X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Flod%2Flod_dev.c;h=9ab7972b52860e5413efb3bcdb94a8e0d4e3de48;hb=3fc039a99f41725798d157a422174005a49200c8;hp=194e4967d2ee26dbbde86e33c70e9353c87a9726;hpb=7ff7b6e6f32e5937bd33f184fc0026d9da756565;p=fs%2Flustre-release.git diff --git a/lustre/lod/lod_dev.c b/lustre/lod/lod_dev.c index 194e496..9ab7972 100644 --- a/lustre/lod/lod_dev.c +++ b/lustre/lod/lod_dev.c @@ -23,7 +23,7 @@ * Copyright 2009 Sun Microsystems, Inc. All rights reserved * Use is subject to license terms. * - * Copyright (c) 2012, Intel Corporation. + * Copyright (c) 2012, 2013, Intel Corporation. * */ /* @@ -38,13 +38,13 @@ * Author: Mikhail Pershin */ -#ifndef EXPORT_SYMTAB -# define EXPORT_SYMTAB -#endif #define DEBUG_SUBSYSTEM S_MDS #include +#include +#include #include +#include #include "lod_internal.h" @@ -54,38 +54,40 @@ * \param lod LOD to be lookup at. * \param fid FID of object to find MDT/OST. * \param tgt MDT/OST index to return. - * \param flags indidcate the FID is on MDS or OST. + * \param type indidcate the FID is on MDS or OST. **/ int lod_fld_lookup(const struct lu_env *env, struct lod_device *lod, - const struct lu_fid *fid, __u32 *tgt, int flags) + const struct lu_fid *fid, __u32 *tgt, int *type) { - struct lu_seq_range range; + struct lu_seq_range range = { 0 }; struct lu_server_fld *server_fld; int rc = 0; ENTRY; LASSERTF(fid_is_sane(fid), "Invalid FID "DFID"\n", PFID(fid)); + if (fid_is_idif(fid)) { *tgt = fid_idif_ost_idx(fid); + *type = LU_SEQ_RANGE_OST; RETURN(rc); } - if (!lod->lod_initialized || !fid_is_norm(fid)) { + if (!lod->lod_initialized || (!fid_seq_in_fldb(fid_seq(fid)))) { LASSERT(lu_site2seq(lod2lu_dev(lod)->ld_site) != NULL); + *tgt = lu_site2seq(lod2lu_dev(lod)->ld_site)->ss_node_id; + *type = LU_SEQ_RANGE_MDT; RETURN(rc); } server_fld = lu_site2seq(lod2lu_dev(lod)->ld_site)->ss_server_fld; - range.lsr_flags = flags; + fld_range_set_type(&range, *type); rc = fld_server_lookup(env, server_fld, fid_seq(fid), &range); - if (rc) { - CERROR("%s: Can't find tgt by seq "LPX64", rc %d\n", - lod2obd(lod)->obd_name, fid_seq(fid), rc); + if (rc) RETURN(rc); - } *tgt = range.lsr_index; + *type = range.lsr_flags; CDEBUG(D_INFO, "LOD: got tgt %x for sequence: " LPX64"\n", *tgt, fid_seq(fid)); @@ -97,7 +99,7 @@ extern struct lu_object_operations lod_lu_obj_ops; extern struct dt_object_operations lod_obj_ops; /* Slab for OSD object allocation */ -cfs_mem_cache_t *lod_object_kmem; +struct kmem_cache *lod_object_kmem; static struct lu_kmem_descr lod_caches[] = { { @@ -117,19 +119,20 @@ struct lu_object *lod_object_alloc(const struct lu_env *env, const struct lu_object_header *hdr, struct lu_device *dev) { - struct lu_object *lu_obj; - struct lod_object *lo; + struct lod_object *lod_obj; + struct lu_object *lu_obj; + ENTRY; - OBD_SLAB_ALLOC_PTR_GFP(lo, lod_object_kmem, CFS_ALLOC_IO); - if (lo == NULL) - return NULL; + OBD_SLAB_ALLOC_PTR_GFP(lod_obj, lod_object_kmem, GFP_NOFS); + if (lod_obj == NULL) + RETURN(ERR_PTR(-ENOMEM)); - lu_obj = lod2lu_obj(lo); - dt_object_init(&lo->ldo_obj, NULL, dev); - lo->ldo_obj.do_ops = &lod_obj_ops; + lu_obj = lod2lu_obj(lod_obj); + dt_object_init(&lod_obj->ldo_obj, NULL, dev); + lod_obj->ldo_obj.do_ops = &lod_obj_ops; lu_obj->lo_ops = &lod_lu_obj_ops; - return lu_obj; + RETURN(lu_obj); } static int lod_cleanup_desc_tgts(const struct lu_env *env, @@ -165,6 +168,62 @@ static int lod_cleanup_desc_tgts(const struct lu_env *env, return rc; } +static int lodname2mdt_index(char *lodname, long *index) +{ + char *ptr, *tmp; + + /* The lodname suppose to be fsname-MDTxxxx-mdtlov */ + ptr = strrchr(lodname, '-'); + if (ptr == NULL) { + CERROR("invalid MDT index in '%s'\n", lodname); + return -EINVAL; + } + + if (strncmp(ptr, "-mdtlov", 7) != 0) { + CERROR("invalid MDT index in '%s'\n", lodname); + return -EINVAL; + } + + if ((unsigned long)ptr - (unsigned long)lodname <= 8) { + CERROR("invalid MDT index in '%s'\n", lodname); + return -EINVAL; + } + + if (strncmp(ptr - 8, "-MDT", 4) != 0) { + CERROR("invalid MDT index in '%s'\n", lodname); + return -EINVAL; + } + + *index = simple_strtol(ptr - 4, &tmp, 16); + if (*tmp != '-' || *index > INT_MAX || *index < 0) { + CERROR("invalid MDT index in '%s'\n", lodname); + return -EINVAL; + } + return 0; +} + +/** + * Procss config log on LOD + * \param env environment info + * \param dev lod device + * \param lcfg config log + * + * Add osc config log, + * marker 20 (flags=0x01, v2.2.49.56) lustre-OST0001 'add osc' + * add_uuid nid=192.168.122.162@tcp(0x20000c0a87aa2) 0: 1:nidxxx + * attach 0:lustre-OST0001-osc-MDT0001 1:osc 2:lustre-MDT0001-mdtlov_UUID + * setup 0:lustre-OST0001-osc-MDT0001 1:lustre-OST0001_UUID 2:nid + * lov_modify_tgts add 0:lustre-MDT0001-mdtlov 1:lustre-OST0001_UUID 2:1 3:1 + * marker 20 (flags=0x02, v2.2.49.56) lustre-OST0001 'add osc' + * + * Add mdc config log + * marker 10 (flags=0x01, v2.2.49.56) lustre-MDT0000 'add osp' + * add_uuid nid=192.168.122.162@tcp(0x20000c0a87aa2) 0: 1:nid + * attach 0:lustre-MDT0000-osp-MDT0001 1:osp 2:lustre-MDT0001-mdtlov_UUID + * setup 0:lustre-MDT0000-osp-MDT0001 1:lustre-MDT0000_UUID 2:nid + * modify_mdc_tgts add 0:lustre-MDT0001 1:lustre-MDT0000_UUID 2:0 3:1 + * marker 10 (flags=0x02, v2.2.49.56) lustre-MDT0000_UUID 'add osp' + **/ static int lod_process_config(const struct lu_env *env, struct lu_device *dev, struct lustre_cfg *lcfg) @@ -172,16 +231,21 @@ static int lod_process_config(const struct lu_env *env, struct lod_device *lod = lu2lod_dev(dev); struct lu_device *next = &lod->lod_child->dd_lu_dev; char *arg1; - int rc; + int rc = 0; ENTRY; switch(lcfg->lcfg_command) { case LCFG_LOV_DEL_OBD: case LCFG_LOV_ADD_INA: - case LCFG_LOV_ADD_OBD: { + case LCFG_LOV_ADD_OBD: + case LCFG_ADD_MDC: { __u32 index; + __u32 mdt_index; int gen; - /* lov_modify_tgts add 0:lov_mdsA 1:osp 2:0 3:1 */ + /* lov_modify_tgts add 0:lov_mdsA 1:osp 2:0 3:1 + * modify_mdc_tgts add 0:lustre-MDT0001 + * 1:lustre-MDT0001-mdc0002 + * 2:2 3:1*/ arg1 = lustre_cfg_string(lcfg, 1); if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", &index) != 1) @@ -189,33 +253,57 @@ static int lod_process_config(const struct lu_env *env, if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", &gen) != 1) GOTO(out, rc = -EINVAL); - if (lcfg->lcfg_command == LCFG_LOV_ADD_OBD) - rc = lod_add_device(env, lod, arg1, index, gen, 1); - else if (lcfg->lcfg_command == LCFG_LOV_ADD_INA) - rc = lod_add_device(env, lod, arg1, index, gen, 0); - else + if (lcfg->lcfg_command == LCFG_LOV_ADD_OBD) { + char *mdt; + mdt = strstr(lustre_cfg_string(lcfg, 0), "-MDT"); + /* 1.8 configs don't have "-MDT0000" at the end */ + if (mdt == NULL) { + mdt_index = 0; + } else { + long long_index; + rc = lodname2mdt_index( + lustre_cfg_string(lcfg, 0), + &long_index); + if (rc != 0) + GOTO(out, rc); + mdt_index = long_index; + } + rc = lod_add_device(env, lod, arg1, index, gen, + mdt_index, LUSTRE_OSC_NAME, 1); + } else if (lcfg->lcfg_command == LCFG_ADD_MDC) { + mdt_index = index; + rc = lod_add_device(env, lod, arg1, index, gen, + mdt_index, LUSTRE_MDC_NAME, 1); + } else if (lcfg->lcfg_command == LCFG_LOV_ADD_INA) { + /*FIXME: Add mdt_index for LCFG_LOV_ADD_INA*/ + mdt_index = 0; + rc = lod_add_device(env, lod, arg1, index, gen, + mdt_index, LUSTRE_OSC_NAME, 0); + } else { rc = lod_del_device(env, lod, &lod->lod_ost_descs, - arg1, index, gen); + arg1, index, gen, true); + } break; } case LCFG_PARAM: { - struct lprocfs_static_vars v = { 0 }; - struct obd_device *obd = lod2obd(lod); + struct obd_device *obd = lod2obd(lod); - lprocfs_lod_init_vars(&v); - - rc = class_process_proc_param(PARAM_LOV, v.obd_vars, lcfg, obd); + rc = class_process_proc_seq_param(PARAM_LOV, obd->obd_vars, + lcfg, obd); if (rc > 0) rc = 0; GOTO(out, rc); } case LCFG_CLEANUP: + case LCFG_PRE_CLEANUP: { lu_dev_del_linkage(dev->ld_site, dev); lod_cleanup_desc_tgts(env, lod, &lod->lod_mdt_descs, lcfg); lod_cleanup_desc_tgts(env, lod, &lod->lod_ost_descs, lcfg); + if (lcfg->lcfg_command == LCFG_PRE_CLEANUP) + break; /* * do cleanup on underlying storage only when * all OSPs are cleaned up, as they use that OSD as well @@ -230,7 +318,7 @@ static int lod_process_config(const struct lu_env *env, if (rc) CERROR("error in disconnect from storage: %d\n", rc); break; - + } default: CERROR("%s: unknown command %u\n", lod2obd(lod)->obd_name, lcfg->lcfg_command); @@ -314,19 +402,60 @@ static int lod_statfs(const struct lu_env *env, static struct thandle *lod_trans_create(const struct lu_env *env, struct dt_device *dev) { - return dt_trans_create(env, dt2lod_dev(dev)->lod_child); + struct thandle *th; + + th = dt_trans_create(env, dt2lod_dev(dev)->lod_child); + if (IS_ERR(th)) + return th; + + return th; } static int lod_trans_start(const struct lu_env *env, struct dt_device *dev, struct thandle *th) { - return dt_trans_start(env, dt2lod_dev(dev)->lod_child, th); + struct lod_device *lod = dt2lod_dev((struct dt_device *) dev); + int rc = 0; + + if (unlikely(th->th_update != NULL)) { + struct thandle_update *tu = th->th_update; + struct dt_update_request *update; + + list_for_each_entry(update, &tu->tu_remote_update_list, + dur_list) { + LASSERT(update->dur_dt != NULL); + rc = dt_trans_start(env, update->dur_dt, th); + if (rc != 0) + return rc; + } + } + return dt_trans_start(env, lod->lod_child, th); } -static int lod_trans_stop(const struct lu_env *env, struct thandle *th) +static int lod_trans_stop(const struct lu_env *env, struct dt_device *dt, + struct thandle *th) { - /* XXX: we don't know next device, will be fixed with DNE */ - return dt_trans_stop(env, th->th_dev, th); + struct thandle_update *tu = th->th_update; + struct dt_update_request *update; + struct dt_update_request *tmp; + int rc2 = 0; + int rc; + ENTRY; + + rc = dt_trans_stop(env, th->th_dev, th); + if (likely(tu == NULL)) + RETURN(rc); + + list_for_each_entry_safe(update, tmp, + &tu->tu_remote_update_list, + dur_list) { + /* update will be freed inside dt_trans_stop */ + rc2 = dt_trans_stop(env, update->dur_dt, th); + if (unlikely(rc2 != 0 && rc == 0)) + rc = rc2; + } + + RETURN(rc); } static void lod_conf_get(const struct lu_env *env, @@ -552,7 +681,7 @@ static struct lu_device *lod_device_free(const struct lu_env *env, struct lu_device *next = &lod->lod_child->dd_lu_dev; ENTRY; - LASSERT(cfs_atomic_read(&lu->ld_ref) == 0); + LASSERT(atomic_read(&lu->ld_ref) == 0); dt_device_fini(&lod->lod_dt_dev); OBD_FREE_PTR(lod); RETURN(next); @@ -594,12 +723,12 @@ static struct lu_device *lod_device_fini(const struct lu_env *env, lod_procfs_fini(lod); - rc = lod_fini_tgt(lod, &lod->lod_ost_descs); + rc = lod_fini_tgt(env, lod, &lod->lod_ost_descs, true); if (rc) CERROR("%s:can not fini ost descs %d\n", lod2obd(lod)->obd_name, rc); - rc = lod_fini_tgt(lod, &lod->lod_mdt_descs); + rc = lod_fini_tgt(env, lod, &lod->lod_mdt_descs, false); if (rc) CERROR("%s:can not fini mdt descs %d\n", lod2obd(lod)->obd_name, rc); @@ -685,6 +814,7 @@ static void lod_key_fini(const struct lu_context *ctx, info->lti_ea_store = NULL; info->lti_ea_store_size = 0; } + lu_buf_free(&info->lti_linkea_buf); OBD_FREE_PTR(info); } @@ -713,25 +843,37 @@ static struct lu_device_type lod_device_type = { .ldt_ctx_tags = LCT_MD_THREAD, }; -static int lod_obd_health_check(const struct lu_env *env, - struct obd_device *obd) +static int lod_obd_get_info(const struct lu_env *env, struct obd_export *exp, + __u32 keylen, void *key, __u32 *vallen, void *val, + struct lov_stripe_md *lsm) { - struct lod_device *d = lu2lod_dev(obd->obd_lu_dev); - struct lod_ost_desc *ost; - int i, rc = 1; - ENTRY; - - LASSERT(d); - lod_getref(&d->lod_ost_descs); - lod_foreach_ost(d, i) { - ost = OST_TGT(d, i); - LASSERT(ost && ost->ltd_ost); - rc = obd_health_check(env, ost->ltd_exp->exp_obd); - /* one healthy device is enough */ - if (rc == 0) - break; + int rc = -EINVAL; + + if (KEY_IS(KEY_OSP_CONNECTED)) { + struct obd_device *obd = exp->exp_obd; + struct lod_device *d; + struct lod_ost_desc *ost; + int i, rc = 1; + + if (!obd->obd_set_up || obd->obd_stopping) + RETURN(-EAGAIN); + + d = lu2lod_dev(obd->obd_lu_dev); + lod_getref(&d->lod_ost_descs); + lod_foreach_ost(d, i) { + ost = OST_TGT(d, i); + LASSERT(ost && ost->ltd_ost); + + rc = obd_get_info(env, ost->ltd_exp, keylen, key, + vallen, val, lsm); + /* one healthy device is enough */ + if (rc == 0) + break; + } + lod_putref(d, &d->lod_ost_descs); + RETURN(rc); } - lod_putref(d, &d->lod_ost_descs); + RETURN(rc); } @@ -739,7 +881,7 @@ static struct obd_ops lod_obd_device_ops = { .o_owner = THIS_MODULE, .o_connect = lod_obd_connect, .o_disconnect = lod_obd_disconnect, - .o_health_check = lod_obd_health_check, + .o_get_info = lod_obd_get_info, .o_pool_new = lod_pool_new, .o_pool_rem = lod_pool_remove, .o_pool_add = lod_pool_add, @@ -748,17 +890,17 @@ static struct obd_ops lod_obd_device_ops = { static int __init lod_mod_init(void) { - struct lprocfs_static_vars lvars = { 0 }; - cfs_proc_dir_entry_t *lov_proc_dir; - int rc; + struct obd_type *type; + int rc; rc = lu_kmem_init(lod_caches); if (rc) return rc; - lprocfs_lod_init_vars(&lvars); - - rc = class_register_type(&lod_obd_device_ops, NULL, lvars.module_vars, + rc = class_register_type(&lod_obd_device_ops, NULL, true, NULL, +#ifndef HAVE_ONLY_PROCFS_SEQ + NULL, +#endif LUSTRE_LOD_NAME, &lod_device_type); if (rc) { lu_kmem_fini(lod_caches); @@ -766,23 +908,23 @@ static int __init lod_mod_init(void) } /* create "lov" entry in procfs for compatibility purposes */ - lov_proc_dir = lprocfs_srch(proc_lustre_root, "lov"); - if (lov_proc_dir == NULL) { - lov_proc_dir = lprocfs_register("lov", proc_lustre_root, - NULL, NULL); - if (IS_ERR(lov_proc_dir)) - CERROR("lod: can't create compat entry \"lov\": %d\n", - (int)PTR_ERR(lov_proc_dir)); - } + type = class_search_type(LUSTRE_LOV_NAME); + if (type != NULL && type->typ_procroot != NULL) + return rc; + type = class_search_type(LUSTRE_LOD_NAME); + type->typ_procsym = lprocfs_seq_register("lov", proc_lustre_root, + NULL, NULL); + if (IS_ERR(type->typ_procsym)) { + CERROR("lod: can't create compat entry \"lov\": %d\n", + (int)PTR_ERR(type->typ_procsym)); + type->typ_procsym = NULL; + } return rc; } static void __exit lod_mod_exit(void) { - - lprocfs_try_remove_proc_entry("lov", proc_lustre_root); - class_unregister_type(LUSTRE_LOD_NAME); lu_kmem_fini(lod_caches); }