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) 2011, 2012, Intel, Inc.
30 * This file is part of Lustre, http://www.lustre.org/
31 * Lustre is a trademark of Sun Microsystems, Inc.
33 * lustre/lod/lod_dev.c
35 * Lustre Logical Object Device
37 * Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
38 * Author: Mikhail Pershin <mike.pershin@intel.com>
42 # define EXPORT_SYMTAB
44 #define DEBUG_SUBSYSTEM S_MDS
46 #include <obd_class.h>
47 #include <lustre_param.h>
49 #include "lod_internal.h"
51 /* Slab for OSD object allocation */
52 cfs_mem_cache_t *lod_object_kmem;
54 static struct lu_kmem_descr lod_caches[] = {
56 .ckd_cache = &lod_object_kmem,
57 .ckd_name = "lod_obj",
58 .ckd_size = sizeof(struct lod_object)
65 static struct lu_device *lod_device_fini(const struct lu_env *env,
68 static int lod_process_config(const struct lu_env *env,
69 struct lu_device *dev,
70 struct lustre_cfg *lcfg)
72 struct lod_device *lod = lu2lod_dev(dev);
73 struct lu_device *next = &lod->lod_child->dd_lu_dev;
78 switch(lcfg->lcfg_command) {
80 case LCFG_LOV_DEL_OBD:
81 case LCFG_LOV_ADD_INA:
82 case LCFG_LOV_ADD_OBD: {
85 /* lov_modify_tgts add 0:lov_mdsA 1:osp 2:0 3:1 */
86 arg1 = lustre_cfg_string(lcfg, 1);
88 if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", &index) != 1)
89 GOTO(out, rc = -EINVAL);
90 if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", &gen) != 1)
91 GOTO(out, rc = -EINVAL);
98 struct lprocfs_static_vars v = { 0 };
99 struct obd_device *obd = lod2obd(lod);
101 lprocfs_lod_init_vars(&v);
103 rc = class_process_proc_param(PARAM_LOV, v.obd_vars, lcfg, obd);
110 lu_dev_del_linkage(dev->ld_site, dev);
112 lod_foreach_ost(lod, i) {
113 struct lod_ost_desc *ost;
114 ost = OST_TGT(lod, i);
115 LASSERT(ost && ost->ltd_ost);
116 next = &ost->ltd_ost->dd_lu_dev;
117 rc = next->ld_ops->ldo_process_config(env, next, lcfg);
119 CERROR("%s: can't process %u: %d\n",
120 lod2obd(lod)->obd_name,
121 lcfg->lcfg_command, rc);
126 * do cleanup on underlying storage only when
127 * all OSPs are cleaned up, as they use that OSD as well
129 next = &lod->lod_child->dd_lu_dev;
130 rc = next->ld_ops->ldo_process_config(env, next, lcfg);
132 CERROR("%s: can't process %u: %d\n",
133 lod2obd(lod)->obd_name, lcfg->lcfg_command, rc);
135 rc = obd_disconnect(lod->lod_child_exp);
137 CERROR("error in disconnect from storage: %d\n", rc);
141 CERROR("%s: unknown command %u\n", lod2obd(lod)->obd_name,
151 static int lod_recovery_complete(const struct lu_env *env,
152 struct lu_device *dev)
154 struct lod_device *lod = lu2lod_dev(dev);
155 struct lu_device *next = &lod->lod_child->dd_lu_dev;
156 struct lod_ost_desc *ost;
160 LASSERT(lod->lod_recovery_completed == 0);
161 lod->lod_recovery_completed = 1;
163 rc = next->ld_ops->ldo_recovery_complete(env, next);
166 lod_foreach_ost(lod, i) {
167 ost = OST_TGT(lod, i);
168 LASSERT(ost && ost->ltd_ost);
169 next = &ost->ltd_ost->dd_lu_dev;
170 rc = next->ld_ops->ldo_recovery_complete(env, next);
172 CERROR("%s: can't complete recovery on #%d: %d\n",
173 lod2obd(lod)->obd_name, i, rc);
180 static int lod_prepare(const struct lu_env *env, struct lu_device *pdev,
181 struct lu_device *cdev)
183 struct lod_device *lod = lu2lod_dev(cdev);
184 struct lu_device *next = &lod->lod_child->dd_lu_dev;
188 rc = next->ld_ops->ldo_prepare(env, pdev, next);
193 const struct lu_device_operations lod_lu_ops = {
194 .ldo_process_config = lod_process_config,
195 .ldo_recovery_complete = lod_recovery_complete,
196 .ldo_prepare = lod_prepare,
199 static int lod_root_get(const struct lu_env *env,
200 struct dt_device *dev, struct lu_fid *f)
202 return dt_root_get(env, dt2lod_dev(dev)->lod_child, f);
205 static int lod_statfs(const struct lu_env *env,
206 struct dt_device *dev, struct obd_statfs *sfs)
208 return dt_statfs(env, dt2lod_dev(dev)->lod_child, sfs);
211 static struct thandle *lod_trans_create(const struct lu_env *env,
212 struct dt_device *dev)
214 return dt_trans_create(env, dt2lod_dev(dev)->lod_child);
217 static int lod_trans_start(const struct lu_env *env, struct dt_device *dev,
220 return dt_trans_start(env, dt2lod_dev(dev)->lod_child, th);
223 static int lod_trans_stop(const struct lu_env *env, struct thandle *th)
225 /* XXX: we don't know next device, will be fixed with DNE */
226 return dt_trans_stop(env, th->th_dev, th);
229 static void lod_conf_get(const struct lu_env *env,
230 const struct dt_device *dev,
231 struct dt_device_param *param)
233 dt_conf_get(env, dt2lod_dev((struct dt_device *)dev)->lod_child, param);
236 static int lod_sync(const struct lu_env *env, struct dt_device *dev)
238 struct lod_device *lod = dt2lod_dev(dev);
239 struct lod_ost_desc *ost;
244 lod_foreach_ost(lod, i) {
245 ost = OST_TGT(lod, i);
246 LASSERT(ost && ost->ltd_ost);
247 rc = dt_sync(env, ost->ltd_ost);
249 CERROR("%s: can't sync %u: %d\n",
250 lod2obd(lod)->obd_name, i, rc);
256 rc = dt_sync(env, lod->lod_child);
261 static int lod_ro(const struct lu_env *env, struct dt_device *dev)
263 return dt_ro(env, dt2lod_dev(dev)->lod_child);
266 static int lod_commit_async(const struct lu_env *env, struct dt_device *dev)
268 return dt_commit_async(env, dt2lod_dev(dev)->lod_child);
271 static int lod_init_capa_ctxt(const struct lu_env *env, struct dt_device *dev,
272 int mode, unsigned long timeout,
273 __u32 alg, struct lustre_capa_key *keys)
275 struct dt_device *next = dt2lod_dev(dev)->lod_child;
276 return dt_init_capa_ctxt(env, next, mode, timeout, alg, keys);
279 static const struct dt_device_operations lod_dt_ops = {
280 .dt_root_get = lod_root_get,
281 .dt_statfs = lod_statfs,
282 .dt_trans_create = lod_trans_create,
283 .dt_trans_start = lod_trans_start,
284 .dt_trans_stop = lod_trans_stop,
285 .dt_conf_get = lod_conf_get,
288 .dt_commit_async = lod_commit_async,
289 .dt_init_capa_ctxt = lod_init_capa_ctxt,
292 static int lod_connect_to_osd(const struct lu_env *env, struct lod_device *lod,
293 struct lustre_cfg *cfg)
295 struct obd_connect_data *data = NULL;
296 struct obd_device *obd;
297 char *nextdev = NULL, *p, *s;
302 LASSERT(lod->lod_child_exp == NULL);
304 /* compatibility hack: we still use old config logs
305 * which specify LOV, but we need to learn underlying
306 * OSD device, which is supposed to be:
307 * <fsname>-MDTxxxx-osd
309 * 2.x MGS generates lines like the following:
310 * #03 (176)lov_setup 0:lustre-MDT0000-mdtlov 1:(struct lov_desc)
311 * 1.8 MGS generates lines like the following:
312 * #03 (168)lov_setup 0:lustre-mdtlov 1:(struct lov_desc)
314 * we use "-MDT" to differentiate 2.x from 1.8 */
316 if ((p = lustre_cfg_string(cfg, 0)) && strstr(p, "-mdtlov")) {
318 OBD_ALLOC(nextdev, len);
320 GOTO(out, rc = -ENOMEM);
323 s = strstr(nextdev, "-mdtlov");
324 if (unlikely(s == NULL)) {
325 CERROR("unable to parse device name %s\n",
326 lustre_cfg_string(cfg, 0));
327 GOTO(out, rc = -EINVAL);
330 if (strstr(nextdev, "-MDT")) {
335 strcpy(s, "-MDT0000-osd");
338 CERROR("unable to parse device name %s\n",
339 lustre_cfg_string(cfg, 0));
340 GOTO(out, rc = -EINVAL);
345 GOTO(out, rc = -ENOMEM);
347 obd = class_name2obd(nextdev);
349 CERROR("can not locate next device: %s\n", nextdev);
350 GOTO(out, rc = -ENOTCONN);
353 data->ocd_connect_flags = OBD_CONNECT_VERSION;
354 data->ocd_version = LUSTRE_VERSION_CODE;
356 rc = obd_connect(env, &lod->lod_child_exp, obd, &obd->obd_uuid,
359 CERROR("cannot connect to next dev %s (%d)\n", nextdev, rc);
363 lod->lod_dt_dev.dd_lu_dev.ld_site =
364 lod->lod_child_exp->exp_obd->obd_lu_dev->ld_site;
365 LASSERT(lod->lod_dt_dev.dd_lu_dev.ld_site);
366 lod->lod_child = lu2dt_dev(lod->lod_child_exp->exp_obd->obd_lu_dev);
372 OBD_FREE(nextdev, len);
376 static int lod_init0(const struct lu_env *env, struct lod_device *lod,
377 struct lu_device_type *ldt, struct lustre_cfg *cfg)
379 struct dt_device_param ddp;
380 struct proc_dir_entry *lov_proc_dir;
381 struct obd_device *obd;
385 obd = class_name2obd(lustre_cfg_string(cfg, 0));
387 CERROR("Cannot find obd with name %s\n",
388 lustre_cfg_string(cfg, 0));
392 obd->obd_lu_dev = &lod->lod_dt_dev.dd_lu_dev;
393 lod->lod_dt_dev.dd_lu_dev.ld_obd = obd;
394 lod->lod_dt_dev.dd_lu_dev.ld_ops = &lod_lu_ops;
395 lod->lod_dt_dev.dd_ops = &lod_dt_ops;
397 rc = lod_connect_to_osd(env, lod, cfg);
401 dt_conf_get(env, &lod->lod_dt_dev, &ddp);
402 lod->lod_osd_max_easize = ddp.ddp_max_ea_size;
404 /* for compatibility we link old procfs's OSC entries to osp ones */
405 lov_proc_dir = lprocfs_srch(proc_lustre_root, "lov");
407 cfs_proc_dir_entry_t *symlink = NULL;
409 OBD_ALLOC(name, strlen(obd->obd_name) + 1);
411 strcpy(name, obd->obd_name);
412 if (strstr(name, "lov"))
413 symlink = lprocfs_add_symlink(name,
417 OBD_FREE(name, strlen(obd->obd_name) + 1);
418 lod->lod_symlink = symlink;
422 cfs_mutex_init(&lod->lod_mutex);
423 cfs_init_rwsem(&lod->lod_rw_sem);
424 cfs_spin_lock_init(&lod->lod_desc_lock);
428 obd_disconnect(lod->lod_child_exp);
432 static struct lu_device *lod_device_free(const struct lu_env *env,
433 struct lu_device *lu)
435 struct lod_device *lod = lu2lod_dev(lu);
436 struct lu_device *next = &lod->lod_child->dd_lu_dev;
439 LASSERT(cfs_atomic_read(&lu->ld_ref) == 0);
440 dt_device_fini(&lod->lod_dt_dev);
445 static struct lu_device *lod_device_alloc(const struct lu_env *env,
446 struct lu_device_type *type,
447 struct lustre_cfg *lcfg)
449 struct lod_device *lod;
450 struct lu_device *lu_dev;
454 lu_dev = ERR_PTR(-ENOMEM);
458 lu_dev = lod2lu_dev(lod);
459 dt_device_init(&lod->lod_dt_dev, type);
460 rc = lod_init0(env, lod, type, lcfg);
462 lod_device_free(env, lu_dev);
463 lu_dev = ERR_PTR(rc);
470 static struct lu_device *lod_device_fini(const struct lu_env *env,
473 struct lod_device *lod = lu2lod_dev(d);
476 if (lod->lod_symlink)
477 lprocfs_remove(&lod->lod_symlink);
483 * we use exports to track all LOD users
485 static int lod_obd_connect(const struct lu_env *env, struct obd_export **exp,
486 struct obd_device *obd, struct obd_uuid *cluuid,
487 struct obd_connect_data *data, void *localdata)
489 struct lod_device *lod = lu2lod_dev(obd->obd_lu_dev);
490 struct lustre_handle conn;
494 CDEBUG(D_CONFIG, "connect #%d\n", lod->lod_connects);
496 rc = class_connect(&conn, obd, cluuid);
500 *exp = class_conn2export(&conn);
502 cfs_mutex_lock(&lod->lod_mutex);
504 /* at the moment we expect the only user */
505 LASSERT(lod->lod_connects == 1);
506 cfs_mutex_unlock(&lod->lod_mutex);
512 * once last export (we don't count self-export) disappeared
513 * lod can be released
515 static int lod_obd_disconnect(struct obd_export *exp)
517 struct obd_device *obd = exp->exp_obd;
518 struct lod_device *lod = lu2lod_dev(obd->obd_lu_dev);
522 /* Only disconnect the underlying layers on the final disconnect. */
523 cfs_mutex_lock(&lod->lod_mutex);
525 if (lod->lod_connects != 0) {
526 /* why should there be more than 1 connect? */
527 cfs_mutex_unlock(&lod->lod_mutex);
528 CERROR("%s: disconnect #%d\n", exp->exp_obd->obd_name,
532 cfs_mutex_unlock(&lod->lod_mutex);
534 /* the last user of lod has gone, let's release the device */
538 rc = class_disconnect(exp); /* bz 9811 */
540 if (rc == 0 && release)
541 class_manual_cleanup(obd);
545 LU_KEY_INIT(lod, struct lod_thread_info);
547 static void lod_key_fini(const struct lu_context *ctx,
548 struct lu_context_key *key, void *data)
550 struct lod_thread_info *info = data;
551 /* allocated in lod_get_lov_ea
552 * XXX: this is overload, a tread may have such store but used only
553 * once. Probably better would be pool of such stores per LOD.
555 if (info->lti_ea_store) {
556 OBD_FREE_LARGE(info->lti_ea_store, info->lti_ea_store_size);
557 info->lti_ea_store = NULL;
558 info->lti_ea_store_size = 0;
563 /* context key: lod_thread_key */
564 LU_CONTEXT_KEY_DEFINE(lod, LCT_MD_THREAD);
566 LU_TYPE_INIT_FINI(lod, &lod_thread_key);
568 static struct lu_device_type_operations lod_device_type_ops = {
569 .ldto_init = lod_type_init,
570 .ldto_fini = lod_type_fini,
572 .ldto_start = lod_type_start,
573 .ldto_stop = lod_type_stop,
575 .ldto_device_alloc = lod_device_alloc,
576 .ldto_device_free = lod_device_free,
578 .ldto_device_fini = lod_device_fini
581 static struct lu_device_type lod_device_type = {
582 .ldt_tags = LU_DEVICE_DT,
583 .ldt_name = LUSTRE_LOD_NAME,
584 .ldt_ops = &lod_device_type_ops,
585 .ldt_ctx_tags = LCT_MD_THREAD,
588 static int lod_obd_health_check(const struct lu_env *env,
589 struct obd_device *obd)
591 struct lod_device *d = lu2lod_dev(obd->obd_lu_dev);
592 struct lod_ost_desc *ost;
598 lod_foreach_ost(d, i) {
600 LASSERT(ost && ost->ltd_ost);
601 rc = obd_health_check(env, ost->ltd_exp->exp_obd);
602 /* one healthy device is enough */
610 static struct obd_ops lod_obd_device_ops = {
611 .o_owner = THIS_MODULE,
612 .o_connect = lod_obd_connect,
613 .o_disconnect = lod_obd_disconnect,
614 .o_health_check = lod_obd_health_check,
617 static int __init lod_mod_init(void)
619 struct lprocfs_static_vars lvars = { 0 };
620 cfs_proc_dir_entry_t *lov_proc_dir;
623 rc = lu_kmem_init(lod_caches);
627 lprocfs_lod_init_vars(&lvars);
629 rc = class_register_type(&lod_obd_device_ops, NULL, lvars.module_vars,
630 LUSTRE_LOD_NAME, &lod_device_type);
632 lu_kmem_fini(lod_caches);
636 /* create "lov" entry in procfs for compatibility purposes */
637 lov_proc_dir = lprocfs_srch(proc_lustre_root, "lov");
638 if (lov_proc_dir == NULL) {
639 lov_proc_dir = lprocfs_register("lov", proc_lustre_root,
641 if (IS_ERR(lov_proc_dir))
642 CERROR("lod: can't create compat entry \"lov\": %d\n",
643 (int)PTR_ERR(lov_proc_dir));
649 static void __exit lod_mod_exit(void)
652 lprocfs_try_remove_proc_entry("lov", proc_lustre_root);
654 class_unregister_type(LUSTRE_LOD_NAME);
655 lu_kmem_fini(lod_caches);
658 MODULE_AUTHOR("Whamcloud, Inc. <http://www.whamcloud.com/>");
659 MODULE_DESCRIPTION("Lustre Logical Object Device ("LUSTRE_LOD_NAME")");
660 MODULE_LICENSE("GPL");
662 module_init(lod_mod_init);
663 module_exit(lod_mod_exit);