_object **ldo_stripe;
248 /* to know how much memory to free, ldo_stripenr can be less */
249 /* default striping for directory represented by this object
250 * is cached in stripenr/stripe_size */
251 unsigned int ldo_stripes_allocated:16,
252 ldo_striping_cached:1,
253 ldo_def_striping_set:1,
254 /* ldo_dir_slave_stripe indicate this is a slave stripe of
256 ldo_dir_slave_stripe:1;
257 __u32 ldo_def_stripe_size;
258 __u16 ldo_def_stripenr;
259 __u16 ldo_def_stripe_offset;
260 struct lod_dir_stripe_info *ldo_dir_stripe;
263 #define ldo_dir_stripe_offset ldo_dir_stripe->ldsi_stripe_offset
264 #define ldo_dir_def_stripenr ldo_dir_stripe->ldsi_def_stripenr
265 #define ldo_dir_hash_type ldo_dir_stripe->ldsi_hash_type
266 #define ldo_dir_def_hash_type ldo_dir_stripe->ldsi_def_hash_type
267 #define ldo_dir_striping_cached ldo_dir_stripe->ldsi_striping_cached
268 #define ldo_dir_striped ldo_dir_stripe->ldsi_striped
269 #define ldo_dir_def_striping_set ldo_dir_stripe->ldsi_def_striping_set
270 #define ldo_dir_def_stripe_offset ldo_dir_stripe->ldsi_def_stripe_offset
273 struct dt_object *lit_obj; /* object from the layer below */
274 /* stripe offset of iteration */
275 __u32 lit_stripe_index;
277 struct dt_it *lit_it; /* iterator from the layer below */
280 struct lod_thread_info {
281 /* per-thread buffer for LOV EA */
283 __u32 lti_ea_store_size;
284 /* per-thread buffer for LMV EA */
285 struct lu_buf lti_buf;
286 struct ost_id lti_ostid;
287 struct lu_fid lti_fid;
288 struct obd_statfs lti_osfs;
289 struct lu_attr lti_attr;
290 struct lod_it lti_it;
291 struct ldlm_res_id lti_res_id;
292 /* used to hold lu_dirent, sizeof(struct lu_dirent) + NAME_MAX */
293 char lti_key[sizeof(struct lu_dirent) + NAME_MAX];
294 struct dt_object_format lti_format;
295 struct lu_name lti_name;
296 struct lu_buf lti_linkea_buf;
297 struct dt_insert_rec lti_dt_rec;
300 extern const struct lu_device_operations lod_lu_ops;
302 static inline int lu_device_is_lod(struct lu_device *d)
304 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &lod_lu_ops);
307 static inline struct lod_device* lu2lod_dev(struct lu_device *d)
309 LASSERT(lu_device_is_lod(d));
310 return container_of0(d, struct lod_device, lod_dt_dev.dd_lu_dev);
313 static inline struct lu_device *lod2lu_dev(struct lod_device *d)
315 return &d->lod_dt_dev.dd_lu_dev;
318 static inline struct obd_device *lod2obd(struct lod_device *d)
320 return d->lod_dt_dev.dd_lu_dev.ld_obd;
323 static inline struct lod_device *dt2lod_dev(struct dt_device *d)
325 LASSERT(lu_device_is_lod(&d->dd_lu_dev));
326 return container_of0(d, struct lod_device, lod_dt_dev);
329 static inline struct lod_object *lu2lod_obj(struct lu_object *o)
331 LASSERT(ergo(o != NULL, lu_device_is_lod(o->lo_dev)));
332 return container_of0(o, struct lod_object, ldo_obj.do_lu);
335 static inline struct lu_object *lod2lu_obj(struct lod_object *obj)
337 return &obj->ldo_obj.do_lu;
340 static inline struct lod_object *lod_obj(const struct lu_object *o)
342 LASSERT(lu_device_is_lod(o->lo_dev));
343 return container_of0(o, struct lod_object, ldo_obj.do_lu);
346 static inline struct lod_object *lod_dt_obj(const struct dt_object *d)
348 return lod_obj(&d->do_lu);
351 static inline struct dt_object* lod_object_child(struct lod_object *o)
353 return container_of0(lu_object_next(lod2lu_obj(o)),
354 struct dt_object, do_lu);
357 static inline struct dt_object *dt_object_child(struct dt_object *o)
359 return container_of0(lu_object_next(&(o)->do_lu),
360 struct dt_object, do_lu);
363 extern struct lu_context_key lod_thread_key;
365 static inline struct lod_thread_info *lod_env_info(const struct lu_env *env)
367 struct lod_thread_info *info;
368 info = lu_context_key_get(&env->le_ctx, &lod_thread_key);
373 static inline struct lu_name *
374 lod_name_get(const struct lu_env *env, const void *area, int len)
376 struct lu_name *lname;
378 lname = &lod_env_info(env)->lti_name;
379 lname->ln_name = area;
380 lname->ln_namelen = len;
384 #define lod_foreach_ost(__dev, index) \
385 if ((__dev)->lod_osts_size > 0) \
386 cfs_foreach_bit((__dev)->lod_ost_bitmap, (index))
389 extern struct kmem_cache *lod_object_kmem;
390 int lod_fld_lookup(const struct lu_env *env, struct lod_device *lod,
391 const struct lu_fid *fid, __u32 *tgt, int *flags);
393 void lod_getref(struct lod_tgt_descs *ltd);
394 void lod_putref(struct lod_device *lod, struct lod_tgt_descs *ltd);
395 int lod_add_device(const struct lu_env *env, struct lod_device *lod,
396 char *osp, unsigned index, unsigned gen, int mdt_index,
397 char *type, int active);
398 int lod_del_device(const struct lu_env *env, struct lod_device *lod,
399 struct lod_tgt_descs *ltd, char *osp, unsigned idx,
400 unsigned gen, bool for_ost);
401 int lod_fini_tgt(const struct lu_env *env, struct lod_device *lod,
402 struct lod_tgt_descs *ltd, bool for_ost);
403 int lod_load_striping_locked(const struct lu_env *env, struct lod_object *lo);
404 int lod_load_striping(const struct lu_env *env, struct lod_object *lo);
406 int lod_get_ea(const struct lu_env *env, struct lod_object *lo,
409 lod_get_lov_ea(const struct lu_env *env, struct lod_object *lo)
411 return lod_get_ea(env, lo, XATTR_NAME_LOV);
415 lod_get_lmv_ea(const struct lu_env *env, struct lod_object *lo)
417 return lod_get_ea(env, lo, XATTR_NAME_LMV);
421 lod_get_default_lmv_ea(const struct lu_env *env, struct lod_object *lo)
423 return lod_get_ea(env, lo, XATTR_NAME_DEFAULT_LMV);
426 void lod_fix_desc(struct lov_desc *desc);
427 void lod_fix_desc_qos_maxage(__u32 *val);
428 void lod_fix_desc_pattern(__u32 *val);
429 void lod_fix_desc_stripe_count(__u32 *val);
430 void lod_fix_desc_stripe_size(__u64 *val);
431 int lod_pools_init(struct lod_device *m, struct lustre_cfg *cfg);
432 int lod_pools_fini(struct lod_device *m);
433 int lod_parse_striping(const struct lu_env *env, struct lod_object *mo,
434 const struct lu_buf *buf);
435 int lod_parse_dir_striping(const struct lu_env *env, struct lod_object *lo,
436 const struct lu_buf *buf);
437 int lod_initialize_objects(const struct lu_env *env, struct lod_object *mo,
438 struct lov_ost_data_v1 *objs);
439 int lod_verify_striping(struct lod_device *d, const struct lu_buf *buf,
441 int lod_generate_and_set_lovea(const struct lu_env *env,
442 struct lod_object *mo, struct thandle *th);
443 int lod_ea_store_resize(struct lod_thread_info *info, size_t size);
445 int lod_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count);
446 int lod_ost_pool_remove(struct ost_pool *op, __u32 idx);
447 int lod_ost_pool_extend(struct ost_pool *op, unsigned int min_count);
448 struct pool_desc *lod_find_pool(struct lod_device *lod, char *poolname);
449 void lod_pool_putref(struct pool_desc *pool);
450 int lod_ost_pool_free(struct ost_pool *op);
451 int lod_pool_del(struct obd_device *obd, char *poolname);
452 int lod_ost_pool_init(struct ost_pool *op, unsigned int count);
453 extern cfs_hash_ops_t pool_hash_operations;
454 int lod_check_index_in_pool(__u32 idx, struct pool_desc *pool);
455 int lod_pool_new(struct obd_device *obd, char *poolname);
456 int lod_pool_add(struct obd_device *obd, char *poolname, char *ostname);
457 int lod_pool_remove(struct obd_device *obd, char *poolname, char *ostname);
460 int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo,
461 struct lu_attr *attr, const struct lu_buf *buf,
463 int qos_add_tgt(struct lod_device*, struct lod_tgt_desc *);
464 int qos_del_tgt(struct lod_device *, struct lod_tgt_desc *);
467 int lod_procfs_init(struct lod_device *lod);
468 void lod_procfs_fini(struct lod_device *lod);
471 extern struct dt_object_operations lod_obj_ops;
472 extern struct lu_object_operations lod_lu_obj_ops;
473 int lod_load_lmv_shards(const struct lu_env *env, struct lod_object *lo,
474 struct lu_buf *buf, bool resize);
475 int lod_object_set_pool(struct lod_object *o, char *pool);
476 int lod_declare_striped_object(const struct lu_env *env, struct dt_object *dt,
477 struct lu_attr *attr,
478 const struct lu_buf *lovea, struct thandle *th);
479 int lod_striping_create(const struct lu_env *env, struct dt_object *dt,
480 struct lu_attr *attr, struct dt_object_format *dof,
482 void lod_object_free_striping(const struct lu_env *env, struct lod_object *lo);