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
48 /* Special values to remove LOV EA from disk */
49 #define LOVEA_DELETE_VALUES(size, count, offset) \
50 ((size) == 0 && (count) == 0 && (offset) == (typeof(offset))(-1))
52 #define LMVEA_DELETE_VALUES(count, offset) \
53 ((count) == 0 && (offset) == (typeof(offset))(-1))
56 __u32 lqr_start_idx; /* start index of new inode */
57 __u32 lqr_offset_idx; /* aliasing for start_idx */
58 int lqr_start_count; /* reseed counter */
59 struct ost_pool lqr_pool; /* round-robin optimized list */
60 unsigned long lqr_dirty:1; /* recalc round-robin list */
64 char pool_name[LOV_MAXPOOLNAME + 1];
65 struct ost_pool pool_obds; /* pool members */
66 atomic_t pool_refcount;
67 struct lod_qos_rr pool_rr;
68 struct hlist_node pool_hash; /* access by poolname */
69 struct list_head pool_list;
70 struct proc_dir_entry *pool_proc_entry;
71 struct obd_device *pool_lobd; /* owner */
74 #define pool_tgt_size(p) ((p)->pool_obds.op_size)
75 #define pool_tgt_count(p) ((p)->pool_obds.op_count)
76 #define pool_tgt_array(p) ((p)->pool_obds.op_array)
77 #define pool_tgt_rw_sem(p) ((p)->pool_obds.op_rw_sem)
80 struct list_head lq_oss_list;
81 struct rw_semaphore lq_rw_sem;
82 __u32 lq_active_oss_count;
83 unsigned int lq_prio_free; /* priority for free space */
84 unsigned int lq_threshold_rr;/* priority for rr */
85 struct lod_qos_rr lq_rr; /* round robin qos data */
86 bool lq_dirty:1, /* recalc qos data */
87 lq_same_space:1,/* the ost's all have approx.
88 the same space avail */
89 lq_reset:1; /* zero current penalties */
93 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
94 struct list_head lqo_oss_list; /* link to lov_qos */
95 __u64 lqo_bavail; /* total bytes avail on OSS */
96 __u64 lqo_penalty; /* current penalty */
97 __u64 lqo_penalty_per_obj; /* penalty decrease
99 time_t lqo_used; /* last used time, seconds */
100 __u32 lqo_ost_count; /* number of osts on this oss */
104 struct lod_qos_oss *ltq_oss; /* oss info */
105 __u64 ltq_penalty; /* current penalty */
106 __u64 ltq_penalty_per_obj; /* penalty decrease
108 __u64 ltq_weight; /* net weighting */
109 time_t ltq_used; /* last used time, seconds */
110 bool ltq_usable:1; /* usable for striping */
113 struct lod_tgt_desc {
114 struct dt_device *ltd_tgt;
115 struct list_head ltd_kill;
116 struct obd_export *ltd_exp;
117 struct obd_uuid ltd_uuid;
120 struct ltd_qos ltd_qos; /* qos info per target */
121 struct obd_statfs ltd_statfs;
122 unsigned long ltd_active:1,/* is this target up for requests */
123 ltd_activate:1,/* should target be activated */
124 ltd_reap:1; /* should this target be deleted */
127 #define TGT_PTRS 256 /* number of pointers at 1st level */
128 #define TGT_PTRS_PER_BLOCK 256 /* number of pointers at 2nd level */
130 struct lod_tgt_desc_idx {
131 struct lod_tgt_desc *ldi_tgt[TGT_PTRS_PER_BLOCK];
134 #define LTD_TGT(ltd, index) \
135 ((ltd)->ltd_tgt_idx[(index) / \
136 TGT_PTRS_PER_BLOCK]->ldi_tgt[(index) % TGT_PTRS_PER_BLOCK])
138 #define OST_TGT(lod, index) LTD_TGT(&lod->lod_ost_descs, index)
139 #define MDT_TGT(lod, index) LTD_TGT(&lod->lod_mdt_descs, index)
140 struct lod_tgt_descs {
141 /* list of known TGTs */
142 struct lod_tgt_desc_idx *ltd_tgt_idx[TGT_PTRS];
143 /* Size of the lod_tgts array, granted to be a power of 2 */
145 /* number of registered TGTs */
147 /* bitmap of TGTs available */
148 cfs_bitmap_t *ltd_tgt_bitmap;
149 /* TGTs scheduled to be deleted */
151 /* Table refcount used for delayed deletion */
153 /* mutex to serialize concurrent updates to the tgt table */
154 struct mutex ltd_mutex;
155 /* read/write semaphore used for array relocation */
156 struct rw_semaphore ltd_rw_sem;
160 struct dt_device lod_dt_dev;
161 struct obd_export *lod_child_exp;
162 struct dt_device *lod_child;
163 struct proc_dir_entry *lod_proc_entry;
164 struct lprocfs_stats *lod_stats;
165 spinlock_t lod_connects_lock;
167 unsigned int lod_recovery_completed:1,
171 /* lov settings descriptor storing static information */
172 struct lov_desc lod_desc;
174 /* use to protect ld_active_tgt_count and all ltd_active */
175 spinlock_t lod_desc_lock;
177 /* Description of OST */
178 struct lod_tgt_descs lod_ost_descs;
179 /* Description of MDT */
180 struct lod_tgt_descs lod_mdt_descs;
182 /* maximum EA size underlied OSD may have */
183 unsigned int lod_osd_max_easize;
185 /*FIXME: When QOS and pool is implemented for MDT, probably these
186 * structure should be moved to lod_tgt_descs as well.
188 /* QoS info per LOD */
189 struct lod_qos lod_qos; /* qos info per lod */
192 struct ost_pool lod_pool_info; /* all OSTs in a packed array */
194 cfs_hash_t *lod_pools_hash_body; /* used for key access */
195 struct list_head lod_pool_list; /* used for sequential access */
196 struct proc_dir_entry *lod_pool_proc_entry;
198 enum lustre_sec_part lod_sp_me;
200 struct proc_dir_entry *lod_symlink;
203 #define lod_osts lod_ost_descs.ltd_tgts
204 #define lod_ost_bitmap lod_ost_descs.ltd_tgt_bitmap
205 #define lod_ostnr lod_ost_descs.ltd_tgtnr
206 #define lod_osts_size lod_ost_descs.ltd_tgts_size
207 #define ltd_ost ltd_tgt
208 #define lod_ost_desc lod_tgt_desc
210 #define lod_mdts lod_mdt_descs.ltd_tgts
211 #define lod_mdt_bitmap lod_mdt_descs.ltd_tgt_bitmap
212 #define lod_remote_mdt_count lod_mdt_descs.ltd_tgtnr
213 #define lod_mdts_size lod_mdt_descs.ltd_tgts_size
214 #define ltd_mdt ltd_tgt
215 #define lod_mdt_desc lod_tgt_desc
217 struct lod_dir_stripe_info {
218 __u32 ldsi_stripe_offset;
219 __u32 ldsi_def_stripenr;
220 __u32 ldsi_def_stripe_offset;
221 __u32 ldsi_def_hash_type;
222 __u32 ldsi_hash_type;
224 unsigned int ldsi_striping_cached:1,
225 ldsi_def_striping_set:1,
230 * XXX: shrink this structure, currently it's 72bytes on 32bit arch,
231 * so, slab will be allocating 128bytes
234 struct dt_object ldo_obj;
236 /* if object is striped, then the next fields describe stripes */
237 /* For striped directory, ldo_stripenr == slave stripe count */
239 __u16 ldo_layout_gen;
240 __u32 ldo_stripe_size;
242 __u16 ldo_released_stripenr;
244 struct dt_object **ldo_stripe;
245 /* to know how much memory to free, ldo_stripenr can be less */
246 /* default striping for directory represented by this object
247 * is cached in stripenr/stripe_size */
248 unsigned int ldo_stripes_allocated:16,
249 ldo_striping_cached:1,
250 ldo_def_striping_set:1,
251 /* ldo_dir_slave_stripe indicate this is a slave stripe of
253 ldo_dir_slave_stripe:1;
254 __u32 ldo_def_stripe_size;
255 __u16 ldo_def_stripenr;
256 __u16 ldo_def_stripe_offset;
257 struct lod_dir_stripe_info *ldo_dir_stripe;
260 #define ldo_dir_stripe_offset ldo_dir_stripe->ldsi_stripe_offset
261 #define ldo_dir_def_stripenr ldo_dir_stripe->ldsi_def_stripenr
262 #define ldo_dir_hash_type ldo_dir_stripe->ldsi_hash_type
263 #define ldo_dir_def_hash_type ldo_dir_stripe->ldsi_def_hash_type
264 #define ldo_dir_striping_cached ldo_dir_stripe->ldsi_striping_cached
265 #define ldo_dir_striped ldo_dir_stripe->ldsi_striped
266 #define ldo_dir_def_striping_set ldo_dir_stripe->ldsi_def_striping_set
267 #define ldo_dir_def_stripe_offset ldo_dir_stripe->ldsi_def_stripe_offset
270 struct dt_object *lit_obj; /* object from the layer below */
271 /* stripe offset of iteration */
272 __u32 lit_stripe_index;
274 struct dt_it *lit_it; /* iterator from the layer below */
277 struct lod_thread_info {
278 /* per-thread buffer for LOV EA */
280 __u32 lti_ea_store_size;
281 /* per-thread buffer for LMV EA */
282 struct lu_buf lti_buf;
283 struct ost_id lti_ostid;
284 struct lu_fid lti_fid;
285 struct obd_statfs lti_osfs;
286 struct lu_attr lti_attr;
287 struct lod_it lti_it;
288 struct ldlm_res_id lti_res_id;
289 /* used to hold lu_dirent, sizeof(struct lu_dirent) + NAME_MAX */
290 char lti_key[sizeof(struct lu_dirent) + NAME_MAX];
291 struct dt_object_format lti_format;
292 struct lu_name lti_name;
293 struct lu_buf lti_linkea_buf;
294 struct dt_insert_rec lti_dt_rec;
297 extern const struct lu_device_operations lod_lu_ops;
299 static inline int lu_device_is_lod(struct lu_device *d)
301 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &lod_lu_ops);
304 static inline struct lod_device* lu2lod_dev(struct lu_device *d)
306 LASSERT(lu_device_is_lod(d));
307 return container_of0(d, struct lod_device, lod_dt_dev.dd_lu_dev);
310 static inline struct lu_device *lod2lu_dev(struct lod_device *d)
312 return &d->lod_dt_dev.dd_lu_dev;
315 static inline struct obd_device *lod2obd(struct lod_device *d)
317 return d->lod_dt_dev.dd_lu_dev.ld_obd;
320 static inline struct lod_device *dt2lod_dev(struct dt_device *d)
322 LASSERT(lu_device_is_lod(&d->dd_lu_dev));
323 return container_of0(d, struct lod_device, lod_dt_dev);
326 static inline struct lod_object *lu2lod_obj(struct lu_object *o)
328 LASSERT(ergo(o != NULL, lu_device_is_lod(o->lo_dev)));
329 return container_of0(o, struct lod_object, ldo_obj.do_lu);
332 static inline struct lu_object *lod2lu_obj(struct lod_object *obj)
334 return &obj->ldo_obj.do_lu;
337 static inline struct lod_object *lod_obj(const struct lu_object *o)
339 LASSERT(lu_device_is_lod(o->lo_dev));
340 return container_of0(o, struct lod_object, ldo_obj.do_lu);
343 static inline struct lod_object *lod_dt_obj(const struct dt_object *d)
345 return lod_obj(&d->do_lu);
348 static inline struct dt_object* lod_object_child(struct lod_object *o)
350 return container_of0(lu_object_next(lod2lu_obj(o)),
351 struct dt_object, do_lu);
354 static inline struct dt_object *dt_object_child(struct dt_object *o)
356 return container_of0(lu_object_next(&(o)->do_lu),
357 struct dt_object, do_lu);
360 extern struct lu_context_key lod_thread_key;
362 static inline struct lod_thread_info *lod_env_info(const struct lu_env *env)
364 struct lod_thread_info *info;
365 info = lu_context_key_get(&env->le_ctx, &lod_thread_key);
370 static inline struct lu_name *
371 lod_name_get(const struct lu_env *env, const void *area, int len)
373 struct lu_name *lname;
375 lname = &lod_env_info(env)->lti_name;
376 lname->ln_name = area;
377 lname->ln_namelen = len;
381 #define lod_foreach_ost(__dev, index) \
382 if ((__dev)->lod_osts_size > 0) \
383 cfs_foreach_bit((__dev)->lod_ost_bitmap, (index))
386 int lod_fld_lookup(const struct lu_env *env, struct lod_device *lod,
387 const struct lu_fid *fid, __u32 *tgt, int *flags);
389 void lod_getref(struct lod_tgt_descs *ltd);
390 void lod_putref(struct lod_device *lod, struct lod_tgt_descs *ltd);
391 int lod_add_device(const struct lu_env *env, struct lod_device *lod,
392 char *osp, unsigned index, unsigned gen, int mdt_index,
393 char *type, int active);
394 int lod_del_device(const struct lu_env *env, struct lod_device *lod,
395 struct lod_tgt_descs *ltd, char *osp, unsigned idx,
396 unsigned gen, bool for_ost);
397 int lod_fini_tgt(const struct lu_env *env, struct lod_device *lod,
398 struct lod_tgt_descs *ltd, bool for_ost);
399 int lod_load_striping_locked(const struct lu_env *env, struct lod_object *lo);
400 int lod_load_striping(const struct lu_env *env, struct lod_object *lo);
402 int lod_get_ea(const struct lu_env *env, struct lod_object *lo,
405 lod_get_lov_ea(const struct lu_env *env, struct lod_object *lo)
407 return lod_get_ea(env, lo, XATTR_NAME_LOV);
411 lod_get_lmv_ea(const struct lu_env *env, struct lod_object *lo)
413 return lod_get_ea(env, lo, XATTR_NAME_LMV);
417 lod_get_default_lmv_ea(const struct lu_env *env, struct lod_object *lo)
419 return lod_get_ea(env, lo, XATTR_NAME_DEFAULT_LMV);
422 void lod_fix_desc(struct lov_desc *desc);
423 void lod_fix_desc_qos_maxage(__u32 *val);
424 void lod_fix_desc_pattern(__u32 *val);
425 void lod_fix_desc_stripe_count(__u32 *val);
426 void lod_fix_desc_stripe_size(__u64 *val);
427 int lod_pools_init(struct lod_device *m, struct lustre_cfg *cfg);
428 int lod_pools_fini(struct lod_device *m);
429 int lod_parse_striping(const struct lu_env *env, struct lod_object *mo,
430 const struct lu_buf *buf);
431 int lod_parse_dir_striping(const struct lu_env *env, struct lod_object *lo,
432 const struct lu_buf *buf);
433 int lod_initialize_objects(const struct lu_env *env, struct lod_object *mo,
434 struct lov_ost_data_v1 *objs);
435 int lod_store_def_striping(const struct lu_env *env, struct dt_object *dt,
437 int lod_verify_striping(struct lod_device *d, const struct lu_buf *buf,
439 int lod_generate_and_set_lovea(const struct lu_env *env,
440 struct lod_object *mo, struct thandle *th);
441 int lod_ea_store_resize(struct lod_thread_info *info, size_t size);
443 int lod_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count);
444 int lod_ost_pool_remove(struct ost_pool *op, __u32 idx);
445 int lod_ost_pool_extend(struct ost_pool *op, unsigned int min_count);
446 struct pool_desc *lod_find_pool(struct lod_device *lod, char *poolname);
447 void lod_pool_putref(struct pool_desc *pool);
448 int lod_ost_pool_free(struct ost_pool *op);
449 int lod_pool_del(struct obd_device *obd, char *poolname);
450 int lod_ost_pool_init(struct ost_pool *op, unsigned int count);
451 extern cfs_hash_ops_t pool_hash_operations;
452 int lod_check_index_in_pool(__u32 idx, struct pool_desc *pool);
453 int lod_pool_new(struct obd_device *obd, char *poolname);
454 int lod_pool_add(struct obd_device *obd, char *poolname, char *ostname);
455 int lod_pool_remove(struct obd_device *obd, char *poolname, char *ostname);
458 int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo,
459 struct lu_attr *attr, const struct lu_buf *buf,
461 int qos_add_tgt(struct lod_device*, struct lod_tgt_desc *);
462 int qos_del_tgt(struct lod_device *, struct lod_tgt_desc *);
465 int lod_procfs_init(struct lod_device *lod);
466 void lod_procfs_fini(struct lod_device *lod);
469 int lod_load_lmv_shards(const struct lu_env *env, struct lod_object *lo,
470 struct lu_buf *buf, bool resize);
471 int lod_object_set_pool(struct lod_object *o, char *pool);
472 int lod_declare_striped_object(const struct lu_env *env, struct dt_object *dt,
473 struct lu_attr *attr,
474 const struct lu_buf *lovea, struct thandle *th);
475 int lod_striping_create(const struct lu_env *env, struct dt_object *dt,
476 struct lu_attr *attr, struct dt_object_format *dof,
478 void lod_object_free_striping(const struct lu_env *env, struct lod_object *lo);