/* -*- MODE: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * * mdd/mdd_handler.c * Lustre Metadata Server (mdd) routines * * Copyright (C) 2006 Cluster File Systems, Inc. * Author: Wang Di * * This file is part of the Lustre file system, http://www.lustre.org * Lustre is a trademark of Cluster File Systems, Inc. * * You may have signed or agreed to another license before downloading * this software. If so, you are bound by the terms and conditions * of that agreement, and the following does not apply to you. See the * LICENSE file included with this distribution for more information. * * If you did not agree to a different license, then this copy of Lustre * is open source software; you can redistribute it and/or modify it * under the terms of version 2 of the GNU General Public License as * published by the Free Software Foundation. * * In either case, Lustre is distributed in the hope that it will be * useful, but WITHOUT ANY WARRANTY; without even the implied warranty * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * license text for more details. */ #ifndef EXPORT_SYMTAB # define EXPORT_SYMTAB #endif #define DEBUG_SUBSYSTEM S_MDS #include #include #include #include #include #include #include #include #include #include #include #include "mdd_internal.h" static const char dot[] = "."; static const char dotdot[] = ".."; static int __mdd_lookup(const struct lu_env *env, struct md_object *pobj, const char *name, struct lu_fid* fid, int mask); static int __mdd_lookup_locked(const struct lu_env *env, struct md_object *pobj, const char *name, struct lu_fid* fid, int mask) { struct mdd_object *mdd_obj = md2mdd_obj(pobj); struct dynlock_handle *dlh; int rc; dlh = mdd_pdo_read_lock(env, mdd_obj, name); if (dlh == NULL) return -ENOMEM; rc = __mdd_lookup(env, pobj, name, fid, mask); mdd_pdo_read_unlock(env, mdd_obj, dlh); return rc; } static int mdd_lookup(const struct lu_env *env, struct md_object *pobj, const char *name, struct lu_fid* fid) { int rc; ENTRY; rc = __mdd_lookup_locked(env, pobj, name, fid, MAY_EXEC); RETURN(rc); } static int mdd_parent_fid(const struct lu_env *env, struct mdd_object *obj, struct lu_fid *fid) { return __mdd_lookup_locked(env, &obj->mod_obj, dotdot, fid, 0); } /* * For root fid use special function, whcih does not compare version component * of fid. Vresion component is different for root fids on all MDTs. */ static int mdd_is_root(struct mdd_device *mdd, const struct lu_fid *fid) { return fid_seq(&mdd->mdd_root_fid) == fid_seq(fid) && fid_oid(&mdd->mdd_root_fid) == fid_oid(fid); } /* * return 1: if lf is the fid of the ancestor of p1; * return 0: if not; * * return -EREMOTE: if remote object is found, in this * case fid of remote object is saved to @pf; * * otherwise: values < 0, errors. */ static int mdd_is_parent(const struct lu_env *env, struct mdd_device *mdd, struct mdd_object *p1, const struct lu_fid *lf, struct lu_fid *pf) { struct mdd_object *parent = NULL; struct lu_fid *pfid; int rc; ENTRY; LASSERT(!lu_fid_eq(mdo2fid(p1), lf)); pfid = &mdd_env_info(env)->mti_fid; /* Check for root first. */ if (mdd_is_root(mdd, mdo2fid(p1))) RETURN(0); for(;;) { rc = mdd_parent_fid(env, p1, pfid); if (rc) GOTO(out, rc); if (mdd_is_root(mdd, pfid)) GOTO(out, rc = 0); if (lu_fid_eq(pfid, lf)) GOTO(out, rc = 1); if (parent) mdd_object_put(env, parent); parent = mdd_object_find(env, mdd, pfid); /* cross-ref parent */ if (parent == NULL) { if (pf != NULL) *pf = *pfid; GOTO(out, rc = -EREMOTE); } else if (IS_ERR(parent)) GOTO(out, rc = PTR_ERR(parent)); p1 = parent; } EXIT; out: if (parent && !IS_ERR(parent)) mdd_object_put(env, parent); return rc; } /* * No permission check is needed. * * returns 1: if fid is ancestor of @mo; * returns 0: if fid is not a ancestor of @mo; * * returns EREMOTE if remote object is found, fid of remote object is saved to * @fid; * * returns < 0: if error */ static int mdd_is_subdir(const struct lu_env *env, struct md_object *mo, const struct lu_fid *fid, struct lu_fid *sfid) { struct mdd_device *mdd = mdo2mdd(mo); int rc; ENTRY; if (!S_ISDIR(mdd_object_type(md2mdd_obj(mo)))) RETURN(0); rc = mdd_is_parent(env, mdd, md2mdd_obj(mo), fid, sfid); if (rc == 0) { /* found root */ fid_zero(sfid); } else if (rc == 1) { /* found @fid is parent */ *sfid = *fid; rc = 0; } RETURN(rc); } /* Check whether it may create the cobj under the pobj */ static int mdd_may_create(const struct lu_env *env, struct mdd_object *pobj, struct mdd_object *cobj, int need_check) { int rc = 0; ENTRY; if (cobj && lu_object_exists(&cobj->mod_obj.mo_lu)) RETURN(-EEXIST); if (mdd_is_dead_obj(pobj)) RETURN(-ENOENT); if (need_check) { rc = mdd_permission_internal_locked(env, pobj, (MAY_WRITE | MAY_EXEC)); } RETURN(rc); } /* * It's inline, so penalty for filesystems that don't use sticky bit is * minimal. */ static inline int mdd_is_sticky(const struct lu_env *env, struct mdd_object *pobj, struct mdd_object *cobj) { struct lu_attr *tmp_la = &mdd_env_info(env)->mti_la; struct md_ucred *uc = md_ucred(env); int rc; rc = mdd_la_get(env, cobj, tmp_la, BYPASS_CAPA); if (rc) { return rc; } else if (tmp_la->la_uid == uc->mu_fsuid) { return 0; } else { mdd_read_lock(env, pobj); rc = mdd_la_get(env, pobj, tmp_la, BYPASS_CAPA); mdd_read_unlock(env, pobj); if (rc) return rc; else if (!(tmp_la->la_mode & S_ISVTX)) return 0; else if (tmp_la->la_uid == uc->mu_fsuid) return 0; else return !mdd_capable(uc, CAP_FOWNER); } } /* Check whether it may delete the cobj under the pobj. */ static int mdd_may_delete(const struct lu_env *env, struct mdd_object *pobj, struct mdd_object *cobj, int is_dir, int need_check) { struct mdd_device *mdd = mdo2mdd(&cobj->mod_obj); int rc = 0; ENTRY; LASSERT(cobj); if (!lu_object_exists(&cobj->mod_obj.mo_lu)) RETURN(-ENOENT); if (mdd_is_immutable(cobj) || mdd_is_append(cobj)) RETURN(-EPERM); if (is_dir) { if (!S_ISDIR(mdd_object_type(cobj))) RETURN(-ENOTDIR); if (lu_fid_eq(mdo2fid(cobj), &mdd->mdd_root_fid)) RETURN(-EBUSY); } else if (S_ISDIR(mdd_object_type(cobj))) { RETURN(-EISDIR); } if (pobj) { if (mdd_is_dead_obj(pobj)) RETURN(-ENOENT); if (mdd_is_sticky(env, pobj, cobj)) RETURN(-EPERM); if (need_check) rc = mdd_permission_internal_locked(env, pobj, MAY_WRITE | MAY_EXEC); } RETURN(rc); } int mdd_link_sanity_check(const struct lu_env *env, struct mdd_object *tgt_obj, struct mdd_object *src_obj) { int rc = 0; ENTRY; if (mdd_is_immutable(src_obj) || mdd_is_append(src_obj)) RETURN(-EPERM); if (S_ISDIR(mdd_object_type(src_obj))) RETURN(-EPERM); LASSERT(src_obj != tgt_obj); if (tgt_obj) { rc = mdd_may_create(env, tgt_obj, NULL, 1); if (rc) RETURN(rc); } RETURN(rc); } const struct dt_rec *__mdd_fid_rec(const struct lu_env *env, const struct lu_fid *fid) { struct mdd_thread_info *info = mdd_env_info(env); fid_cpu_to_be(&info->mti_fid2, fid); return (const struct dt_rec *)&info->mti_fid2; } /* insert new index, add reference if isdir, update times */ static int __mdd_index_insert(const struct lu_env *env, struct mdd_object *pobj, const struct lu_fid *lf, const char *name, int is_dir, struct thandle *handle, struct lustre_capa *capa) { struct dt_object *next = mdd_object_child(pobj); struct timeval start; int rc; ENTRY; mdd_lproc_time_start(mdo2mdd(&pobj->mod_obj), &start, LPROC_MDD_INDEX_INSERT); if (dt_try_as_dir(env, next)) { rc = next->do_index_ops->dio_insert(env, next, __mdd_fid_rec(env, lf), (const struct dt_key *)name, handle, capa); } else { rc = -ENOTDIR; } if (rc == 0) { if (is_dir) { mdd_write_lock(env, pobj); mdd_ref_add_internal(env, pobj, handle); mdd_write_unlock(env, pobj); } } mdd_lproc_time_end(mdo2mdd(&pobj->mod_obj), &start, LPROC_MDD_INDEX_INSERT); RETURN(rc); } static int __mdd_index_delete(const struct lu_env *env, struct mdd_object *pobj, const char *name, int is_dir, struct thandle *handle, struct lustre_capa *capa) { struct dt_object *next = mdd_object_child(pobj); struct timeval start; int rc; ENTRY; mdd_lproc_time_start(mdo2mdd(&pobj->mod_obj), &start, LPROC_MDD_INDEX_DELETE); if (dt_try_as_dir(env, next)) { rc = next->do_index_ops->dio_delete(env, next, (struct dt_key *)name, handle, capa); if (rc == 0 && is_dir) { mdd_write_lock(env, pobj); mdd_ref_del_internal(env, pobj, handle); mdd_write_unlock(env, pobj); } } else rc = -ENOTDIR; mdd_lproc_time_end(mdo2mdd(&pobj->mod_obj), &start, LPROC_MDD_INDEX_DELETE); RETURN(rc); } static int __mdd_index_insert_only(const struct lu_env *env, struct mdd_object *pobj, const struct lu_fid *lf, const char *name, struct thandle *handle, struct lustre_capa *capa) { struct dt_object *next = mdd_object_child(pobj); int rc; ENTRY; if (dt_try_as_dir(env, next)) { rc = next->do_index_ops->dio_insert(env, next, __mdd_fid_rec(env, lf), (const struct dt_key *)name, handle, capa); } else { rc = -ENOTDIR; } RETURN(rc); } static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj, struct md_object *src_obj, const char *name, struct md_attr *ma) { struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; struct mdd_object *mdd_tobj = md2mdd_obj(tgt_obj); struct mdd_object *mdd_sobj = md2mdd_obj(src_obj); struct mdd_device *mdd = mdo2mdd(src_obj); struct dynlock_handle *dlh; struct thandle *handle; int rc; ENTRY; mdd_txn_param_build(env, mdd, MDD_TXN_LINK_OP); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); dlh = mdd_pdo_write_lock(env, mdd_tobj, name); if (dlh == NULL) GOTO(out_trans, rc = -ENOMEM); mdd_write_lock(env, mdd_sobj); rc = mdd_link_sanity_check(env, mdd_tobj, mdd_sobj); if (rc) GOTO(out_unlock, rc); rc = __mdd_index_insert_only(env, mdd_tobj, mdo2fid(mdd_sobj), name, handle, mdd_object_capa(env, mdd_tobj)); if (rc) GOTO(out_unlock, rc); mdd_ref_add_internal(env, mdd_sobj, handle); *la = ma->ma_attr; la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_tobj, la, handle, 0); if (rc) GOTO(out_unlock, rc); la->la_valid = LA_CTIME; rc = mdd_attr_set_internal(env, mdd_sobj, la, handle, 0); EXIT; out_unlock: mdd_write_unlock(env, mdd_sobj); mdd_pdo_write_unlock(env, mdd_tobj, dlh); out_trans: mdd_trans_stop(env, mdd, rc, handle); return rc; } /* caller should take a lock before calling */ int mdd_finish_unlink(const struct lu_env *env, struct mdd_object *obj, struct md_attr *ma, struct thandle *th) { int rc; ENTRY; rc = mdd_iattr_get(env, obj, ma); if (rc == 0 && ma->ma_attr.la_nlink == 0) { /* add new orphan and the object * will be deleted during the object_put() */ if (__mdd_orphan_add(env, obj, th) == 0) obj->mod_flags |= ORPHAN_OBJ; if (obj->mod_count == 0) rc = mdd_object_kill(env, obj, ma); else /* clear MA_LOV | MA_COOKIE, if we do not * unlink it in case we get it somewhere */ ma->ma_valid &= ~(MA_LOV | MA_COOKIE); } else ma->ma_valid &= ~(MA_LOV | MA_COOKIE); RETURN(rc); } /* * Check that @dir contains no entries except (possibly) dot and dotdot. * * Returns: * * 0 empty * -ENOTEMPTY not empty * -ve other error * */ static int mdd_dir_is_empty(const struct lu_env *env, struct mdd_object *dir) { struct dt_it *it; struct dt_object *obj; struct dt_it_ops *iops; int result; ENTRY; obj = mdd_object_child(dir); iops = &obj->do_index_ops->dio_it; it = iops->init(env, obj, 0, BYPASS_CAPA); if (it != NULL) { result = iops->get(env, it, (const void *)""); if (result > 0) { int i; for (result = 0, i = 0; result == 0 && i < 3; ++i) result = iops->next(env, it); if (result == 0) result = -ENOTEMPTY; else if (result == +1) result = 0; } else if (result == 0) /* * Huh? Index contains no zero key? */ result = -EIO; iops->put(env, it); iops->fini(env, it); } else result = -ENOMEM; RETURN(result); } int mdd_unlink_sanity_check(const struct lu_env *env, struct mdd_object *pobj, struct mdd_object *cobj, struct md_attr *ma) { struct dt_object *dt_cobj = mdd_object_child(cobj); int rc = 0; ENTRY; rc = mdd_may_delete(env, pobj, cobj, S_ISDIR(ma->ma_attr.la_mode), 1); if (rc) RETURN(rc); if (S_ISDIR(mdd_object_type(cobj))) { if (dt_try_as_dir(env, dt_cobj)) rc = mdd_dir_is_empty(env, cobj); else rc = -ENOTDIR; } RETURN(rc); } static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, struct md_object *cobj, const char *name, struct md_attr *ma) { struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; struct mdd_object *mdd_pobj = md2mdd_obj(pobj); struct mdd_object *mdd_cobj = md2mdd_obj(cobj); struct mdd_device *mdd = mdo2mdd(pobj); struct dynlock_handle *dlh; struct thandle *handle; int rc, is_dir; ENTRY; rc = mdd_log_txn_param_build(env, cobj, ma, MDD_TXN_UNLINK_OP); if (rc) RETURN(rc); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); dlh = mdd_pdo_write_lock(env, mdd_pobj, name); if (dlh == NULL) GOTO(out_trans, rc = -ENOMEM); mdd_write_lock(env, mdd_cobj); rc = mdd_unlink_sanity_check(env, mdd_pobj, mdd_cobj, ma); if (rc) GOTO(cleanup, rc); is_dir = S_ISDIR(lu_object_attr(&cobj->mo_lu)); rc = __mdd_index_delete(env, mdd_pobj, name, is_dir, handle, mdd_object_capa(env, mdd_pobj)); if (rc) GOTO(cleanup, rc); mdd_ref_del_internal(env, mdd_cobj, handle); if (is_dir) { /* unlink dot */ mdd_ref_del_internal(env, mdd_cobj, handle); } *la = ma->ma_attr; la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_pobj, la, handle, 0); if (rc) GOTO(cleanup, rc); la->la_valid = LA_CTIME; rc = mdd_attr_set_internal(env, mdd_cobj, la, handle, 0); if (rc) GOTO(cleanup, rc); rc = mdd_finish_unlink(env, mdd_cobj, ma, handle); if (rc == 0) obd_set_info_async(mdd2obd_dev(mdd)->u.mds.mds_osc_exp, strlen("unlinked"), "unlinked", 0, NULL, NULL); EXIT; cleanup: mdd_write_unlock(env, mdd_cobj); mdd_pdo_write_unlock(env, mdd_pobj, dlh); out_trans: mdd_trans_stop(env, mdd, rc, handle); return rc; } /* * Partial operation. Be aware, this is called with write lock taken, so we use * locksless version of __mdd_lookup() here. */ static int mdd_ni_sanity_check(const struct lu_env *env, struct md_object *pobj, const char *name, const struct lu_fid *fid) { struct mdd_object *obj = md2mdd_obj(pobj); ENTRY; /* EEXIST check */ if (mdd_is_dead_obj(obj)) RETURN(-ENOENT); /* The exist of the name will be checked in _index_insert. */ RETURN(mdd_permission_internal_locked(env, obj, MAY_WRITE | MAY_EXEC)); } static int mdd_name_insert(const struct lu_env *env, struct md_object *pobj, const char *name, const struct lu_fid *fid, int is_dir) { struct lu_attr *la = &mdd_env_info(env)->mti_la; struct mdd_object *mdd_obj = md2mdd_obj(pobj); struct mdd_device *mdd = mdo2mdd(pobj); struct dynlock_handle *dlh; struct thandle *handle; int rc; ENTRY; mdd_txn_param_build(env, mdd, MDD_TXN_INDEX_INSERT_OP); handle = mdd_trans_start(env, mdo2mdd(pobj)); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); dlh = mdd_pdo_write_lock(env, mdd_obj, name); if (dlh == NULL) GOTO(out_trans, rc = -ENOMEM); rc = mdd_ni_sanity_check(env, pobj, name, fid); if (rc) GOTO(out_unlock, rc); rc = __mdd_index_insert(env, mdd_obj, fid, name, is_dir, handle, BYPASS_CAPA); la->la_ctime = la->la_atime = CURRENT_SECONDS; la->la_valid = LA_ATIME | LA_CTIME; rc = mdd_attr_set_internal_locked(env, mdd_obj, la, handle, 0); EXIT; out_unlock: mdd_pdo_write_unlock(env, mdd_obj, dlh); out_trans: mdd_trans_stop(env, mdo2mdd(pobj), rc, handle); return rc; } /* * Be aware, this is called with write lock taken, so we use locksless version * of __mdd_lookup() here. */ static int mdd_nr_sanity_check(const struct lu_env *env, struct md_object *pobj, const char *name) { struct mdd_object *obj = md2mdd_obj(pobj); int rc; ENTRY; /* EEXIST check */ if (mdd_is_dead_obj(obj)) { CWARN("Dir "DFID" is dead?\n", PFID(mdo2fid(obj))); RETURN(-ENOENT); } /* Name presense will be checked in _index_delete. */ rc = mdd_permission_internal_locked(env, obj, MAY_WRITE | MAY_EXEC); RETURN(rc); } static int mdd_name_remove(const struct lu_env *env, struct md_object *pobj, const char *name, int is_dir) { struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; struct mdd_object *mdd_obj = md2mdd_obj(pobj); struct mdd_device *mdd = mdo2mdd(pobj); struct dynlock_handle *dlh; struct thandle *handle; int rc; ENTRY; mdd_txn_param_build(env, mdd, MDD_TXN_INDEX_DELETE_OP); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); dlh = mdd_pdo_write_lock(env, mdd_obj, name); if (dlh == NULL) GOTO(out_trans, rc = -ENOMEM); rc = mdd_nr_sanity_check(env, pobj, name); if (rc) GOTO(out_unlock, rc); rc = __mdd_index_delete(env, mdd_obj, name, is_dir, handle, BYPASS_CAPA); if (rc) GOTO(out_unlock, rc); la->la_ctime = la->la_mtime = CURRENT_SECONDS; la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_obj, la, handle, 0); EXIT; out_unlock: mdd_pdo_write_unlock(env, mdd_obj, dlh); out_trans: mdd_trans_stop(env, mdd, rc, handle); return rc; } static int mdd_rt_sanity_check(const struct lu_env *env, struct mdd_object *tgt_pobj, struct mdd_object *tobj, const struct lu_fid *sfid, const char *name, struct md_attr *ma) { int rc, src_is_dir; ENTRY; /* EEXIST check */ if (mdd_is_dead_obj(tgt_pobj)) RETURN(-ENOENT); src_is_dir = S_ISDIR(ma->ma_attr.la_mode); if (tobj) { rc = mdd_may_delete(env, tgt_pobj, tobj, src_is_dir, 1); if (!rc && S_ISDIR(mdd_object_type(tobj)) && mdd_dir_is_empty(env, tobj)) RETURN(-ENOTEMPTY); } else { rc = mdd_may_create(env, tgt_pobj, NULL, 1); } RETURN(rc); } static int mdd_rename_tgt(const struct lu_env *env, struct md_object *pobj, struct md_object *tobj, const struct lu_fid *lf, const char *name, struct md_attr *ma) { struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; struct mdd_object *mdd_tpobj = md2mdd_obj(pobj); struct mdd_object *mdd_tobj = md2mdd_obj(tobj); struct mdd_device *mdd = mdo2mdd(pobj); struct dynlock_handle *dlh; struct thandle *handle; int rc; ENTRY; mdd_txn_param_build(env, mdd, MDD_TXN_RENAME_TGT_OP); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); dlh = mdd_pdo_write_lock(env, mdd_tpobj, name); if (dlh == NULL) GOTO(out_trans, rc = -ENOMEM); if (mdd_tobj) mdd_write_lock(env, mdd_tobj); /* XXX: Rename sanity checking. */ rc = mdd_rt_sanity_check(env, mdd_tpobj, mdd_tobj, lf, name, ma); if (rc) GOTO(cleanup, rc); /* * If rename_tgt is called then we should just re-insert name with * correct fid, no need to dec/inc parent nlink if obj is dir. */ rc = __mdd_index_delete(env, mdd_tpobj, name, 0, handle, BYPASS_CAPA); if (rc) GOTO(cleanup, rc); rc = __mdd_index_insert_only(env, mdd_tpobj, lf, name, handle, BYPASS_CAPA); if (rc) GOTO(cleanup, rc); *la = ma->ma_attr; la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_tpobj, la, handle, 0); if (rc) GOTO(cleanup, rc); if (tobj && lu_object_exists(&tobj->mo_lu)) { mdd_ref_del_internal(env, mdd_tobj, handle); la->la_valid = LA_CTIME; rc = mdd_attr_set_internal(env, mdd_tobj, la, handle, 0); } EXIT; cleanup: if (tobj) mdd_write_unlock(env, mdd_tobj); mdd_pdo_write_unlock(env, mdd_tpobj, dlh); out_trans: mdd_trans_stop(env, mdd, rc, handle); return rc; } /* * The permission has been checked when obj created, no need check again. */ static int mdd_cd_sanity_check(const struct lu_env *env, struct mdd_object *obj) { ENTRY; /* EEXIST check */ if (!obj || mdd_is_dead_obj(obj)) RETURN(-ENOENT); RETURN(0); } static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, struct md_object *cobj, const struct md_create_spec *spec, struct md_attr *ma) { struct mdd_device *mdd = mdo2mdd(cobj); struct mdd_object *mdd_pobj = md2mdd_obj(pobj);/* XXX maybe NULL */ struct mdd_object *son = md2mdd_obj(cobj); struct lu_attr *attr = &ma->ma_attr; struct lov_mds_md *lmm = NULL; int lmm_size = 0; struct thandle *handle; int rc; ENTRY; rc = mdd_cd_sanity_check(env, son); if (rc) RETURN(rc); if (spec->sp_cr_flags & MDS_OPEN_DELAY_CREATE || !(spec->sp_cr_flags & FMODE_WRITE)) RETURN(0); rc = mdd_lov_create(env, mdd, mdd_pobj, son, &lmm, &lmm_size, spec, attr); if (rc) RETURN(rc); mdd_txn_param_build(env, mdd, MDD_TXN_CREATE_DATA_OP); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(rc = PTR_ERR(handle)); /* * XXX: Setting the lov ea is not locked but setting the attr is locked? * Should this be fixed? */ /* Replay creates has objects already */ if (spec->u.sp_ea.no_lov_create) { CDEBUG(D_INFO, "we already have lov ea\n"); rc = mdd_lov_set_md(env, mdd_pobj, son, (struct lov_mds_md *)spec->u.sp_ea.eadata, spec->u.sp_ea.eadatalen, handle, 0); } else rc = mdd_lov_set_md(env, mdd_pobj, son, lmm, lmm_size, handle, 0); if (rc == 0) rc = mdd_attr_get_internal_locked(env, son, ma); /* Finish mdd_lov_create() stuff. */ mdd_lov_create_finish(env, mdd, rc); mdd_trans_stop(env, mdd, rc, handle); if (lmm) OBD_FREE(lmm, lmm_size); RETURN(rc); } static int __mdd_lookup(const struct lu_env *env, struct md_object *pobj, const char *name, struct lu_fid* fid, int mask) { const struct dt_key *key = (const struct dt_key *)name; struct mdd_object *mdd_obj = md2mdd_obj(pobj); struct dt_object *dir = mdd_object_child(mdd_obj); struct dt_rec *rec = (struct dt_rec *)fid; struct timeval start; int rc; ENTRY; mdd_lproc_time_start(mdo2mdd(pobj), &start, LPROC_MDD_LOOKUP); if (mdd_is_dead_obj(mdd_obj)) RETURN(-ESTALE); rc = lu_object_exists(mdd2lu_obj(mdd_obj)); if (rc == 0) RETURN(-ESTALE); else if (rc < 0) { CERROR("Object "DFID" locates on remote server\n", PFID(mdo2fid(mdd_obj))); LBUG(); } rc = mdd_permission_internal_locked(env, mdd_obj, mask); if (rc) RETURN(rc); if (S_ISDIR(mdd_object_type(mdd_obj)) && dt_try_as_dir(env, dir)) { rc = dir->do_index_ops->dio_lookup(env, dir, rec, key, mdd_object_capa(env, mdd_obj)); if (rc == 0) fid_be_to_cpu(fid, fid); } else rc = -ENOTDIR; mdd_lproc_time_end(mdo2mdd(pobj), &start, LPROC_MDD_LOOKUP); RETURN(rc); } int mdd_object_initialize(const struct lu_env *env, const struct lu_fid *pfid, struct mdd_object *child, struct md_attr *ma, struct thandle *handle) { int rc; ENTRY; /* * Update attributes for child. * * FIXME: * (1) the valid bits should be converted between Lustre and Linux; * (2) maybe, the child attributes should be set in OSD when creation. */ rc = mdd_attr_set_internal(env, child, &ma->ma_attr, handle, 0); if (rc != 0) RETURN(rc); if (S_ISDIR(ma->ma_attr.la_mode)) { /* Add "." and ".." for newly created dir */ mdd_ref_add_internal(env, child, handle); rc = __mdd_index_insert_only(env, child, mdo2fid(child), dot, handle, BYPASS_CAPA); if (rc == 0) { rc = __mdd_index_insert_only(env, child, pfid, dotdot, handle, BYPASS_CAPA); if (rc != 0) { int rc2; rc2 = __mdd_index_delete(env, child, dot, 0, handle, BYPASS_CAPA); if (rc2 != 0) CERROR("Failure to cleanup after dotdot" " creation: %d (%d)\n", rc2, rc); else mdd_ref_del_internal(env, child, handle); } } } RETURN(rc); } static int mdd_create_sanity_check(const struct lu_env *env, struct md_object *pobj, const char *name, struct md_attr *ma, int lookup) { struct mdd_thread_info *info = mdd_env_info(env); struct lu_attr *la = &info->mti_la; struct lu_fid *fid = &info->mti_fid; struct mdd_object *obj = md2mdd_obj(pobj); int rc; ENTRY; /* EEXIST check */ if (mdd_is_dead_obj(obj)) RETURN(-ENOENT); /* * In some cases this lookup is not needed - we know before that if name * exists or not. */ if (lookup) { /* * Check if the name already exist, though it will be checked in * _index_insert also, for avoiding rolling back if exists * _index_insert. */ rc = __mdd_lookup_locked(env, pobj, name, fid, MAY_WRITE | MAY_EXEC); if (rc != -ENOENT) RETURN(rc ? : -EEXIST); } else { /* * Check if has WRITE permission for the parent. */ rc = mdd_permission_internal_locked(env, obj, MAY_WRITE); if (rc) RETURN(rc); } /* sgid check */ mdd_read_lock(env, obj); rc = mdd_la_get(env, obj, la, BYPASS_CAPA); mdd_read_unlock(env, obj); if (rc != 0) RETURN(rc); if (la->la_mode & S_ISGID) { ma->ma_attr.la_gid = la->la_gid; if (S_ISDIR(ma->ma_attr.la_mode)) { ma->ma_attr.la_mode |= S_ISGID; ma->ma_attr.la_valid |= LA_MODE; } } switch (ma->ma_attr.la_mode & S_IFMT) { case S_IFREG: case S_IFDIR: case S_IFLNK: case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK: rc = 0; break; default: rc = -EINVAL; break; } RETURN(rc); } /* * Create object and insert it into namespace. */ static int mdd_create(const struct lu_env *env, struct md_object *pobj, const char *name, struct md_object *child, struct md_create_spec *spec, struct md_attr* ma) { struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; struct mdd_object *mdd_pobj = md2mdd_obj(pobj); struct mdd_object *son = md2mdd_obj(child); struct mdd_device *mdd = mdo2mdd(pobj); struct lu_attr *attr = &ma->ma_attr; struct lov_mds_md *lmm = NULL; struct thandle *handle; int rc, created = 0, inserted = 0, lmm_size = 0; struct dynlock_handle *dlh; struct timeval start; ENTRY; mdd_lproc_time_start(mdd, &start, LPROC_MDD_CREATE); /* * Two operations have to be performed: * * - allocation of new object (->do_create()), and * * - insertion into parent index (->dio_insert()). * * Due to locking, operation order is not important, when both are * successful, *but* error handling cases are quite different: * * - if insertion is done first, and following object creation fails, * insertion has to be rolled back, but this operation might fail * also leaving us with dangling index entry. * * - if creation is done first, is has to be undone if insertion * fails, leaving us with leaked space, which is neither good, nor * fatal. * * It seems that creation-first is simplest solution, but it is * sub-optimal in the frequent * * $ mkdir foo * $ mkdir foo * * case, because second mkdir is bound to create object, only to * destroy it immediately. * * To avoid this follow local file systems that do double lookup: * * 0. lookup -> -EEXIST (mdd_create_sanity_check()) * * 1. create (mdd_object_create_internal()) * * 2. insert (__mdd_index_insert(), lookup again) */ /* Sanity checks before big job. */ rc = mdd_create_sanity_check(env, pobj, name, ma, spec->sp_cr_lookup); if (rc) RETURN(rc); /* * No RPC inside the transaction, so OST objects should be created at * first. */ if (S_ISREG(attr->la_mode)) { rc = mdd_lov_create(env, mdd, mdd_pobj, son, &lmm, &lmm_size, spec, attr); if (rc) RETURN(rc); } mdd_txn_param_build(env, mdd, MDD_TXN_MKDIR_OP); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); dlh = mdd_pdo_write_lock(env, mdd_pobj, name); if (dlh == NULL) GOTO(out_trans, rc = -ENOMEM); /* * XXX: Check that link can be added to the parent in mkdir case. */ mdd_write_lock(env, son); rc = mdd_object_create_internal(env, son, ma, handle); if (rc) { mdd_write_unlock(env, son); GOTO(cleanup, rc); } created = 1; #ifdef CONFIG_FS_POSIX_ACL mdd_read_lock(env, mdd_pobj); rc = mdd_acl_init(env, mdd_pobj, son, &ma->ma_attr.la_mode, handle); mdd_read_unlock(env, mdd_pobj); if (rc) { mdd_write_unlock(env, son); GOTO(cleanup, rc); } else { ma->ma_attr.la_valid |= LA_MODE; } #endif rc = mdd_object_initialize(env, mdo2fid(mdd_pobj), son, ma, handle); mdd_write_unlock(env, son); if (rc) /* * Object has no links, so it will be destroyed when last * reference is released. (XXX not now.) */ GOTO(cleanup, rc); rc = __mdd_index_insert(env, mdd_pobj, mdo2fid(son), name, S_ISDIR(attr->la_mode), handle, mdd_object_capa(env, mdd_pobj)); if (rc) GOTO(cleanup, rc); inserted = 1; /* Replay creates has objects already. */ if (spec->u.sp_ea.no_lov_create) { CDEBUG(D_INFO, "we already have lov ea\n"); LASSERT(lmm == NULL); lmm = (struct lov_mds_md *)spec->u.sp_ea.eadata; lmm_size = spec->u.sp_ea.eadatalen; } rc = mdd_lov_set_md(env, mdd_pobj, son, lmm, lmm_size, handle, 0); if (rc) { CERROR("error on stripe info copy %d \n", rc); GOTO(cleanup, rc); } if (lmm && lmm_size > 0) { /* Set Lov here, do not get lmm again later */ memcpy(ma->ma_lmm, lmm, lmm_size); ma->ma_lmm_size = lmm_size; ma->ma_valid |= MA_LOV; } if (S_ISLNK(attr->la_mode)) { struct dt_object *dt = mdd_object_child(son); const char *target_name = spec->u.sp_symname; int sym_len = strlen(target_name); const struct lu_buf *buf; loff_t pos = 0; buf = mdd_buf_get_const(env, target_name, sym_len); rc = dt->do_body_ops->dbo_write(env, dt, buf, &pos, handle, mdd_object_capa(env, son)); if (rc == sym_len) rc = 0; else rc = -EFAULT; } *la = ma->ma_attr; la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_pobj, la, handle, 0); if (rc) GOTO(cleanup, rc); /* Return attr back. */ rc = mdd_attr_get_internal_locked(env, son, ma); EXIT; cleanup: if (rc && created) { int rc2 = 0; if (inserted) { rc2 = __mdd_index_delete(env, mdd_pobj, name, S_ISDIR(attr->la_mode), handle, BYPASS_CAPA); if (rc2) CERROR("error can not cleanup destroy %d\n", rc2); } if (rc2 == 0) { mdd_write_lock(env, son); mdd_ref_del_internal(env, son, handle); mdd_write_unlock(env, son); } } /* Finish mdd_lov_create() stuff */ mdd_lov_create_finish(env, mdd, rc); if (lmm && !spec->u.sp_ea.no_lov_create) OBD_FREE(lmm, lmm_size); mdd_pdo_write_unlock(env, mdd_pobj, dlh); out_trans: mdd_trans_stop(env, mdd, rc, handle); mdd_lproc_time_end(mdd, &start, LPROC_MDD_CREATE); return rc; } /* * Get locks on parents in proper order * RETURN: < 0 - error, rename_order if successful */ enum rename_order { MDD_RN_SAME, MDD_RN_SRCTGT, MDD_RN_TGTSRC }; static int mdd_rename_order(const struct lu_env *env, struct mdd_device *mdd, struct mdd_object *src_pobj, struct mdd_object *tgt_pobj) { /* order of locking, 1 - tgt-src, 0 - src-tgt*/ int rc; ENTRY; if (src_pobj == tgt_pobj) RETURN(MDD_RN_SAME); /* compared the parent child relationship of src_p&tgt_p */ if (lu_fid_eq(&mdd->mdd_root_fid, mdo2fid(src_pobj))){ rc = MDD_RN_SRCTGT; } else if (lu_fid_eq(&mdd->mdd_root_fid, mdo2fid(tgt_pobj))) { rc = MDD_RN_TGTSRC; } else { rc = mdd_is_parent(env, mdd, src_pobj, mdo2fid(tgt_pobj), NULL); if (rc == -EREMOTE) rc = 0; if (rc == 1) rc = MDD_RN_TGTSRC; else if (rc == 0) rc = MDD_RN_SRCTGT; } RETURN(rc); } static int mdd_rename_sanity_check(const struct lu_env *env, struct mdd_object *src_pobj, struct mdd_object *tgt_pobj, const struct lu_fid *sfid, int src_is_dir, struct mdd_object *tobj) { int rc; ENTRY; if (mdd_is_dead_obj(src_pobj)) RETURN(-ENOENT); /* The sobj maybe on the remote, check parent permission only here */ rc = mdd_permission_internal_locked(env, src_pobj, MAY_WRITE | MAY_EXEC); if (rc) RETURN(rc); if (!tobj) { rc = mdd_may_create(env, tgt_pobj, NULL, (src_pobj != tgt_pobj)); } else { mdd_read_lock(env, tobj); rc = mdd_may_delete(env, tgt_pobj, tobj, src_is_dir, (src_pobj != tgt_pobj)); if (rc == 0) if (S_ISDIR(mdd_object_type(tobj)) && mdd_dir_is_empty(env, tobj)) rc = -ENOTEMPTY; mdd_read_unlock(env, tobj); } RETURN(rc); } /* src object can be remote that is why we use only fid and type of object */ static int mdd_rename(const struct lu_env *env, struct md_object *src_pobj, struct md_object *tgt_pobj, const struct lu_fid *lf, const char *sname, struct md_object *tobj, const char *tname, struct md_attr *ma) { struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; struct mdd_object *mdd_spobj = md2mdd_obj(src_pobj); struct mdd_object *mdd_tpobj = md2mdd_obj(tgt_pobj); struct mdd_device *mdd = mdo2mdd(src_pobj); struct mdd_object *mdd_sobj = NULL; struct mdd_object *mdd_tobj = NULL; struct dynlock_handle *sdlh, *tdlh; struct thandle *handle; int is_dir; int rc; ENTRY; LASSERT(ma->ma_attr.la_mode & S_IFMT); is_dir = S_ISDIR(ma->ma_attr.la_mode); if (ma->ma_attr.la_valid & LA_FLAGS && ma->ma_attr.la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL)) RETURN(-EPERM); if (tobj) mdd_tobj = md2mdd_obj(tobj); mdd_txn_param_build(env, mdd, MDD_TXN_RENAME_OP); handle = mdd_trans_start(env, mdd); if (IS_ERR(handle)) RETURN(PTR_ERR(handle)); /* FIXME: Should consider tobj and sobj too in rename_lock. */ rc = mdd_rename_order(env, mdd, mdd_spobj, mdd_tpobj); if (rc < 0) GOTO(cleanup_unlocked, rc); /* Get locks in determined order */ if (rc == MDD_RN_SAME) { sdlh = mdd_pdo_write_lock(env, mdd_spobj, sname); /* check hashes to determine do we need one lock or two */ if (mdd_name2hash(sname) != mdd_name2hash(tname)) tdlh = mdd_pdo_write_lock(env, mdd_tpobj, tname); else tdlh = sdlh; } else if (rc == MDD_RN_SRCTGT) { sdlh = mdd_pdo_write_lock(env, mdd_spobj, sname); tdlh = mdd_pdo_write_lock(env, mdd_tpobj, tname); } else { tdlh = mdd_pdo_write_lock(env, mdd_tpobj, tname); sdlh = mdd_pdo_write_lock(env, mdd_spobj, sname); } if (sdlh == NULL || tdlh == NULL) GOTO(cleanup, rc = -ENOMEM); rc = mdd_rename_sanity_check(env, mdd_spobj, mdd_tpobj, lf, is_dir, mdd_tobj); if (rc) GOTO(cleanup, rc); rc = __mdd_index_delete(env, mdd_spobj, sname, is_dir, handle, mdd_object_capa(env, mdd_spobj)); if (rc) GOTO(cleanup, rc); /* * Here tobj can be remote one, so we do index_delete unconditionally * and -ENOENT is allowed. */ rc = __mdd_index_delete(env, mdd_tpobj, tname, is_dir, handle, mdd_object_capa(env, mdd_tpobj)); if (rc != 0 && rc != -ENOENT) GOTO(cleanup, rc); rc = __mdd_index_insert(env, mdd_tpobj, lf, tname, is_dir, handle, mdd_object_capa(env, mdd_tpobj)); if (rc) GOTO(cleanup, rc); *la = ma->ma_attr; mdd_sobj = mdd_object_find(env, mdd, lf); if (mdd_sobj) { la->la_valid = LA_CTIME; /* XXX: How to update ctime for remote sobj? */ rc = mdd_attr_set_internal_locked(env, mdd_sobj, la, handle, 1); if (rc) GOTO(cleanup, rc); } if (tobj && lu_object_exists(&tobj->mo_lu)) { mdd_write_lock(env, mdd_tobj); mdd_ref_del_internal(env, mdd_tobj, handle); /* Remove dot reference. */ if (is_dir) mdd_ref_del_internal(env, mdd_tobj, handle); la->la_valid = LA_CTIME; rc = mdd_attr_set_internal(env, mdd_tobj, la, handle, 0); if (rc) GOTO(cleanup, rc); rc = mdd_finish_unlink(env, mdd_tobj, ma, handle); mdd_write_unlock(env, mdd_tobj); if (rc) GOTO(cleanup, rc); } la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_spobj, la, handle, 0); if (rc) GOTO(cleanup, rc); if (mdd_spobj != mdd_tpobj) { la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_attr_set_internal_locked(env, mdd_tpobj, la, handle, 0); } EXIT; cleanup: if (likely(tdlh) && sdlh != tdlh) mdd_pdo_write_unlock(env, mdd_tpobj, tdlh); if (likely(sdlh)) mdd_pdo_write_unlock(env, mdd_spobj, sdlh); cleanup_unlocked: mdd_trans_stop(env, mdd, rc, handle); if (mdd_sobj) mdd_object_put(env, mdd_sobj); return rc; } struct md_dir_operations mdd_dir_ops = { .mdo_is_subdir = mdd_is_subdir, .mdo_lookup = mdd_lookup, .mdo_create = mdd_create, .mdo_rename = mdd_rename, .mdo_link = mdd_link, .mdo_unlink = mdd_unlink, .mdo_name_insert = mdd_name_insert, .mdo_name_remove = mdd_name_remove, .mdo_rename_tgt = mdd_rename_tgt, .mdo_create_data = mdd_create_data };