X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Fmdd%2Fmdd_dir.c;h=a353f8bf72eacb96c73737d30f3751e71cbd5e8b;hb=3442db6faf685fbdbd092bdfdc8d273e4990a141;hp=d6bf64b109f11b0050bb57aa9027d7f9b2063981;hpb=1f55f2a9071d5e7db4042b959723086dee1c379a;p=fs%2Flustre-release.git diff --git a/lustre/mdd/mdd_dir.c b/lustre/mdd/mdd_dir.c index d6bf64b..a353f8b 100644 --- a/lustre/mdd/mdd_dir.c +++ b/lustre/mdd/mdd_dir.c @@ -15,11 +15,7 @@ * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see - * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ @@ -27,7 +23,7 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, 2013, Intel Corporation. + * Copyright (c) 2011, 2016, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -53,10 +49,21 @@ static const char dot[] = "."; static const char dotdot[] = ".."; static struct lu_name lname_dotdot = { - (char *) dotdot, - sizeof(dotdot) - 1 + .ln_name = (char *) dotdot, + .ln_namelen = sizeof(dotdot) - 1, }; +static inline int +mdd_name_check(struct mdd_device *m, const struct lu_name *ln) +{ + if (!lu_name_is_valid(ln)) + return -EINVAL; + else if (ln->ln_namelen > m->mdd_dt_conf.ddp_max_name_len) + return -ENAMETOOLONG; + else + return 0; +} + /* Get FID from name and parent */ static int __mdd_lookup(const struct lu_env *env, struct md_object *pobj, @@ -74,33 +81,23 @@ __mdd_lookup(const struct lu_env *env, struct md_object *pobj, if (unlikely(mdd_is_dead_obj(mdd_obj))) RETURN(-ESTALE); + if (!mdd_object_exists(mdd_obj)) + RETURN(-ESTALE); + if (mdd_object_remote(mdd_obj)) { CDEBUG(D_INFO, "%s: Object "DFID" locates on remote server\n", mdd2obd_dev(m)->obd_name, PFID(mdo2fid(mdd_obj))); - } else if (!mdd_object_exists(mdd_obj)) { - RETURN(-ESTALE); } - /* The common filename length check. */ - if (unlikely(lname->ln_namelen > m->mdd_dt_conf.ddp_max_name_len)) - RETURN(-ENAMETOOLONG); - rc = mdd_permission_internal_locked(env, mdd_obj, pattr, mask, MOR_TGT_PARENT); if (rc) RETURN(rc); if (likely(S_ISDIR(mdd_object_type(mdd_obj)) && - dt_try_as_dir(env, dir))) { - - rc = dir->do_index_ops->dio_lookup(env, dir, - (struct dt_rec *)fid, key, - mdd_object_capa(env, mdd_obj)); - if (rc > 0) - rc = 0; - else if (rc == 0) - rc = -ENOENT; - } else + dt_try_as_dir(env, dir))) + rc = dt_lookup(env, dir, (struct dt_rec *)fid, key); + else rc = -ENOTDIR; RETURN(rc); @@ -114,20 +111,137 @@ int mdd_lookup(const struct lu_env *env, int rc; ENTRY; - rc = mdd_la_get(env, md2mdd_obj(pobj), pattr, BYPASS_CAPA); + rc = mdd_la_get(env, md2mdd_obj(pobj), pattr); if (rc != 0) RETURN(rc); - rc = __mdd_lookup(env, pobj, pattr, lname, fid, MAY_EXEC); + rc = __mdd_lookup(env, pobj, pattr, lname, fid, + (spec != NULL && spec->sp_permitted) ? 0 : MAY_EXEC); RETURN(rc); } +/** Read the link EA into a temp buffer. + * Uses the mdd_thread_info::mti_big_buf since it is generally large. + * A pointer to the buffer is stored in \a ldata::ld_buf. + * + * \retval 0 or error + */ +static int __mdd_links_read(const struct lu_env *env, + struct mdd_object *mdd_obj, + struct linkea_data *ldata) +{ + int rc; + + if (!mdd_object_exists(mdd_obj)) + return -ENODATA; + + /* First try a small buf */ + LASSERT(env != NULL); + ldata->ld_buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mti_link_buf, + PAGE_SIZE); + if (ldata->ld_buf->lb_buf == NULL) + return -ENOMEM; + + rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, XATTR_NAME_LINK); + if (rc == -ERANGE) { + /* Buf was too small, figure out what we need. */ + lu_buf_free(ldata->ld_buf); + rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, + XATTR_NAME_LINK); + if (rc < 0) + return rc; + ldata->ld_buf = lu_buf_check_and_alloc(ldata->ld_buf, rc); + if (ldata->ld_buf->lb_buf == NULL) + return -ENOMEM; + rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, + XATTR_NAME_LINK); + } + if (rc < 0) { + lu_buf_free(ldata->ld_buf); + ldata->ld_buf = NULL; + return rc; + } + + return linkea_init(ldata); +} + +static int mdd_links_read(const struct lu_env *env, + struct mdd_object *mdd_obj, + struct linkea_data *ldata) +{ + int rc; + + rc = __mdd_links_read(env, mdd_obj, ldata); + if (!rc) + rc = linkea_init(ldata); + + return rc; +} + +static int mdd_links_read_with_rec(const struct lu_env *env, + struct mdd_object *mdd_obj, + struct linkea_data *ldata) +{ + int rc; + + rc = __mdd_links_read(env, mdd_obj, ldata); + if (!rc) + rc = linkea_init_with_rec(ldata); + + return rc; +} + +/** + * Get parent FID of the directory + * + * Read parent FID from linkEA, if that fails, then do lookup + * dotdot to get the parent FID. + * + * \param[in] env execution environment + * \param[in] obj object from which to find the parent FID + * \param[in] attr attribute of the object + * \param[out] fid fid to get the parent FID + * + * \retval 0 if getting the parent FID succeeds. + * \retval negative errno if getting the parent FID fails. + **/ static inline int mdd_parent_fid(const struct lu_env *env, struct mdd_object *obj, const struct lu_attr *attr, struct lu_fid *fid) { - return __mdd_lookup(env, &obj->mod_obj, attr, &lname_dotdot, fid, 0); + struct mdd_thread_info *info = mdd_env_info(env); + struct linkea_data ldata = { NULL }; + struct lu_buf *buf = &info->mti_link_buf; + struct lu_name lname; + int rc = 0; + + ENTRY; + + LASSERT(S_ISDIR(mdd_object_type(obj))); + + buf = lu_buf_check_and_alloc(buf, PATH_MAX); + if (buf->lb_buf == NULL) + GOTO(lookup, rc = 0); + + ldata.ld_buf = buf; + rc = mdd_links_read_with_rec(env, obj, &ldata); + if (rc != 0) + GOTO(lookup, rc); + + LASSERT(ldata.ld_leh != NULL); + /* Directory should only have 1 parent */ + if (ldata.ld_leh->leh_reccount > 1) + GOTO(lookup, rc); + + ldata.ld_lee = (struct link_ea_entry *)(ldata.ld_leh + 1); + + linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, &lname, fid); + if (likely(fid_is_sane(fid))) + RETURN(0); +lookup: + rc = __mdd_lookup(env, &obj->mod_obj, attr, &lname_dotdot, fid, 0); + RETURN(rc); } /* @@ -169,27 +283,26 @@ static int mdd_is_parent(const struct lu_env *env, RETURN(0); for(;;) { - /* this is done recursively, bypass capa for each obj */ - mdd_set_capainfo(env, 4, p1, BYPASS_CAPA); + /* this is done recursively */ rc = mdd_parent_fid(env, p1, attr, pfid); if (rc) GOTO(out, rc); if (mdd_is_root(mdd, pfid)) GOTO(out, rc = 0); + if (lu_fid_eq(pfid, &mdd->mdd_local_root_fid)) + GOTO(out, rc = 0); if (lu_fid_eq(pfid, lf)) GOTO(out, rc = 1); - if (parent) - mdd_object_put(env, parent); + if (parent != NULL) + mdd_object_put(env, parent); parent = mdd_object_find(env, mdd, pfid); - if (IS_ERR(parent)) { + if (IS_ERR(parent)) GOTO(out, rc = PTR_ERR(parent)); - } else if (mdd_object_remote(parent)) { - /*FIXME: Because of the restriction of rename in Phase I. - * If the parent is remote, we just assumed lf is not the - * parent of P1 for now */ - GOTO(out, rc = 0); - } + + if (!mdd_object_exists(parent)) + GOTO(out, rc = -EINVAL); + p1 = parent; } EXIT; @@ -203,7 +316,7 @@ out: * No permission check is needed. * * returns 1: if fid is ancestor of @mo; - * returns 0: if fid is not a ancestor of @mo; + * returns 0: if fid is not an ancestor of @mo; * * returns EREMOTE if remote object is found, fid of remote object is saved to * @fid; @@ -221,7 +334,7 @@ int mdd_is_subdir(const struct lu_env *env, struct md_object *mo, if (!S_ISDIR(mdd_object_type(md2mdd_obj(mo)))) RETURN(0); - rc = mdd_la_get(env, md2mdd_obj(mo), attr, BYPASS_CAPA); + rc = mdd_la_get(env, md2mdd_obj(mo), attr); if (rc != 0) RETURN(rc); @@ -251,41 +364,55 @@ int mdd_is_subdir(const struct lu_env *env, struct md_object *mo, static int mdd_dir_is_empty(const struct lu_env *env, struct mdd_object *dir) { - struct dt_it *it; - struct dt_object *obj; - const struct dt_it_ops *iops; - int result; - ENTRY; + struct dt_it *it; + struct dt_object *obj; + const struct dt_it_ops *iops; + int result; + ENTRY; - obj = mdd_object_child(dir); - if (!dt_try_as_dir(env, obj)) - RETURN(-ENOTDIR); - - iops = &obj->do_index_ops->dio_it; - it = iops->init(env, obj, LUDA_64BITHASH, BYPASS_CAPA); - if (!IS_ERR(it)) { - 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 = PTR_ERR(it); - RETURN(result); + obj = mdd_object_child(dir); + if (!dt_try_as_dir(env, obj)) + RETURN(-ENOTDIR); + + iops = &obj->do_index_ops->dio_it; + it = iops->init(env, obj, LUDA_64BITHASH); + if (!IS_ERR(it)) { + result = iops->get(env, it, (const struct dt_key *)""); + 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 = PTR_ERR(it); + RETURN(result); } +/** + * Determine if the target object can be hard linked, and right now it only + * checks if the link count reach the maximum limit. Note: for ldiskfs, the + * directory nlink count might exceed the maximum link count(see + * osd_object_ref_add), so it only check nlink for non-directories. + * + * \param[in] env thread environment + * \param[in] obj object being linked to + * \param[in] la attributes of \a obj + * + * \retval 0 if \a obj can be hard linked + * \retval negative error if \a obj is a directory or has too + * many links + */ static int __mdd_may_link(const struct lu_env *env, struct mdd_object *obj, const struct lu_attr *la) { @@ -294,25 +421,31 @@ static int __mdd_may_link(const struct lu_env *env, struct mdd_object *obj, LASSERT(la != NULL); - if (!S_ISDIR(la->la_mode)) - RETURN(0); - - /* - * Subdir count limitation can be broken through. - */ - if (la->la_nlink >= m->mdd_dt_conf.ddp_max_nlink) + /* Subdir count limitation can be broken through + * (see osd_object_ref_add), so only check non-directory here. */ + if (!S_ISDIR(la->la_mode) && + la->la_nlink >= m->mdd_dt_conf.ddp_max_nlink) RETURN(-EMLINK); - else - RETURN(0); + + RETURN(0); } -/* +/** * Check whether it may create the cobj under the pobj. - * cobj maybe NULL + * + * \param[in] env execution environment + * \param[in] pobj the parent directory + * \param[in] pattr the attribute of the parent directory + * \param[in] cobj the child to be created + * \param[in] check_perm if check WRITE|EXEC permission for parent + * + * \retval = 0 create the child under this dir is allowed + * \retval negative errno create the child under this dir is + * not allowed */ -int mdd_may_create(const struct lu_env *env, - struct mdd_object *pobj, const struct lu_attr *pattr, - struct mdd_object *cobj, int check_perm, int check_nlink) +int mdd_may_create(const struct lu_env *env, struct mdd_object *pobj, + const struct lu_attr *pattr, struct mdd_object *cobj, + bool check_perm) { int rc = 0; ENTRY; @@ -327,9 +460,6 @@ int mdd_may_create(const struct lu_env *env, rc = mdd_permission_internal_locked(env, pobj, pattr, MAY_WRITE | MAY_EXEC, MOR_TGT_PARENT); - if (!rc && check_nlink) - rc = __mdd_may_link(env, pobj, pattr); - RETURN(rc); } @@ -345,17 +475,16 @@ int mdd_may_unlink(const struct lu_env *env, struct mdd_object *pobj, if (mdd_is_dead_obj(pobj)) RETURN(-ENOENT); - if ((attr->la_valid & LA_FLAGS) && - (attr->la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL))) + if (attr->la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL)) RETURN(-EPERM); rc = mdd_permission_internal_locked(env, pobj, pattr, MAY_WRITE | MAY_EXEC, MOR_TGT_PARENT); - if (rc) + if (rc != 0) RETURN(rc); - if (mdd_is_append(pobj)) + if (pattr->la_flags & LUSTRE_APPEND_FL) RETURN(-EPERM); RETURN(rc); @@ -410,7 +539,7 @@ static int mdd_may_delete_entry(const struct lu_env *env, RETURN(rc); } - if (mdd_is_append(pobj)) + if (pattr->la_flags & LUSTRE_APPEND_FL) RETURN(-EPERM); RETURN(0); @@ -447,11 +576,7 @@ int mdd_may_delete(const struct lu_env *env, struct mdd_object *tpobj, if (mdd_is_sticky(env, tpobj, tpattr, tobj, tattr)) RETURN(-EPERM); - if (mdd_is_immutable(tobj) || mdd_is_append(tobj)) - RETURN(-EPERM); - - if ((tattr->la_valid & LA_FLAGS) && - (tattr->la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL))) + if (tattr->la_flags & (LUSTRE_APPEND_FL | LUSTRE_IMMUTABLE_FL)) RETURN(-EPERM); /* additional check the rename case */ @@ -474,9 +599,20 @@ int mdd_may_delete(const struct lu_env *env, struct mdd_object *tpobj, RETURN(rc); } -/* - * tgt maybe NULL - * has mdd_write_lock on src already, but not on tgt yet +/** + * Check whether it can create the link file(linked to @src_obj) under + * the target directory(@tgt_obj), and src_obj has been locked by + * mdd_write_lock. + * + * \param[in] env execution environment + * \param[in] tgt_obj the target directory + * \param[in] tattr attributes of target directory + * \param[in] lname the link name + * \param[in] src_obj source object for link + * \param[in] cattr attributes for source object + * + * \retval = 0 it is allowed to create the link file under tgt_obj + * \retval negative error not allowed to create the link file */ static int mdd_link_sanity_check(const struct lu_env *env, struct mdd_object *tgt_obj, @@ -496,21 +632,20 @@ static int mdd_link_sanity_check(const struct lu_env *env, RETURN(-ESTALE); /* Local ops, no lookup before link, check filename length here. */ - if (lname && (lname->ln_namelen > m->mdd_dt_conf.ddp_max_name_len)) - RETURN(-ENAMETOOLONG); + rc = mdd_name_check(m, lname); + if (rc < 0) + RETURN(rc); - if (mdd_is_immutable(src_obj) || mdd_is_append(src_obj)) - RETURN(-EPERM); + if (cattr->la_flags & (LUSTRE_IMMUTABLE_FL | LUSTRE_APPEND_FL)) + RETURN(-EPERM); - if (S_ISDIR(mdd_object_type(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, tattr, NULL, 1, 0); - if (rc) - RETURN(rc); - } + rc = mdd_may_create(env, tgt_obj, tattr, NULL, true); + if (rc != 0) + RETURN(rc); rc = __mdd_may_link(env, src_obj, cattr); @@ -518,42 +653,41 @@ static int mdd_link_sanity_check(const struct lu_env *env, } static int __mdd_index_delete_only(const struct lu_env *env, struct mdd_object *pobj, - const char *name, struct thandle *handle, - struct lustre_capa *capa) + const char *name, struct thandle *handle) { - struct dt_object *next = mdd_object_child(pobj); - int rc; - ENTRY; + struct dt_object *next = mdd_object_child(pobj); + int rc; + ENTRY; - if (dt_try_as_dir(env, next)) { - rc = next->do_index_ops->dio_delete(env, next, - (struct dt_key *)name, - handle, capa); - } else - rc = -ENOTDIR; + if (dt_try_as_dir(env, next)) + rc = dt_delete(env, next, (struct dt_key *)name, handle); + else + rc = -ENOTDIR; - RETURN(rc); + 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) + const struct lu_fid *lf, __u32 type, + const char *name, + struct thandle *handle) { struct dt_object *next = mdd_object_child(pobj); int rc; ENTRY; if (dt_try_as_dir(env, next)) { - struct lu_ucred *uc = lu_ucred_check(env); - int ignore_quota; + struct dt_insert_rec *rec = &mdd_env_info(env)->mti_dt_rec; + struct lu_ucred *uc = lu_ucred_check(env); + int ignore_quota; + rec->rec_fid = lf; + rec->rec_type = type; ignore_quota = uc ? uc->uc_cap & CFS_CAP_SYS_RESOURCE_MASK : 1; - rc = next->do_index_ops->dio_insert(env, next, - (struct dt_rec*)lf, - (const struct dt_key *)name, - handle, capa, ignore_quota); + rc = dt_insert(env, next, (const struct dt_rec *)rec, + (const struct dt_key *)name, handle, + ignore_quota); } else { rc = -ENOTDIR; } @@ -562,30 +696,31 @@ static int __mdd_index_insert_only(const struct lu_env *env, /* insert named index, add reference if isdir */ 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) + const struct lu_fid *lf, __u32 type, + const char *name, struct thandle *handle) { - int rc; - ENTRY; + int rc; + ENTRY; - rc = __mdd_index_insert_only(env, pobj, lf, name, handle, capa); - if (rc == 0 && is_dir) { - mdd_write_lock(env, pobj, MOR_TGT_PARENT); - mdo_ref_add(env, pobj, handle); - mdd_write_unlock(env, pobj); - } - RETURN(rc); + rc = __mdd_index_insert_only(env, pobj, lf, type, name, handle); + if (rc == 0 && S_ISDIR(type)) { + mdd_write_lock(env, pobj, MOR_TGT_PARENT); + mdo_ref_add(env, pobj, handle); + mdd_write_unlock(env, pobj); + } + + RETURN(rc); } /* delete named index, drop reference if isdir */ 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) + const char *name, int is_dir, + struct thandle *handle) { int rc; ENTRY; - rc = __mdd_index_delete_only(env, pobj, name, handle, capa); + rc = __mdd_index_delete_only(env, pobj, name, handle); if (rc == 0 && is_dir) { mdd_write_lock(env, pobj, MOR_TGT_PARENT); mdo_ref_del(env, pobj, handle); @@ -595,15 +730,37 @@ static int __mdd_index_delete(const struct lu_env *env, struct mdd_object *pobj, RETURN(rc); } +static int mdd_llog_record_calc_size(const struct lu_env *env, + const struct lu_name *tname, + const struct lu_name *sname) +{ + const struct lu_ucred *uc = lu_ucred(env); + enum changelog_rec_flags crf = CLF_EXTRA_FLAGS; + enum changelog_rec_extra_flags crfe = CLFE_UIDGID; + + if (sname != NULL) + crf |= CLF_RENAME; + + if (uc != NULL && uc->uc_jobid[0] != '\0') + crf |= CLF_JOBID; + + return llog_data_len(LLOG_CHANGELOG_HDR_SZ + + changelog_rec_offset(crf, crfe) + + (tname != NULL ? tname->ln_namelen : 0) + + (sname != NULL ? 1 + sname->ln_namelen : 0)); +} + int mdd_declare_changelog_store(const struct lu_env *env, struct mdd_device *mdd, - const struct lu_name *fname, + const struct lu_name *tname, + const struct lu_name *sname, struct thandle *handle) { struct obd_device *obd = mdd2obd_dev(mdd); struct llog_ctxt *ctxt; struct llog_changelog_rec *rec; struct lu_buf *buf; + struct thandle *llog_th; int reclen; int rc; @@ -611,8 +768,7 @@ int mdd_declare_changelog_store(const struct lu_env *env, if (!(mdd->mdd_cl.mc_flags & CLM_ON)) return 0; - reclen = llog_data_len(sizeof(*rec) + - (fname != NULL ? fname->ln_namelen : 0)); + reclen = mdd_llog_record_calc_size(env, tname, sname); buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mti_big_buf, reclen); if (buf->lb_buf == NULL) return -ENOMEM; @@ -625,47 +781,145 @@ int mdd_declare_changelog_store(const struct lu_env *env, if (ctxt == NULL) return -ENXIO; - rc = llog_declare_add(env, ctxt->loc_handle, &rec->cr_hdr, handle); + llog_th = thandle_get_sub(env, handle, ctxt->loc_handle->lgh_obj); + if (IS_ERR(llog_th)) + GOTO(out_put, rc = PTR_ERR(llog_th)); + + rc = llog_declare_add(env, ctxt->loc_handle, &rec->cr_hdr, llog_th); + +out_put: llog_ctxt_put(ctxt); return rc; } -static int mdd_declare_changelog_ext_store(const struct lu_env *env, - struct mdd_device *mdd, - const struct lu_name *tname, - const struct lu_name *sname, - struct thandle *handle) +struct mdd_changelog_gc { + struct mdd_device *mcgc_mdd; + bool mcgc_found; + __u32 mcgc_maxtime; + __u64 mcgc_maxindexes; + __u32 mcgc_id; +}; + +/* return first registered ChangeLog user idle since too long + * use ChangeLog's user plain LLOG mtime for this */ +static int mdd_changelog_gc_cb(const struct lu_env *env, + struct llog_handle *llh, + struct llog_rec_hdr *hdr, void *data) { - struct obd_device *obd = mdd2obd_dev(mdd); - struct llog_ctxt *ctxt; - struct llog_changelog_ext_rec *rec; - struct lu_buf *buf; - int reclen; - int rc; + struct llog_changelog_user_rec *rec; + struct mdd_changelog_gc *mcgc = (struct mdd_changelog_gc *)data; + struct mdd_device *mdd = mcgc->mcgc_mdd; + ENTRY; - /* Not recording */ - if (!(mdd->mdd_cl.mc_flags & CLM_ON)) - return 0; + if ((llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN) == 0) + RETURN(-ENXIO); - reclen = llog_data_len(sizeof(*rec) + - (tname != NULL ? tname->ln_namelen : 0) + - (sname != NULL ? 1 + sname->ln_namelen : 0)); - buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mti_big_buf, reclen); - if (buf->lb_buf == NULL) - return -ENOMEM; + rec = container_of(hdr, struct llog_changelog_user_rec, + cur_hdr); - rec = buf->lb_buf; - rec->cr_hdr.lrh_len = reclen; - rec->cr_hdr.lrh_type = CHANGELOG_REC; + /* find oldest idle user, based on last record update/cancel time (new + * behavior), or for old user records, last record index vs current + * ChangeLog index. Late users with old record format will be treated + * first as we assume they could be idle since longer + */ + if (rec->cur_time != 0) { + __u32 time_now = (__u32)get_seconds(); + __u32 time_out = rec->cur_time + + mdd->mdd_changelog_max_idle_time; + __u32 idle_time = time_now - rec->cur_time; + + /* treat oldest idle user first, and if no old format user + * has been already selected + */ + if (time_after32(time_now, time_out) && + idle_time > mcgc->mcgc_maxtime && + mcgc->mcgc_maxindexes == 0) { + mcgc->mcgc_maxtime = idle_time; + mcgc->mcgc_id = rec->cur_id; + mcgc->mcgc_found = true; + } + } else { + /* old user record with no idle time stamp, so use empirical + * method based on its current index/position + */ + __u64 idle_indexes; - ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT); - if (ctxt == NULL) - return -ENXIO; + idle_indexes = mdd->mdd_cl.mc_index - rec->cur_endrec; - rc = llog_declare_add(env, ctxt->loc_handle, &rec->cr_hdr, handle); - llog_ctxt_put(ctxt); + /* treat user with the oldest/smallest current index first */ + if (idle_indexes >= mdd->mdd_changelog_max_idle_indexes && + idle_indexes > mcgc->mcgc_maxindexes) { + mcgc->mcgc_maxindexes = idle_indexes; + mcgc->mcgc_id = rec->cur_id; + mcgc->mcgc_found = true; + } + + } + RETURN(0); +} + +/* recover space from long-term inactive ChangeLog users */ +static int mdd_chlg_garbage_collect(void *data) +{ + struct mdd_device *mdd = (struct mdd_device *)data; + struct lu_env *env = NULL; + int rc; + struct llog_ctxt *ctxt; + struct mdd_changelog_gc mcgc = { + .mcgc_mdd = mdd, + .mcgc_found = false, + .mcgc_maxtime = 0, + .mcgc_maxindexes = 0, + }; + ENTRY; + + CDEBUG(D_HA, "%s: ChangeLog garbage collect thread start\n", + mdd2obd_dev(mdd)->obd_name); + + OBD_ALLOC_PTR(env); + if (env == NULL) + GOTO(out, rc = -ENOMEM); + rc = lu_env_init(env, LCT_MD_THREAD); + if (rc) + GOTO(out, rc); + + for (;;) { + ctxt = llog_get_context(mdd2obd_dev(mdd), + LLOG_CHANGELOG_USER_ORIG_CTXT); + if (ctxt == NULL || + (ctxt->loc_handle->lgh_hdr->llh_flags & LLOG_F_IS_CAT) == 0) + GOTO(out_env, rc = -ENXIO); + + rc = llog_cat_process(env, ctxt->loc_handle, + mdd_changelog_gc_cb, &mcgc, 0, 0); + if (rc != 0 || mcgc.mcgc_found == false) + break; + llog_ctxt_put(ctxt); + + CWARN("%s: Force deregister of ChangeLog user cl%d idle more " + "than %us\n", mdd2obd_dev(mdd)->obd_name, mcgc.mcgc_id, + mcgc.mcgc_maxtime); + + mdd_changelog_user_purge(env, mdd, mcgc.mcgc_id); + + /* try again to search for another candidate */ + mcgc.mcgc_found = false; + mcgc.mcgc_maxtime = 0; + mcgc.mcgc_maxindexes = 0; + } + +out_env: + if (ctxt != NULL) + llog_ctxt_put(ctxt); + + lu_env_fini(env); + GOTO(out, rc); +out: + if (env) + OBD_FREE_PTR(env); + mdd->mdd_cl.mc_gc_task = NULL; return rc; } @@ -673,7 +927,7 @@ static int mdd_declare_changelog_ext_store(const struct lu_env *env, * \param mdd * \param rec * \param handle - currently ignored since llogs start their own transaction; - * this will hopefully be fixed in llog rewrite + * this will hopefully be fixed in llog rewrite * \retval 0 ok */ int mdd_changelog_store(const struct lu_env *env, struct mdd_device *mdd, @@ -681,9 +935,14 @@ int mdd_changelog_store(const struct lu_env *env, struct mdd_device *mdd, { struct obd_device *obd = mdd2obd_dev(mdd); struct llog_ctxt *ctxt; + struct thandle *llog_th; int rc; + bool run_gc_task; - rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) + rec->cr.cr_namelen); + rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) + + changelog_rec_varsize(&rec->cr)); + + /* llog_lvfs_write_rec sets the llog tail len */ rec->cr_hdr.lrh_type = CHANGELOG_REC; rec->cr.cr_time = cl_time(); @@ -698,182 +957,195 @@ int mdd_changelog_store(const struct lu_env *env, struct mdd_device *mdd, if (ctxt == NULL) return -ENXIO; - rc = llog_add(env, ctxt->loc_handle, &rec->cr_hdr, NULL, NULL, th); + llog_th = thandle_get_sub(env, th, ctxt->loc_handle->lgh_obj); + if (IS_ERR(llog_th)) + GOTO(out_put, rc = PTR_ERR(llog_th)); + + /* nested journal transaction */ + rc = llog_add(env, ctxt->loc_handle, &rec->cr_hdr, NULL, llog_th); + + /* time to recover some space ?? */ + spin_lock(&mdd->mdd_cl.mc_lock); + if (unlikely(mdd->mdd_changelog_gc && (ktime_get_real_seconds() - + mdd->mdd_cl.mc_gc_time > mdd->mdd_changelog_min_gc_interval) && + mdd->mdd_cl.mc_gc_task == NULL && + llog_cat_free_space(ctxt->loc_handle) <= + mdd->mdd_changelog_min_free_cat_entries)) { + CWARN("%s: low on changelog_catalog free entries, starting " + "ChangeLog garbage collection thread\n", obd->obd_name); + + /* indicate further kthread run will occur outside right after + * critical section + */ + mdd->mdd_cl.mc_gc_task = (struct task_struct *)(-1); + run_gc_task = true; + } + spin_unlock(&mdd->mdd_cl.mc_lock); + if (run_gc_task) { + struct task_struct *gc_task; + + gc_task = kthread_run(mdd_chlg_garbage_collect, mdd, + "chlg_gc_thread"); + if (IS_ERR(gc_task)) { + CERROR("%s: cannot start ChangeLog garbage collection " + "thread: rc = %ld\n", obd->obd_name, + PTR_ERR(gc_task)); + mdd->mdd_cl.mc_gc_task = NULL; + } else { + CDEBUG(D_HA, "%s: ChangeLog garbage collection thread " + "has started with Pid %d\n", obd->obd_name, + gc_task->pid); + mdd->mdd_cl.mc_gc_task = gc_task; + mdd->mdd_cl.mc_gc_time = ktime_get_real_seconds(); + } + } +out_put: llog_ctxt_put(ctxt); if (rc > 0) rc = 0; return rc; } -/** Add a changelog_ext entry \a rec to the changelog llog - * \param mdd - * \param rec - * \param handle - currently ignored since llogs start their own transaction; - * this will hopefully be fixed in llog rewrite - * \retval 0 ok - */ -int mdd_changelog_ext_store(const struct lu_env *env, struct mdd_device *mdd, - struct llog_changelog_ext_rec *rec, - struct thandle *th) +static void mdd_changelog_rec_ext_rename(struct changelog_rec *rec, + const struct lu_fid *sfid, + const struct lu_fid *spfid, + const struct lu_name *sname) { - struct obd_device *obd = mdd2obd_dev(mdd); - struct llog_ctxt *ctxt; - int rc; + struct changelog_ext_rename *rnm = changelog_rec_rename(rec); + size_t extsize = sname->ln_namelen + 1; - rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) + rec->cr.cr_namelen); - /* llog_lvfs_write_rec sets the llog tail len */ - rec->cr_hdr.lrh_type = CHANGELOG_REC; - rec->cr.cr_time = cl_time(); + LASSERT(sfid != NULL); + LASSERT(spfid != NULL); + LASSERT(sname != NULL); - spin_lock(&mdd->mdd_cl.mc_lock); - /* NB: I suppose it's possible llog_add adds out of order wrt cr_index, - * but as long as the MDD transactions are ordered correctly for e.g. - * rename conflicts, I don't think this should matter. */ - rec->cr.cr_index = ++mdd->mdd_cl.mc_index; - spin_unlock(&mdd->mdd_cl.mc_lock); + rnm->cr_sfid = *sfid; + rnm->cr_spfid = *spfid; - ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT); - if (ctxt == NULL) - return -ENXIO; + changelog_rec_name(rec)[rec->cr_namelen] = '\0'; + strlcpy(changelog_rec_sname(rec), sname->ln_name, extsize); + rec->cr_namelen += extsize; +} - /* nested journal transaction */ - rc = llog_add(env, ctxt->loc_handle, &rec->cr_hdr, NULL, NULL, th); - llog_ctxt_put(ctxt); - if (rc > 0) - rc = 0; +void mdd_changelog_rec_ext_jobid(struct changelog_rec *rec, const char *jobid) +{ + struct changelog_ext_jobid *jid = changelog_rec_jobid(rec); - return rc; + if (jobid == NULL || jobid[0] == '\0') + return; + + strlcpy(jid->cr_jobid, jobid, sizeof(jid->cr_jobid)); +} + +void mdd_changelog_rec_ext_extra_flags(struct changelog_rec *rec, __u64 eflags) +{ + struct changelog_ext_extra_flags *ef = changelog_rec_extra_flags(rec); + + ef->cr_extra_flags = eflags; +} + +void mdd_changelog_rec_extra_uidgid(struct changelog_rec *rec, + __u64 uid, __u64 gid) +{ + struct changelog_ext_uidgid *uidgid = changelog_rec_uidgid(rec); + + uidgid->cr_uid = uid; + uidgid->cr_gid = gid; } /** Store a namespace change changelog record * If this fails, we must fail the whole transaction; we don't * want the change to commit without the log entry. * \param target - mdd_object of change - * \param parent - parent dir/object + * \param tpfid - target parent dir/object fid + * \param sfid - source object fid + * \param spfid - source parent fid * \param tname - target name string - * \param handle - transacion handle + * \param sname - source name string + * \param handle - transaction handle */ -int mdd_changelog_ns_store(const struct lu_env *env, struct mdd_device *mdd, - enum changelog_rec_type type, unsigned flags, - struct mdd_object *target, struct mdd_object *parent, - const struct lu_name *tname, struct thandle *handle) -{ - struct llog_changelog_rec *rec; - struct lu_buf *buf; - int reclen; - int rc; - ENTRY; - +int mdd_changelog_ns_store(const struct lu_env *env, + struct mdd_device *mdd, + enum changelog_rec_type type, + enum changelog_rec_flags crf, + struct mdd_object *target, + const struct lu_fid *tpfid, + const struct lu_fid *sfid, + const struct lu_fid *spfid, + const struct lu_name *tname, + const struct lu_name *sname, + struct thandle *handle) +{ + const struct lu_ucred *uc = lu_ucred(env); + struct llog_changelog_rec *rec; + struct lu_buf *buf; + int reclen; + __u64 xflags = CLFE_INVALID; + int rc; + ENTRY; + /* Not recording */ if (!(mdd->mdd_cl.mc_flags & CLM_ON)) RETURN(0); + if ((mdd->mdd_cl.mc_mask & (1 << type)) == 0) RETURN(0); - LASSERT(target != NULL); - LASSERT(parent != NULL); + LASSERT(tpfid != NULL); LASSERT(tname != NULL); LASSERT(handle != NULL); - reclen = llog_data_len(sizeof(*rec) + tname->ln_namelen); + reclen = mdd_llog_record_calc_size(env, tname, sname); buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mti_big_buf, reclen); if (buf->lb_buf == NULL) RETURN(-ENOMEM); rec = buf->lb_buf; - rec->cr.cr_flags = CLF_VERSION | (CLF_FLAGMASK & flags); - rec->cr.cr_type = (__u32)type; - rec->cr.cr_tfid = *mdo2fid(target); - rec->cr.cr_pfid = *mdo2fid(parent); - rec->cr.cr_namelen = tname->ln_namelen; - memcpy(rec->cr.cr_name, tname->ln_name, tname->ln_namelen); - - target->mod_cltime = cfs_time_current_64(); - - rc = mdd_changelog_store(env, mdd, rec, handle); - if (rc < 0) { - CERROR("changelog failed: rc=%d, op%d %s c"DFID" p"DFID"\n", - rc, type, tname->ln_name, PFID(&rec->cr.cr_tfid), - PFID(&rec->cr.cr_pfid)); - RETURN(-EFAULT); - } - - RETURN(0); -} + crf &= CLF_FLAGMASK; + crf |= CLF_EXTRA_FLAGS; + if (uc != NULL && uc->uc_jobid[0] != '\0') + crf |= CLF_JOBID; -/** Store a namespace change changelog record - * If this fails, we must fail the whole transaction; we don't - * want the change to commit without the log entry. - * \param target - mdd_object of change - * \param tpfid - target parent dir/object fid - * \param sfid - source object fid - * \param spfid - source parent fid - * \param tname - target name string - * \param sname - source name string - * \param handle - transacion handle - */ -static int mdd_changelog_ext_ns_store(const struct lu_env *env, - struct mdd_device *mdd, - enum changelog_rec_type type, - unsigned flags, - struct mdd_object *target, - const struct lu_fid *tpfid, - const struct lu_fid *sfid, - const struct lu_fid *spfid, - const struct lu_name *tname, - const struct lu_name *sname, - struct thandle *handle) -{ - struct llog_changelog_ext_rec *rec; - struct lu_buf *buf; - int reclen; - int rc; - ENTRY; + if (sname != NULL) + crf |= CLF_RENAME; + else + crf |= CLF_VERSION; - /* Not recording */ - if (!(mdd->mdd_cl.mc_flags & CLM_ON)) - RETURN(0); - if ((mdd->mdd_cl.mc_mask & (1 << type)) == 0) - RETURN(0); + xflags |= CLFE_UIDGID; - LASSERT(sfid != NULL); - LASSERT(tpfid != NULL); - LASSERT(tname != NULL); - LASSERT(handle != NULL); + rec->cr.cr_flags = crf; - reclen = llog_data_len(sizeof(*rec) + - sname != NULL ? 1 + sname->ln_namelen : 0); - buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mti_big_buf, reclen); - if (buf->lb_buf == NULL) - RETURN(-ENOMEM); - rec = buf->lb_buf; + if (crf & CLF_EXTRA_FLAGS) { + mdd_changelog_rec_ext_extra_flags(&rec->cr, xflags); + if (xflags & CLFE_UIDGID) + mdd_changelog_rec_extra_uidgid(&rec->cr, + uc->uc_uid, uc->uc_gid); + } - rec->cr.cr_flags = CLF_EXT_VERSION | (CLF_FLAGMASK & flags); rec->cr.cr_type = (__u32)type; rec->cr.cr_pfid = *tpfid; - rec->cr.cr_sfid = *sfid; - rec->cr.cr_spfid = *spfid; rec->cr.cr_namelen = tname->ln_namelen; - memcpy(rec->cr.cr_name, tname->ln_name, tname->ln_namelen); - if (sname) { - rec->cr.cr_name[tname->ln_namelen] = '\0'; - memcpy(rec->cr.cr_name + tname->ln_namelen + 1, sname->ln_name, - sname->ln_namelen); - rec->cr.cr_namelen += 1 + sname->ln_namelen; - } + memcpy(changelog_rec_name(&rec->cr), tname->ln_name, tname->ln_namelen); + + if (crf & CLF_RENAME) + mdd_changelog_rec_ext_rename(&rec->cr, sfid, spfid, sname); + + if (crf & CLF_JOBID) + mdd_changelog_rec_ext_jobid(&rec->cr, uc->uc_jobid); if (likely(target != NULL)) { rec->cr.cr_tfid = *mdo2fid(target); - target->mod_cltime = cfs_time_current_64(); + target->mod_cltime = ktime_get(); } else { fid_zero(&rec->cr.cr_tfid); } - rc = mdd_changelog_ext_store(env, mdd, rec, handle); + rc = mdd_changelog_store(env, mdd, rec, handle); if (rc < 0) { - CERROR("changelog failed: rc=%d, op%d %s c"DFID" p"DFID"\n", - rc, type, tname->ln_name, PFID(sfid), PFID(tpfid)); + CERROR("%s: cannot store changelog record: type = %d, " + "name = '%s', t = "DFID", p = "DFID": rc = %d\n", + mdd2obd_dev(mdd)->obd_name, type, tname->ln_name, + PFID(&rec->cr.cr_tfid), PFID(&rec->cr.cr_pfid), rc); return -EFAULT; } @@ -955,41 +1227,32 @@ static int mdd_linkea_prepare(const struct lu_env *env, struct linkea_data *ldata) { int rc = 0; - int rc2 = 0; ENTRY; if (OBD_FAIL_CHECK(OBD_FAIL_FID_IGIF)) - return 0; + RETURN(0); LASSERT(oldpfid != NULL || newpfid != NULL); - if (mdd_obj->mod_flags & DEAD_OBJ) { - /* Prevent linkea to be updated which is NOT necessary. */ - ldata->ld_reclen = 0; - /* No more links, don't bother */ + if (mdd_obj->mod_flags & DEAD_OBJ) + /* Unnecessary to update linkEA for dead object. */ RETURN(0); - } if (oldpfid != NULL) { rc = __mdd_links_del(env, mdd_obj, ldata, oldlname, oldpfid); if (rc) { - if ((check == 1) || - (rc != -ENODATA && rc != -ENOENT)) + if ((check == 1) || (rc != -ENODATA && rc != -ENOENT)) RETURN(rc); + /* No changes done. */ rc = 0; } } /* If renaming, add the new record */ - if (newpfid != NULL) { - /* even if the add fails, we still delete the out-of-date - * old link */ - rc2 = __mdd_links_add(env, mdd_obj, ldata, newlname, newpfid, - first, check); - } - - rc = rc != 0 ? rc : rc2; + if (newpfid != NULL) + rc = __mdd_links_add(env, mdd_obj, ldata, newlname, newpfid, + first, check); RETURN(rc); } @@ -1004,7 +1267,6 @@ int mdd_links_rename(const struct lu_env *env, struct linkea_data *ldata, int first, int check) { - int rc2 = 0; int rc = 0; ENTRY; @@ -1012,42 +1274,37 @@ int mdd_links_rename(const struct lu_env *env, ldata = &mdd_env_info(env)->mti_link_data; memset(ldata, 0, sizeof(*ldata)); rc = mdd_linkea_prepare(env, mdd_obj, oldpfid, oldlname, - newpfid, newlname, first, check, - ldata); - if (rc != 0) + newpfid, newlname, first, check, ldata); + if (rc) GOTO(out, rc); } - if (ldata->ld_reclen != 0) + if (!(mdd_obj->mod_flags & DEAD_OBJ)) rc = mdd_links_write(env, mdd_obj, ldata, handle); - EXIT; + + GOTO(out, rc); + out: - if (rc == 0) - rc = rc2; - if (rc) { - int error = 1; - if (rc == -EOVERFLOW || rc == -ENOSPC) - error = 0; - if (oldpfid == NULL) - CDEBUG(error ? D_ERROR : D_OTHER, - "link_ea add '%.*s' failed %d "DFID"\n", - newlname->ln_namelen, newlname->ln_name, + if (rc != 0) { + if (newlname == NULL) + CERROR("link_ea add failed %d "DFID"\n", rc, PFID(mdd_object_fid(mdd_obj))); + else if (oldpfid == NULL) + CERROR("link_ea add '%.*s' failed %d "DFID"\n", + newlname->ln_namelen, newlname->ln_name, rc, + PFID(mdd_object_fid(mdd_obj))); else if (newpfid == NULL) - CDEBUG(error ? D_ERROR : D_OTHER, - "link_ea del '%.*s' failed %d "DFID"\n", - oldlname->ln_namelen, oldlname->ln_name, - rc, PFID(mdd_object_fid(mdd_obj))); + CERROR("link_ea del '%.*s' failed %d "DFID"\n", + oldlname->ln_namelen, oldlname->ln_name, rc, + PFID(mdd_object_fid(mdd_obj))); else - CDEBUG(error ? D_ERROR : D_OTHER, - "link_ea rename '%.*s'->'%.*s' failed %d " - DFID"\n", - oldlname->ln_namelen, oldlname->ln_name, - newlname->ln_namelen, newlname->ln_name, - rc, PFID(mdd_object_fid(mdd_obj))); + CERROR("link_ea rename '%.*s'->'%.*s' failed %d "DFID + "\n", oldlname->ln_namelen, oldlname->ln_name, + newlname->ln_namelen, newlname->ln_name, rc, + PFID(mdd_object_fid(mdd_obj))); } - if (ldata->ld_buf && ldata->ld_buf->lb_len > OBD_ALLOC_BIG) + if (is_vmalloc_addr(ldata->ld_buf)) /* if we vmalloced a large buffer drop it */ lu_buf_free(ldata->ld_buf); @@ -1059,10 +1316,10 @@ static inline int mdd_links_add(const struct lu_env *env, const struct lu_fid *pfid, const struct lu_name *lname, struct thandle *handle, - struct linkea_data *data, int first) + struct linkea_data *ldata, int first) { return mdd_links_rename(env, mdd_obj, NULL, NULL, - pfid, lname, handle, data, first, 0); + pfid, lname, handle, ldata, first, 0); } static inline int mdd_links_del(const struct lu_env *env, @@ -1076,48 +1333,6 @@ static inline int mdd_links_del(const struct lu_env *env, } /** Read the link EA into a temp buffer. - * Uses the mdd_thread_info::mti_big_buf since it is generally large. - * A pointer to the buffer is stored in \a ldata::ld_buf. - * - * \retval 0 or error - */ -int mdd_links_read(const struct lu_env *env, struct mdd_object *mdd_obj, - struct linkea_data *ldata) -{ - int rc; - - /* First try a small buf */ - LASSERT(env != NULL); - ldata->ld_buf = lu_buf_check_and_alloc(&mdd_env_info(env)->mti_link_buf, - PAGE_CACHE_SIZE); - if (ldata->ld_buf->lb_buf == NULL) - return -ENOMEM; - - if (!mdd_object_exists(mdd_obj)) - return -ENODATA; - - rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, XATTR_NAME_LINK, - BYPASS_CAPA); - if (rc == -ERANGE) { - /* Buf was too small, figure out what we need. */ - lu_buf_free(ldata->ld_buf); - rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, - XATTR_NAME_LINK, BYPASS_CAPA); - if (rc < 0) - return rc; - ldata->ld_buf = lu_buf_check_and_alloc(ldata->ld_buf, rc); - if (ldata->ld_buf->lb_buf == NULL) - return -ENOMEM; - rc = mdo_xattr_get(env, mdd_obj, ldata->ld_buf, - XATTR_NAME_LINK, BYPASS_CAPA); - } - if (rc < 0) - return rc; - - return linkea_init(ldata); -} - -/** Read the link EA into a temp buffer. * Uses the name_buf since it is generally large. * \retval IS_ERR err * \retval ptr to \a lu_buf (always \a mti_big_buf) @@ -1125,7 +1340,7 @@ int mdd_links_read(const struct lu_env *env, struct mdd_object *mdd_obj, struct lu_buf *mdd_links_get(const struct lu_env *env, struct mdd_object *mdd_obj) { - struct linkea_data ldata = { 0 }; + struct linkea_data ldata = { NULL }; int rc; rc = mdd_links_read(env, mdd_obj, &ldata); @@ -1135,31 +1350,50 @@ struct lu_buf *mdd_links_get(const struct lu_env *env, int mdd_links_write(const struct lu_env *env, struct mdd_object *mdd_obj, struct linkea_data *ldata, struct thandle *handle) { - const struct lu_buf *buf = mdd_buf_get_const(env, ldata->ld_buf->lb_buf, - ldata->ld_leh->leh_len); - return mdo_xattr_set(env, mdd_obj, buf, XATTR_NAME_LINK, 0, handle, - mdd_object_capa(env, mdd_obj)); + const struct lu_buf *buf; + int rc; + + if (ldata == NULL || ldata->ld_buf == NULL || + ldata->ld_leh == NULL) + return 0; + + if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_LINKEA)) + return 0; + +again: + buf = mdd_buf_get_const(env, ldata->ld_buf->lb_buf, + ldata->ld_leh->leh_len); + rc = mdo_xattr_set(env, mdd_obj, buf, XATTR_NAME_LINK, 0, handle); + if (unlikely(rc == -ENOSPC)) { + rc = linkea_overflow_shrink(ldata); + if (likely(rc > 0)) + goto again; + } + + return rc; } -int mdd_declare_links_add(const struct lu_env *env, struct mdd_object *mdd_obj, - struct thandle *handle, struct linkea_data *ldata) +static int mdd_declare_links_add(const struct lu_env *env, + struct mdd_object *mdd_obj, + struct thandle *handle, + struct linkea_data *ldata) { int rc; int ea_len; void *linkea; - if (ldata != NULL && ldata->ld_lee != NULL) { + if (ldata != NULL && ldata->ld_leh != NULL) { ea_len = ldata->ld_leh->leh_len; linkea = ldata->ld_buf->lb_buf; } else { - ea_len = DEFAULT_LINKEA_SIZE; + ea_len = MAX_LINKEA_SIZE; linkea = NULL; } - /* XXX: max size? */ rc = mdo_declare_xattr_set(env, mdd_obj, mdd_buf_get_const(env, linkea, ea_len), XATTR_NAME_LINK, 0, handle); + return rc; } @@ -1178,23 +1412,28 @@ static inline int mdd_declare_links_del(const struct lu_env *env, } static int mdd_declare_link(const struct lu_env *env, - struct mdd_device *mdd, - struct mdd_object *p, - struct mdd_object *c, - const struct lu_name *name, + struct mdd_device *mdd, + struct mdd_object *p, + struct mdd_object *c, + const struct lu_name *name, struct thandle *handle, struct lu_attr *la, struct linkea_data *data) { - int rc; + struct lu_fid tfid = *mdo2fid(c); + int rc; - rc = mdo_declare_index_insert(env, p, mdo2fid(c), name->ln_name,handle); - if (rc) - return rc; + if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING3)) + tfid.f_oid = cfs_fail_val; - rc = mdo_declare_ref_add(env, c, handle); - if (rc) - return rc; + rc = mdo_declare_index_insert(env, p, &tfid, mdd_object_type(c), + name->ln_name, handle); + if (rc != 0) + return rc; + + rc = mdo_declare_ref_add(env, c, handle); + if (rc != 0) + return rc; la->la_valid = LA_CTIME | LA_MTIME; rc = mdo_declare_attr_set(env, p, la, handle); @@ -1203,16 +1442,16 @@ static int mdd_declare_link(const struct lu_env *env, la->la_valid = LA_CTIME; rc = mdo_declare_attr_set(env, c, la, handle); - if (rc) - return rc; + if (rc != 0) + return rc; rc = mdd_declare_links_add(env, c, handle, data); - if (rc) - return rc; + if (rc != 0) + return rc; - rc = mdd_declare_changelog_store(env, mdd, name, handle); + rc = mdd_declare_changelog_store(env, mdd, name, NULL, handle); - return rc; + return rc; } static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj, @@ -1225,20 +1464,30 @@ static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj, struct mdd_object *mdd_sobj = md2mdd_obj(src_obj); struct lu_attr *cattr = MDD_ENV_VAR(env, cattr); struct lu_attr *tattr = MDD_ENV_VAR(env, tattr); - struct mdd_device *mdd = mdo2mdd(src_obj); - struct thandle *handle; + struct mdd_device *mdd = mdo2mdd(src_obj); + struct thandle *handle; + struct lu_fid *tfid = &mdd_env_info(env)->mti_fid2; struct linkea_data *ldata = &mdd_env_info(env)->mti_link_data; int rc; ENTRY; - rc = mdd_la_get(env, mdd_sobj, cattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_sobj, cattr); if (rc != 0) RETURN(rc); - rc = mdd_la_get(env, mdd_tobj, tattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_tobj, tattr); if (rc != 0) RETURN(rc); + /* + * If we are using project inheritance, we only allow hard link + * creation in our tree when the project IDs are the same; + * otherwise the tree quota mechanism could be circumvented. + */ + if ((tattr->la_flags & LUSTRE_PROJINHERIT_FL) && + (tattr->la_projid != cattr->la_projid)) + RETURN(-EXDEV); + handle = mdd_trans_create(env, mdd); if (IS_ERR(handle)) GOTO(out_pending, rc = PTR_ERR(handle)); @@ -1248,6 +1497,14 @@ static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj, LASSERT(ma->ma_attr.la_valid & LA_CTIME); la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime; + /* Note: even this function will change ldata, but it comes from + * thread_info, which is completely temporary and only seen in + * this function, so we do not need reset ldata once it fails.*/ + rc = mdd_linkea_prepare(env, mdd_sobj, NULL, NULL, mdo2fid(mdd_tobj), + lname, 0, 0, ldata); + if (rc != 0) + GOTO(stop, rc); + rc = mdd_declare_link(env, mdd, mdd_tobj, mdd_sobj, lname, handle, la, ldata); if (rc) @@ -1263,14 +1520,18 @@ static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj, if (rc) GOTO(out_unlock, rc); - rc = mdo_ref_add(env, mdd_sobj, handle); - if (rc) - GOTO(out_unlock, rc); + if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_LESS_NLINK)) { + rc = mdo_ref_add(env, mdd_sobj, handle); + if (rc != 0) + GOTO(out_unlock, rc); + } + *tfid = *mdo2fid(mdd_sobj); + if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING3)) + tfid->f_oid = cfs_fail_val; - rc = __mdd_index_insert_only(env, mdd_tobj, mdo2fid(mdd_sobj), - name, handle, - mdd_object_capa(env, mdd_tobj)); + rc = __mdd_index_insert_only(env, mdd_tobj, tfid, + mdd_object_type(mdd_sobj), name, handle); if (rc != 0) { mdo_ref_del(env, mdd_sobj, handle); GOTO(out_unlock, rc); @@ -1283,82 +1544,118 @@ static int mdd_link(const struct lu_env *env, struct md_object *tgt_obj, la->la_valid = LA_CTIME; rc = mdd_update_time(env, mdd_sobj, cattr, la, handle); - if (rc == 0) { - rc = mdd_linkea_prepare(env, mdd_sobj, NULL, NULL, - mdo2fid(mdd_tobj), lname, 0, 0, - ldata); - if (rc == 0) - mdd_links_add(env, mdd_sobj, mdo2fid(mdd_tobj), - lname, handle, ldata, 0); - /* The failure of links_add should not cause the link - * failure, reset rc here */ - rc = 0; - } - EXIT; + if (rc == 0) + /* Note: The failure of links_add should not cause the + * link failure, so do not check return value. */ + mdd_links_add(env, mdd_sobj, mdo2fid(mdd_tobj), + lname, handle, ldata, 0); + + EXIT; out_unlock: - mdd_write_unlock(env, mdd_sobj); - if (rc == 0) + mdd_write_unlock(env, mdd_sobj); + if (rc == 0) rc = mdd_changelog_ns_store(env, mdd, CL_HARDLINK, 0, mdd_sobj, - mdd_tobj, lname, handle); + mdo2fid(mdd_tobj), NULL, NULL, + lname, NULL, handle); stop: - mdd_trans_stop(env, mdd, rc, handle); - - if (ldata->ld_buf && ldata->ld_buf->lb_len > OBD_ALLOC_BIG) + rc = mdd_trans_stop(env, mdd, rc, handle); + if (is_vmalloc_addr(ldata->ld_buf)) /* if we vmalloced a large buffer drop it */ lu_buf_free(ldata->ld_buf); out_pending: - return rc; + return rc; } -int mdd_declare_finish_unlink(const struct lu_env *env, - struct mdd_object *obj, - struct md_attr *ma, - struct thandle *handle) +static int mdd_mark_orphan_object(const struct lu_env *env, + struct mdd_object *obj, struct thandle *handle, + bool declare) +{ + struct lu_attr *attr = MDD_ENV_VAR(env, la_for_start); + int rc; + + if (!S_ISDIR(mdd_object_type(obj))) + return 0; + + attr->la_valid = LA_FLAGS; + attr->la_flags = LUSTRE_ORPHAN_FL; + + if (declare) + rc = mdo_declare_attr_set(env, obj, attr, handle); + else + rc = mdo_attr_set(env, obj, attr, handle); + + return rc; +} + +static int mdd_declare_finish_unlink(const struct lu_env *env, + struct mdd_object *obj, + struct thandle *handle) { int rc; + /* Sigh, we do not know if the unlink object will become orphan in + * declare phase, but fortunately the flags here does not matter + * in current declare implementation */ + rc = mdd_mark_orphan_object(env, obj, handle, true); + if (rc != 0) + return rc; + + rc = mdo_declare_destroy(env, obj, handle); + if (rc != 0) + return rc; + rc = orph_declare_index_insert(env, obj, mdd_object_type(obj), handle); - if (rc) + if (rc != 0) return rc; - return mdo_declare_destroy(env, obj, handle); + return mdd_declare_links_del(env, obj, handle); } /* 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) + struct mdd_object *obj, struct md_attr *ma, + const struct mdd_object *pobj, + const struct lu_name *lname, + struct thandle *th) { int rc = 0; - int is_dir = S_ISDIR(ma->ma_attr.la_mode); - ENTRY; + int is_dir = S_ISDIR(ma->ma_attr.la_mode); + ENTRY; - LASSERT(mdd_write_locked(env, obj) != 0); + LASSERT(mdd_write_locked(env, obj) != 0); if (ma->ma_attr.la_nlink == 0 || is_dir) { - obj->mod_flags |= DEAD_OBJ; - - /* add new orphan and the object - * will be deleted during mdd_close() */ - if (obj->mod_count) { - rc = __mdd_orphan_add(env, obj, th); - if (rc == 0) - CDEBUG(D_HA, "Object "DFID" is inserted into " - "orphan list, open count = %d\n", - PFID(mdd_object_fid(obj)), - obj->mod_count); - else - CERROR("Object "DFID" fail to be an orphan, " - "open count = %d, maybe cause failed " - "open replay\n", - PFID(mdd_object_fid(obj)), - obj->mod_count); - } else { + /* add new orphan and the object + * will be deleted during mdd_close() */ + obj->mod_flags |= DEAD_OBJ; + if (obj->mod_count) { + rc = __mdd_orphan_add(env, obj, th); + if (rc == 0) + CDEBUG(D_HA, "Object "DFID" is inserted into " + "orphan list, open count = %d\n", + PFID(mdd_object_fid(obj)), + obj->mod_count); + else + CERROR("Object "DFID" fail to be an orphan, " + "open count = %d, maybe cause failed " + "open replay\n", + PFID(mdd_object_fid(obj)), + obj->mod_count); + + /* mark object as an orphan here, not + * before __mdd_orphan_add() as racing + * mdd_la_get() may propagate ORPHAN_OBJ + * causing the asserition */ + rc = mdd_mark_orphan_object(env, obj, th, false); + } else { rc = mdo_destroy(env, obj, th); - } - } + } + } else if (!is_dir) { + /* old files may not have link ea; ignore errors */ + mdd_links_del(env, obj, mdo2fid(pobj), lname, th); + } - RETURN(rc); + RETURN(rc); } /* @@ -1381,20 +1678,25 @@ int mdd_unlink_sanity_check(const struct lu_env *env, struct mdd_object *pobj, static int mdd_declare_unlink(const struct lu_env *env, struct mdd_device *mdd, struct mdd_object *p, struct mdd_object *c, const struct lu_name *name, struct md_attr *ma, - struct thandle *handle, int no_name) + struct thandle *handle, int no_name, int is_dir) { - struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; - int rc; + struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; + int rc; - if (likely(no_name == 0)) { - rc = mdo_declare_index_delete(env, p, name->ln_name, handle); - if (rc) - return rc; - } + if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING2)) { + if (likely(no_name == 0)) { + rc = mdo_declare_index_delete(env, p, name->ln_name, + handle); + if (rc != 0) + return rc; + } - rc = mdo_declare_ref_del(env, p, handle); - if (rc) - return rc; + if (is_dir != 0) { + rc = mdo_declare_ref_del(env, p, handle); + if (rc != 0) + return rc; + } + } LASSERT(ma->ma_attr.la_valid & LA_CTIME); la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime; @@ -1417,16 +1719,12 @@ static int mdd_declare_unlink(const struct lu_env *env, struct mdd_device *mdd, if (rc) return rc; - rc = mdd_declare_finish_unlink(env, c, ma, handle); + rc = mdd_declare_finish_unlink(env, c, handle); if (rc) return rc; - rc = mdd_declare_links_del(env, c, handle); - if (rc != 0) - return rc; - /* FIXME: need changelog for remove entry */ - rc = mdd_declare_changelog_store(env, mdd, name, handle); + rc = mdd_declare_changelog_store(env, mdd, name, NULL, handle); } return rc; @@ -1451,14 +1749,13 @@ static bool mdd_hsm_archive_exists(const struct lu_env *env, hsm_buf = mdd_buf_get(env, NULL, 0); lu_buf_alloc(hsm_buf, buflen); - rc = mdo_xattr_get(env, obj, hsm_buf, XATTR_NAME_HSM, - mdd_object_capa(env, obj)); + rc = mdo_xattr_get(env, obj, hsm_buf, XATTR_NAME_HSM); rc = lustre_buf2hsm(hsm_buf->lb_buf, rc, &ma->ma_hsm); lu_buf_free(hsm_buf); if (rc < 0) RETURN(false); - ma->ma_valid = MA_HSM; + ma->ma_valid |= MA_HSM; } if (ma->ma_hsm.mh_flags & HS_EXISTS) RETURN(true); @@ -1486,7 +1783,7 @@ static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, struct mdd_object *mdd_cobj = NULL; struct mdd_device *mdd = mdo2mdd(pobj); struct thandle *handle; - int rc, is_dir = 0; + int rc, is_dir = 0, cl_flags = 0; ENTRY; /* cobj == NULL means only delete name entry */ @@ -1496,17 +1793,22 @@ static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, RETURN(-ENOENT); } - rc = mdd_la_get(env, mdd_pobj, pattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_pobj, pattr); if (rc) RETURN(rc); if (likely(mdd_cobj != NULL)) { /* fetch cattr */ - rc = mdd_la_get(env, mdd_cobj, cattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_cobj, cattr); if (rc) RETURN(rc); is_dir = S_ISDIR(cattr->la_mode); + /* search for an existing archive. + * we should check ahead as the object + * can be destroyed in this transaction */ + if (mdd_hsm_archive_exists(env, mdd_cobj, ma)) + cl_flags |= CLF_UNLINK_HSM_EXISTS; } rc = mdd_unlink_sanity_check(env, mdd_pobj, pattr, mdd_cobj, cattr); @@ -1518,7 +1820,7 @@ static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, RETURN(PTR_ERR(handle)); rc = mdd_declare_unlink(env, mdd, mdd_pobj, mdd_cobj, - lname, ma, handle, no_name); + lname, ma, handle, no_name, is_dir); if (rc) GOTO(stop, rc); @@ -1529,20 +1831,23 @@ static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, if (likely(mdd_cobj != NULL)) mdd_write_lock(env, mdd_cobj, MOR_TGT_CHILD); - if (likely(no_name == 0)) { - rc = __mdd_index_delete(env, mdd_pobj, name, is_dir, handle, - mdd_object_capa(env, mdd_pobj)); + if (likely(no_name == 0) && !OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DANGLING2)) { + rc = __mdd_index_delete(env, mdd_pobj, name, is_dir, handle); if (rc) GOTO(cleanup, rc); } + if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_MUL_REF) || + OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_NAMEENTRY)) + GOTO(cleanup, rc = 0); + if (likely(mdd_cobj != NULL)) { rc = mdo_ref_del(env, mdd_cobj, handle); if (rc != 0) { __mdd_index_insert_only(env, mdd_pobj, mdo2fid(mdd_cobj), - name, handle, - mdd_object_capa(env, mdd_pobj)); + mdd_object_type(mdd_cobj), + name, handle); GOTO(cleanup, rc); } @@ -1551,7 +1856,7 @@ static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, mdo_ref_del(env, mdd_cobj, handle); /* fetch updated nlink */ - rc = mdd_la_get(env, mdd_cobj, cattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_cobj, cattr); if (rc) GOTO(cleanup, rc); } @@ -1580,41 +1885,43 @@ static int mdd_unlink(const struct lu_env *env, struct md_object *pobj, /* XXX: this transfer to ma will be removed with LOD/OSP */ ma->ma_attr = *cattr; ma->ma_valid |= MA_INODE; - rc = mdd_finish_unlink(env, mdd_cobj, ma, handle); + rc = mdd_finish_unlink(env, mdd_cobj, ma, mdd_pobj, lname, handle); + if (rc != 0) + GOTO(cleanup, rc); /* fetch updated nlink */ - if (rc == 0) - rc = mdd_la_get(env, mdd_cobj, cattr, BYPASS_CAPA); - - if (!is_dir) - /* old files may not have link ea; ignore errors */ - mdd_links_del(env, mdd_cobj, mdo2fid(mdd_pobj), lname, handle); + rc = mdd_la_get(env, mdd_cobj, cattr); + /* if object is removed then we can't get its attrs, + * use last get */ + if (rc == -ENOENT) { + cattr->la_nlink = 0; + rc = 0; + } - /* if object is removed then we can't get its attrs, use last get */ if (cattr->la_nlink == 0) { ma->ma_attr = *cattr; ma->ma_valid |= MA_INODE; } + EXIT; cleanup: - mdd_write_unlock(env, mdd_cobj); - if (rc == 0) { - int cl_flags = 0; + if (likely(mdd_cobj != NULL)) + mdd_write_unlock(env, mdd_cobj); - if (cattr->la_nlink == 0) { + if (rc == 0) { + if (cattr->la_nlink == 0) cl_flags |= CLF_UNLINK_LAST; - /* search for an existing archive */ - if (mdd_hsm_archive_exists(env, mdd_cobj, ma)) - cl_flags |= CLF_UNLINK_HSM_EXISTS; - } + else + cl_flags &= ~CLF_UNLINK_HSM_EXISTS; rc = mdd_changelog_ns_store(env, mdd, is_dir ? CL_RMDIR : CL_UNLINK, cl_flags, - mdd_cobj, mdd_pobj, lname, handle); + mdd_cobj, mdo2fid(mdd_pobj), NULL, NULL, lname, NULL, + handle); } stop: - mdd_trans_stop(env, mdd, rc, handle); + rc = mdd_trans_stop(env, mdd, rc, handle); return rc; } @@ -1634,9 +1941,11 @@ static int mdd_cd_sanity_check(const struct lu_env *env, RETURN(0); } -static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, - struct md_object *cobj, const struct md_op_spec *spec, - struct md_attr *ma) +static int mdd_create_data(const struct lu_env *env, + struct md_object *pobj, + struct md_object *cobj, + const struct md_op_spec *spec, + struct md_attr *ma) { struct mdd_device *mdd = mdo2mdd(cobj); struct mdd_object *mdd_pobj = md2mdd_obj(pobj); @@ -1644,6 +1953,7 @@ static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, struct thandle *handle; const struct lu_buf *buf; struct lu_attr *attr = MDD_ENV_VAR(env, cattr); + struct dt_allocation_hint *hint = &mdd_env_info(env)->mti_hint; int rc; ENTRY; @@ -1660,22 +1970,22 @@ static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, * striping can be specified or not * 2) CMD? */ - rc = mdd_la_get(env, son, attr, BYPASS_CAPA); + rc = mdd_la_get(env, son, attr); if (rc) RETURN(rc); /* calling ->ah_make_hint() is used to transfer information from parent */ - mdd_object_make_hint(env, mdd_pobj, son, attr, spec); + mdd_object_make_hint(env, mdd_pobj, son, attr, spec, hint); - handle = mdd_trans_create(env, mdd); - if (IS_ERR(handle)) - GOTO(out_free, rc = PTR_ERR(handle)); + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + GOTO(out_free, rc = PTR_ERR(handle)); - /* - * XXX: Setting the lov ea is not locked but setting the attr is locked? - * Should this be fixed? - */ - CDEBUG(D_OTHER, "ea %p/%u, cr_flags "LPO64", no_create %u\n", + /* + * XXX: Setting the lov ea is not locked but setting the attr is locked? + * Should this be fixed? + */ + CDEBUG(D_OTHER, "ea %p/%u, cr_flags %#llo, no_create %u\n", spec->u.sp_ea.eadata, spec->u.sp_ea.eadatalen, spec->sp_cr_flags, spec->no_create); @@ -1691,7 +2001,7 @@ static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, if (rc) GOTO(stop, rc); - rc = mdd_declare_changelog_store(env, mdd, NULL, handle); + rc = mdd_declare_changelog_store(env, mdd, NULL, NULL, handle); if (rc) GOTO(stop, rc); @@ -1700,7 +2010,7 @@ static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, GOTO(stop, rc); rc = dt_xattr_set(env, mdd_object_child(son), buf, XATTR_NAME_LOV, - 0, handle, mdd_object_capa(env, son)); + 0, handle); if (rc) GOTO(stop, rc); @@ -1708,7 +2018,8 @@ static int mdd_create_data(const struct lu_env *env, struct md_object *pobj, rc = mdd_changelog_data_store(env, mdd, CL_LAYOUT, 0, son, handle); stop: - mdd_trans_stop(env, mdd, rc, handle); + rc = mdd_trans_stop(env, mdd, rc, handle); + out_free: RETURN(rc); } @@ -1716,80 +2027,72 @@ out_free: static int mdd_declare_object_initialize(const struct lu_env *env, struct mdd_object *parent, struct mdd_object *child, - struct lu_attr *attr, - struct thandle *handle, - struct linkea_data *ldata) + const struct lu_attr *attr, + struct thandle *handle) { - int rc; + int rc; ENTRY; - /* - * inode mode has been set in creation time, and it's based on umask, - * la_mode and acl, don't set here again! (which will go wrong - * because below function doesn't consider umask). - * I'd suggest set all object attributes in creation time, see above. - */ LASSERT(attr->la_valid & (LA_MODE | LA_TYPE)); - attr->la_valid &= ~(LA_MODE | LA_TYPE); - rc = mdo_declare_attr_set(env, child, attr, handle); - attr->la_valid |= LA_MODE | LA_TYPE; - if (rc == 0 && S_ISDIR(attr->la_mode)) { - rc = mdo_declare_index_insert(env, child, mdo2fid(child), - dot, handle); - if (rc == 0) - rc = mdo_declare_ref_add(env, child, handle); - - rc = mdo_declare_index_insert(env, child, mdo2fid(parent), - dotdot, handle); - } + if (!S_ISDIR(attr->la_mode)) + RETURN(0); - if (rc == 0) - mdd_declare_links_add(env, child, handle, ldata); + rc = mdo_declare_index_insert(env, child, mdo2fid(child), S_IFDIR, + dot, handle); + if (rc != 0) + RETURN(rc); + + rc = mdo_declare_ref_add(env, child, handle); + if (rc != 0) + RETURN(rc); + + rc = mdo_declare_index_insert(env, child, mdo2fid(parent), S_IFDIR, + dotdot, handle); RETURN(rc); } static int mdd_object_initialize(const struct lu_env *env, const struct lu_fid *pfid, - const struct lu_name *lname, struct mdd_object *child, struct lu_attr *attr, struct thandle *handle, - const struct md_op_spec *spec, - struct linkea_data *ldata) + const struct md_op_spec *spec) { - 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. - */ + int rc = 0; + ENTRY; - rc = mdd_attr_set_internal(env, child, attr, handle, 0); - /* arguments are supposed to stay the same */ if (S_ISDIR(attr->la_mode)) { /* Add "." and ".." for newly created dir */ mdo_ref_add(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) - mdo_ref_del(env, child, handle); - } - - if (rc == 0) - mdd_links_add(env, child, pfid, lname, handle, ldata, 1); + S_IFDIR, dot, handle); + if (rc == 0) + rc = __mdd_index_insert_only(env, child, pfid, S_IFDIR, + dotdot, handle); + if (rc != 0) + mdo_ref_del(env, child, handle); + } RETURN(rc); } -/* has not lock on pobj yet */ +/** + * This function checks whether it can create a file/dir under the + * directory(@pobj). The directory(@pobj) is not being locked by + * mdd lock. + * + * \param[in] env execution environment + * \param[in] pobj the directory to create files + * \param[in] pattr the attributes of the directory + * \param[in] lname the name of the created file/dir + * \param[in] cattr the attributes of the file/dir + * \param[in] spec create specification + * + * \retval = 0 it is allowed to create file/dir under + * the directory + * \retval negative error not allowed to create file/dir + * under the directory + */ static int mdd_create_sanity_check(const struct lu_env *env, struct md_object *pobj, const struct lu_attr *pattr, @@ -1797,16 +2100,17 @@ static int mdd_create_sanity_check(const struct lu_env *env, struct lu_attr *cattr, struct md_op_spec *spec) { - struct mdd_thread_info *info = mdd_env_info(env); - struct lu_fid *fid = &info->mti_fid; - struct mdd_object *obj = md2mdd_obj(pobj); - struct mdd_device *m = mdo2mdd(pobj); - int rc; - ENTRY; + struct mdd_thread_info *info = mdd_env_info(env); + struct lu_fid *fid = &info->mti_fid; + struct mdd_object *obj = md2mdd_obj(pobj); + struct mdd_device *m = mdo2mdd(pobj); + bool check_perm = true; + int rc; + ENTRY; - /* EEXIST check */ - if (mdd_is_dead_obj(obj)) - RETURN(-ENOENT); + /* EEXIST check */ + if (mdd_is_dead_obj(obj)) + RETURN(-ENOENT); /* * In some cases this lookup is not needed - we know before if name @@ -1821,20 +2125,35 @@ static int mdd_create_sanity_check(const struct lu_env *env, */ rc = __mdd_lookup(env, pobj, pattr, lname, fid, MAY_WRITE | MAY_EXEC); - if (rc != -ENOENT) - RETURN(rc ? : -EEXIST); - } else { - /* - * Check WRITE permission for the parent. - * EXEC permission have been checked - * when lookup before create already. - */ - rc = mdd_permission_internal_locked(env, obj, pattr, MAY_WRITE, - MOR_TGT_PARENT); - if (rc) - RETURN(rc); + if (rc != -ENOENT) + RETURN(rc ? : -EEXIST); + + /* Permission is already being checked in mdd_lookup */ + check_perm = false; + } + + if (S_ISDIR(cattr->la_mode) && + unlikely(spec != NULL && spec->sp_cr_flags & MDS_OPEN_HAS_EA) && + spec->u.sp_ea.eadata != NULL && spec->u.sp_ea.eadatalen > 0) { + const struct lmv_user_md *lum = spec->u.sp_ea.eadata; + + if (unlikely(le32_to_cpu(lum->lum_magic) != LMV_USER_MAGIC) && + le32_to_cpu(lum->lum_magic) != LMV_USER_MAGIC_V0) { + rc = -EINVAL; + CERROR("%s: invalid lmv_user_md: magic = %x, " + "stripe_offset = %d, stripe_count = %u: " + "rc = %d\n", mdd2obd_dev(m)->obd_name, + le32_to_cpu(lum->lum_magic), + (int)le32_to_cpu(lum->lum_stripe_offset), + le32_to_cpu(lum->lum_stripe_count), rc); + return rc; + } } + rc = mdd_may_create(env, obj, pattr, NULL, check_perm); + if (rc != 0) + RETURN(rc); + /* sgid check */ if (pattr->la_mode & S_ISGID) { cattr->la_gid = pattr->la_gid; @@ -1844,15 +2163,29 @@ static int mdd_create_sanity_check(const struct lu_env *env, } } + /* Inherit project ID from parent directory */ + if (pattr->la_flags & LUSTRE_PROJINHERIT_FL) { + cattr->la_projid = pattr->la_projid; + if (S_ISDIR(cattr->la_mode)) { + cattr->la_flags |= LUSTRE_PROJINHERIT_FL; + cattr->la_valid |= LA_FLAGS; + } + cattr->la_valid |= LA_PROJID; + } + + rc = mdd_name_check(m, lname); + if (rc < 0) + RETURN(rc); + switch (cattr->la_mode & S_IFMT) { - case S_IFLNK: { - unsigned int symlen = strlen(spec->u.sp_symname) + 1; + case S_IFLNK: { + unsigned int symlen = strlen(spec->u.sp_symname) + 1; - if (symlen > (1 << m->mdd_dt_conf.ddp_block_shift)) - RETURN(-ENAMETOOLONG); - else - RETURN(0); - } + if (symlen > m->mdd_dt_conf.ddp_symlink_max) + RETURN(-ENAMETOOLONG); + else + RETURN(0); + } case S_IFDIR: case S_IFREG: case S_IFCHR: @@ -1868,21 +2201,23 @@ static int mdd_create_sanity_check(const struct lu_env *env, RETURN(rc); } -static int mdd_declare_create(const struct lu_env *env, struct mdd_device *mdd, - struct mdd_object *p, struct mdd_object *c, - const struct lu_name *name, - struct lu_attr *attr, - struct thandle *handle, - const struct md_op_spec *spec, - struct linkea_data *ldata, - struct lu_buf *def_acl_buf, - struct lu_buf *acl_buf) +static int mdd_declare_create_object(const struct lu_env *env, + struct mdd_device *mdd, + struct mdd_object *p, struct mdd_object *c, + struct lu_attr *attr, + struct thandle *handle, + const struct md_op_spec *spec, + struct lu_buf *def_acl_buf, + struct lu_buf *acl_buf, + struct dt_allocation_hint *hint) { + const struct lu_buf *buf; int rc; - rc = mdd_declare_object_create_internal(env, p, c, attr, handle, spec); - if (rc) - GOTO(out, rc); + rc = mdd_declare_create_object_internal(env, p, c, attr, handle, spec, + hint); + if (rc) + GOTO(out, rc); #ifdef CONFIG_FS_POSIX_ACL if (def_acl_buf->lb_len > 0 && S_ISDIR(attr->la_mode)) { @@ -1905,30 +2240,13 @@ static int mdd_declare_create(const struct lu_env *env, struct mdd_device *mdd, GOTO(out, rc); } #endif - - if (S_ISDIR(attr->la_mode)) { - rc = mdo_declare_ref_add(env, p, handle); - if (rc) - GOTO(out, rc); - } - - rc = mdd_declare_object_initialize(env, p, c, attr, handle, ldata); - if (rc) - GOTO(out, rc); - - if (spec->sp_cr_flags & MDS_OPEN_VOLATILE) - rc = orph_declare_index_insert(env, c, attr->la_mode, handle); - else - rc = mdo_declare_index_insert(env, p, mdo2fid(c), - name->ln_name, handle); + rc = mdd_declare_object_initialize(env, p, c, attr, handle); if (rc) GOTO(out, rc); /* replay case, create LOV EA from client data */ if (spec->no_create || (spec->sp_cr_flags & MDS_OPEN_HAS_EA && S_ISREG(attr->la_mode))) { - const struct lu_buf *buf; - buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata, spec->u.sp_ea.eadatalen); rc = mdo_declare_xattr_set(env, c, buf, XATTR_NAME_LOV, 0, @@ -1938,33 +2256,80 @@ static int mdd_declare_create(const struct lu_env *env, struct mdd_device *mdd, } if (S_ISLNK(attr->la_mode)) { + const char *target_name = spec->u.sp_symname; + int sym_len = strlen(target_name); + const struct lu_buf *buf; + + buf = mdd_buf_get_const(env, target_name, sym_len); rc = dt_declare_record_write(env, mdd_object_child(c), - strlen(spec->u.sp_symname), 0, - handle); + buf, 0, handle); if (rc) GOTO(out, rc); } - if (!(spec->sp_cr_flags & MDS_OPEN_VOLATILE)) { - struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; + if (spec->sp_cr_file_secctx_name != NULL) { + buf = mdd_buf_get_const(env, spec->sp_cr_file_secctx, + spec->sp_cr_file_secctx_size); + rc = mdo_declare_xattr_set(env, c, buf, + spec->sp_cr_file_secctx_name, 0, + handle); + if (rc < 0) + GOTO(out, rc); + } +out: + return rc; +} + +static int mdd_declare_create(const struct lu_env *env, struct mdd_device *mdd, + struct mdd_object *p, struct mdd_object *c, + const struct lu_name *name, + struct lu_attr *attr, + struct thandle *handle, + const struct md_op_spec *spec, + struct linkea_data *ldata, + struct lu_buf *def_acl_buf, + struct lu_buf *acl_buf, + struct dt_allocation_hint *hint) +{ + int rc; + + rc = mdd_declare_create_object(env, mdd, p, c, attr, handle, spec, + def_acl_buf, acl_buf, hint); + if (rc) + GOTO(out, rc); + + if (S_ISDIR(attr->la_mode)) { + rc = mdo_declare_ref_add(env, p, handle); + if (rc) + GOTO(out, rc); + } + + if (unlikely(spec->sp_cr_flags & MDS_OPEN_VOLATILE)) { + rc = orph_declare_index_insert(env, c, attr->la_mode, handle); + if (rc) + GOTO(out, rc); + } else { + struct lu_attr *la = &mdd_env_info(env)->mti_la_for_fix; + + rc = mdo_declare_index_insert(env, p, mdo2fid(c), attr->la_mode, + name->ln_name, handle); + if (rc != 0) + return rc; + + rc = mdd_declare_links_add(env, c, handle, ldata); + if (rc) + return rc; *la = *attr; la->la_valid = LA_CTIME | LA_MTIME; rc = mdo_declare_attr_set(env, p, la, handle); if (rc) return rc; - } - - rc = mdd_declare_changelog_store(env, mdd, name, handle); - if (rc) - return rc; - /* XXX: For remote create, it should indicate the remote RPC - * will be sent after local transaction is finished, which - * is not very nice, but it will be removed once we fully support - * async update */ - if (mdd_object_remote(p) && handle->th_update != NULL) - handle->th_update->tu_sent_after_local_trans = 1; + rc = mdd_declare_changelog_store(env, mdd, name, NULL, handle); + if (rc) + return rc; + } out: return rc; } @@ -1984,7 +2349,7 @@ static int mdd_acl_init(const struct lu_env *env, struct mdd_object *pobj, mdd_read_lock(env, pobj, MOR_TGT_PARENT); rc = mdo_xattr_get(env, pobj, def_acl_buf, - XATTR_NAME_ACL_DEFAULT, BYPASS_CAPA); + XATTR_NAME_ACL_DEFAULT); mdd_read_unlock(env, pobj); if (rc > 0) { /* If there are default ACL, fix mode/ACL by default ACL */ @@ -2011,12 +2376,217 @@ static int mdd_acl_init(const struct lu_env *env, struct mdd_object *pobj, RETURN(rc); } -/* +/** + * Create a metadata object and initialize it, set acl, xattr. + **/ +static int mdd_create_object(const struct lu_env *env, struct mdd_object *pobj, + struct mdd_object *son, struct lu_attr *attr, + struct md_op_spec *spec, struct lu_buf *acl_buf, + struct lu_buf *def_acl_buf, + struct dt_allocation_hint *hint, + struct thandle *handle) +{ + const struct lu_buf *buf; + int rc; + + mdd_write_lock(env, son, MOR_TGT_CHILD); + rc = mdd_create_object_internal(env, NULL, son, attr, handle, spec, + hint); + if (rc) + GOTO(unlock, rc); + + /* Note: In DNE phase I, for striped dir, though sub-stripes will be + * created in declare phase, they also needs to be added to master + * object as sub-directory entry. So it has to initialize the master + * object, then set dir striped EA.(in mdo_xattr_set) */ + rc = mdd_object_initialize(env, mdo2fid(pobj), son, attr, handle, + spec); + if (rc != 0) + GOTO(err_destroy, rc); + + /* + * in case of replay we just set LOVEA provided by the client + * XXX: I think it would be interesting to try "old" way where + * MDT calls this xattr_set(LOV) in a different transaction. + * probably this way we code can be made better. + */ + + /* During creation, there are only a few cases we need do xattr_set to + * create stripes. + * 1. regular file: see comments above. + * 2. dir: inherit default striping or pool settings from parent. + * 3. create striped directory with provided stripeEA. + * 4. create striped directory because inherit default layout from the + * parent. + */ + if (spec->no_create || + (S_ISREG(attr->la_mode) && spec->sp_cr_flags & MDS_OPEN_HAS_EA) || + S_ISDIR(attr->la_mode)) { + buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata, + spec->u.sp_ea.eadatalen); + rc = mdo_xattr_set(env, son, buf, + S_ISDIR(attr->la_mode) ? XATTR_NAME_LMV : + XATTR_NAME_LOV, 0, + handle); + if (rc != 0) + GOTO(err_destroy, rc); + } + +#ifdef CONFIG_FS_POSIX_ACL + if (def_acl_buf != NULL && def_acl_buf->lb_len > 0 && + S_ISDIR(attr->la_mode)) { + /* set default acl */ + rc = mdo_xattr_set(env, son, def_acl_buf, + XATTR_NAME_ACL_DEFAULT, 0, + handle); + if (rc) + GOTO(err_destroy, rc); + } + /* set its own acl */ + if (acl_buf != NULL && acl_buf->lb_len > 0) { + rc = mdo_xattr_set(env, son, acl_buf, + XATTR_NAME_ACL_ACCESS, + 0, handle); + if (rc) + GOTO(err_destroy, rc); + } +#endif + + if (S_ISLNK(attr->la_mode)) { + struct lu_ucred *uc = lu_ucred_assert(env); + struct dt_object *dt = mdd_object_child(son); + const char *target_name = spec->u.sp_symname; + int sym_len = strlen(target_name); + 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, + uc->uc_cap & + CFS_CAP_SYS_RESOURCE_MASK); + + if (rc == sym_len) + rc = 0; + else + GOTO(err_initlized, rc = -EFAULT); + } + + if (spec->sp_cr_file_secctx_name != NULL) { + buf = mdd_buf_get_const(env, spec->sp_cr_file_secctx, + spec->sp_cr_file_secctx_size); + rc = mdo_xattr_set(env, son, buf, spec->sp_cr_file_secctx_name, + 0, handle); + if (rc < 0) + GOTO(err_initlized, rc); + } + +err_initlized: + if (unlikely(rc != 0)) { + int rc2; + if (S_ISDIR(attr->la_mode)) { + /* Drop the reference, no need to delete "."/"..", + * because the object to be destroied directly. */ + rc2 = mdo_ref_del(env, son, handle); + if (rc2 != 0) + GOTO(unlock, rc); + } + rc2 = mdo_ref_del(env, son, handle); + if (rc2 != 0) + GOTO(unlock, rc); +err_destroy: + mdo_destroy(env, son, handle); + } +unlock: + mdd_write_unlock(env, son); + RETURN(rc); +} + +static int mdd_index_delete(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct lu_attr *cattr, + const struct lu_name *lname) +{ + struct mdd_device *mdd = mdo2mdd(&mdd_pobj->mod_obj); + struct thandle *handle; + int rc; + ENTRY; + + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + RETURN(PTR_ERR(handle)); + + rc = mdo_declare_index_delete(env, mdd_pobj, lname->ln_name, + handle); + if (rc != 0) + GOTO(stop, rc); + + if (S_ISDIR(cattr->la_mode)) { + rc = mdo_declare_ref_del(env, mdd_pobj, handle); + if (rc != 0) + GOTO(stop, rc); + } + + /* Since this will only be used in the error handler path, + * Let's set the thandle to be local and not mess the transno */ + handle->th_local = 1; + rc = mdd_trans_start(env, mdd, handle); + if (rc) + GOTO(stop, rc); + + rc = __mdd_index_delete(env, mdd_pobj, lname->ln_name, + S_ISDIR(cattr->la_mode), handle); + if (rc) + GOTO(stop, rc); +stop: + rc = mdd_trans_stop(env, mdd, rc, handle); + + RETURN(rc); +} + +/** * Create object and insert it into namespace. + * + * Two operations have to be performed: + * + * - an allocation of a new object (->do_create()), and + * - an insertion into a 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 not good but not 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_create_object_internal()) + * 2. insert (__mdd_index_insert(), lookup again) + * + * \param[in] pobj parent object + * \param[in] lname name of child being created + * \param[in,out] child child object being created + * \param[in] spec additional create parameters + * \param[in] ma attributes for new child object + * + * \retval 0 on success + * \retval negative errno on failure */ static int mdd_create(const struct lu_env *env, struct md_object *pobj, const struct lu_name *lname, struct md_object *child, - struct md_op_spec *spec, struct md_attr* ma) + struct md_op_spec *spec, struct md_attr *ma) { struct mdd_thread_info *info = mdd_env_info(env); struct lu_attr *la = &info->mti_la_for_fix; @@ -2030,46 +2600,12 @@ static int mdd_create(const struct lu_env *env, struct md_object *pobj, struct lu_buf def_acl_buf; struct linkea_data *ldata = &info->mti_link_data; const char *name = lname->ln_name; - int rc, created = 0, initialized = 0, inserted = 0; + struct dt_allocation_hint *hint = &mdd_env_info(env)->mti_hint; + int rc; + int rc2; ENTRY; - /* - * Two operations have to be performed: - * - * - an allocation of a new object (->do_create()), and - * - * - an insertion into a 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) - */ - - rc = mdd_la_get(env, mdd_pobj, pattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_pobj, pattr); if (rc != 0) RETURN(rc); @@ -2078,198 +2614,144 @@ static int mdd_create(const struct lu_env *env, struct md_object *pobj, if (rc) RETURN(rc); - if (OBD_FAIL_CHECK(OBD_FAIL_MDS_DQACQ_NET)) + if (OBD_FAIL_CHECK(OBD_FAIL_MDS_DQACQ_NET)) GOTO(out_free, rc = -EINPROGRESS); - acl_buf.lb_buf = info->mti_xattr_buf; - acl_buf.lb_len = sizeof(info->mti_xattr_buf); + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + GOTO(out_free, rc = PTR_ERR(handle)); + + lu_buf_check_and_alloc(&info->mti_xattr_buf, + mdd->mdd_dt_conf.ddp_max_ea_size); + acl_buf = info->mti_xattr_buf; def_acl_buf.lb_buf = info->mti_key; def_acl_buf.lb_len = sizeof(info->mti_key); rc = mdd_acl_init(env, mdd_pobj, attr, &def_acl_buf, &acl_buf); if (rc < 0) - GOTO(out_free, rc); + GOTO(out_stop, rc); - mdd_object_make_hint(env, mdd_pobj, son, attr, spec); - - handle = mdd_trans_create(env, mdd); - if (IS_ERR(handle)) - GOTO(out_free, rc = PTR_ERR(handle)); + mdd_object_make_hint(env, mdd_pobj, son, attr, spec, hint); memset(ldata, 0, sizeof(*ldata)); - mdd_linkea_prepare(env, son, NULL, NULL, mdd_object_fid(mdd_pobj), - lname, 1, 0, ldata); + if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_PARENT)) { + struct lu_fid tfid = *mdd_object_fid(mdd_pobj); - rc = mdd_declare_create(env, mdd, mdd_pobj, son, lname, attr, - handle, spec, ldata, &def_acl_buf, &acl_buf); - if (rc) - GOTO(out_stop, rc); + tfid.f_oid--; + rc = mdd_linkea_prepare(env, son, NULL, NULL, + &tfid, lname, 1, 0, ldata); + } else { + rc = mdd_linkea_prepare(env, son, NULL, NULL, + mdd_object_fid(mdd_pobj), + lname, 1, 0, ldata); + } - rc = mdd_trans_start(env, mdd, handle); - if (rc) - GOTO(out_stop, rc); + rc = mdd_declare_create(env, mdd, mdd_pobj, son, lname, attr, + handle, spec, ldata, &def_acl_buf, &acl_buf, + hint); + if (rc) + GOTO(out_stop, rc); - mdd_write_lock(env, son, MOR_TGT_CHILD); - rc = mdd_object_create_internal(env, NULL, son, attr, handle, spec); - if (rc) { - mdd_write_unlock(env, son); - GOTO(cleanup, rc); - } + rc = mdd_trans_start(env, mdd, handle); + if (rc) + GOTO(out_stop, rc); - created = 1; + rc = mdd_create_object(env, mdd_pobj, son, attr, spec, &acl_buf, + &def_acl_buf, hint, handle); + if (rc != 0) + GOTO(out_stop, rc); -#ifdef CONFIG_FS_POSIX_ACL - if (def_acl_buf.lb_len > 0 && S_ISDIR(attr->la_mode)) { - /* set default acl */ - rc = mdo_xattr_set(env, son, &def_acl_buf, - XATTR_NAME_ACL_DEFAULT, 0, - handle, BYPASS_CAPA); - if (rc) { - mdd_write_unlock(env, son); - GOTO(cleanup, rc); - } - } - /* set its own acl */ - if (acl_buf.lb_len > 0) { - rc = mdo_xattr_set(env, son, &acl_buf, - XATTR_NAME_ACL_ACCESS, - 0, handle, BYPASS_CAPA); - if (rc) { - mdd_write_unlock(env, son); - GOTO(cleanup, rc); - } - } -#endif + if (unlikely(spec->sp_cr_flags & MDS_OPEN_VOLATILE)) { + mdd_write_lock(env, son, MOR_TGT_CHILD); + son->mod_flags |= VOLATILE_OBJ; + rc = __mdd_orphan_add(env, son, handle); + GOTO(out_volatile, rc); + } else { + rc = __mdd_index_insert(env, mdd_pobj, mdo2fid(son), + attr->la_mode, name, handle); + if (rc != 0) + GOTO(err_created, rc); - rc = mdd_object_initialize(env, mdo2fid(mdd_pobj), lname, - son, attr, handle, spec, ldata); + mdd_links_add(env, son, mdo2fid(mdd_pobj), lname, handle, + ldata, 1); - /* - * in case of replay we just set LOVEA provided by the client - * XXX: I think it would be interesting to try "old" way where - * MDT calls this xattr_set(LOV) in a different transaction. - * probably this way we code can be made better. - */ - if (rc == 0 && (spec->no_create || - (spec->sp_cr_flags & MDS_OPEN_HAS_EA && - S_ISREG(attr->la_mode)))) { - const struct lu_buf *buf; - - buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata, - spec->u.sp_ea.eadatalen); - rc = mdo_xattr_set(env, son, buf, XATTR_NAME_LOV, 0, handle, - BYPASS_CAPA); + /* update parent directory mtime/ctime */ + *la = *attr; + la->la_valid = LA_CTIME | LA_MTIME; + rc = mdd_update_time(env, mdd_pobj, pattr, la, handle); + if (rc) + GOTO(err_insert, rc); } - if (rc == 0 && spec->sp_cr_flags & MDS_OPEN_VOLATILE) - rc = __mdd_orphan_add(env, son, handle); - - mdd_write_unlock(env, son); - - if (rc != 0) - /* - * Object has no links, so it will be destroyed when last - * reference is released. (XXX not now.) - */ - GOTO(cleanup, rc); - - initialized = 1; - - if (!(spec->sp_cr_flags & MDS_OPEN_VOLATILE)) - rc = __mdd_index_insert(env, mdd_pobj, mdo2fid(son), - name, S_ISDIR(attr->la_mode), handle, - mdd_object_capa(env, mdd_pobj)); - - if (rc != 0) - GOTO(cleanup, rc); - - inserted = 1; - - if (S_ISLNK(attr->la_mode)) { - struct lu_ucred *uc = lu_ucred_assert(env); - 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), - uc->uc_cap & - CFS_CAP_SYS_RESOURCE_MASK); - - if (rc == sym_len) - rc = 0; - else - GOTO(cleanup, rc = -EFAULT); - } - - /* volatile file creation does not update parent directory times */ - if (spec->sp_cr_flags & MDS_OPEN_VOLATILE) - GOTO(cleanup, rc = 0); - - /* update parent directory mtime/ctime */ - *la = *attr; - la->la_valid = LA_CTIME | LA_MTIME; - rc = mdd_update_time(env, mdd_pobj, pattr, la, handle); - if (rc) - GOTO(cleanup, rc); - - EXIT; -cleanup: - if (rc != 0 && created != 0) { - int rc2; - - if (inserted != 0) { - if (spec->sp_cr_flags & MDS_OPEN_VOLATILE) - rc2 = __mdd_orphan_del(env, son, handle); - else - rc2 = __mdd_index_delete(env, mdd_pobj, name, - S_ISDIR(attr->la_mode), - handle, BYPASS_CAPA); - if (rc2 != 0) - goto out_stop; - } + EXIT; +err_insert: + if (rc != 0) { + if (spec->sp_cr_flags & MDS_OPEN_VOLATILE) + rc2 = __mdd_orphan_del(env, son, handle); + else + rc2 = __mdd_index_delete(env, mdd_pobj, name, + S_ISDIR(attr->la_mode), + handle); + if (rc2 != 0) + goto out_stop; +err_created: mdd_write_lock(env, son, MOR_TGT_CHILD); - if (initialized != 0 && S_ISDIR(attr->la_mode)) { + if (S_ISDIR(attr->la_mode)) { /* Drop the reference, no need to delete "."/"..", - * because the object to be destroied directly. */ + * because the object is to be destroyed directly. */ rc2 = mdo_ref_del(env, son, handle); if (rc2 != 0) { mdd_write_unlock(env, son); goto out_stop; } } - +out_volatile: + /* For volatile files drop one link immediately, since there is + * no filename in the namespace, and save any error returned. */ rc2 = mdo_ref_del(env, son, handle); if (rc2 != 0) { mdd_write_unlock(env, son); + if (unlikely(rc == 0)) + rc = rc2; goto out_stop; } - mdo_destroy(env, son, handle); + /* Don't destroy the volatile object on success */ + if (likely(rc != 0)) + mdo_destroy(env, son, handle); mdd_write_unlock(env, son); - } + } - if (rc == 0 && fid_is_namespace_visible(mdo2fid(son))) + if (rc == 0 && fid_is_namespace_visible(mdo2fid(son)) && + likely((spec->sp_cr_flags & MDS_OPEN_VOLATILE) == 0)) rc = mdd_changelog_ns_store(env, mdd, - S_ISDIR(attr->la_mode) ? CL_MKDIR : - S_ISREG(attr->la_mode) ? CL_CREATE : - S_ISLNK(attr->la_mode) ? CL_SOFTLINK : CL_MKNOD, - 0, son, mdd_pobj, lname, handle); + S_ISDIR(attr->la_mode) ? CL_MKDIR : + S_ISREG(attr->la_mode) ? CL_CREATE : + S_ISLNK(attr->la_mode) ? CL_SOFTLINK : CL_MKNOD, + 0, son, mdo2fid(mdd_pobj), NULL, NULL, lname, + NULL, handle); out_stop: - mdd_trans_stop(env, mdd, rc, handle); + rc2 = mdd_trans_stop(env, mdd, rc, handle); + if (rc == 0) { + /* If creation fails, it is most likely due to the remote update + * failure, because local transaction will mostly succeed at + * this stage. There is no easy way to rollback all of previous + * updates, so let's remove the object from namespace, and + * LFSCK should handle the orphan object. */ + if (rc2 < 0 && !mdd_object_remote(mdd_pobj)) + mdd_index_delete(env, mdd_pobj, attr, lname); + rc = rc2; + } out_free: - if (ldata->ld_buf && ldata->ld_buf->lb_len > OBD_ALLOC_BIG) + if (is_vmalloc_addr(ldata->ld_buf)) /* if we vmalloced a large buffer drop it */ lu_buf_free(ldata->ld_buf); - /* The child object shouldn't be cached anymore */ - if (rc) + /* The child object shouldn't be cached anymore */ + if (rc) set_bit(LU_OBJECT_HEARD_BANSHEE, - &child->mo_lu.lo_header->loh_flags); - return rc; + &child->mo_lu.lo_header->loh_flags); + return rc; } /* @@ -2334,6 +2816,17 @@ static int mdd_rename_sanity_check(const struct lu_env *env, * before mdd_rename and enable MDS_PERM_BYPASS. */ LASSERT(sobj); + /* + * If we are using project inheritance, we only allow renames + * into our tree when the project IDs are the same; otherwise + * tree quota mechanism would be circumvented. + */ + if (((tpattr->la_flags & LUSTRE_PROJINHERIT_FL) && + tpattr->la_projid != cattr->la_projid) || + ((pattr->la_flags & LUSTRE_PROJINHERIT_FL) && + (pattr->la_projid != tpattr->la_projid))) + RETURN(-EXDEV); + rc = mdd_may_delete(env, src_pobj, pattr, sobj, cattr, NULL, 1, 0); if (rc) RETURN(rc); @@ -2343,9 +2836,9 @@ static int mdd_rename_sanity_check(const struct lu_env *env, * processed in cld_rename before mdd_rename and enable * MDS_PERM_BYPASS). * So check may_create, but not check may_unlink. */ - if (!tobj) + if (tobj == NULL) rc = mdd_may_create(env, tgt_pobj, tpattr, NULL, - (src_pobj != tgt_pobj), 0); + (src_pobj != tgt_pobj)); else rc = mdd_may_delete(env, tgt_pobj, tpattr, tobj, tattr, cattr, (src_pobj != tgt_pobj), 1); @@ -2362,8 +2855,8 @@ static int mdd_declare_rename(const struct lu_env *env, struct mdd_object *mdd_tpobj, struct mdd_object *mdd_sobj, struct mdd_object *mdd_tobj, - const struct lu_name *tname, const struct lu_name *sname, + const struct lu_name *tname, struct md_attr *ma, struct linkea_data *ldata, struct thandle *handle) @@ -2393,21 +2886,22 @@ static int mdd_declare_rename(const struct lu_env *env, if (mdd_spobj != mdd_tpobj) { rc = mdo_declare_index_delete(env, mdd_sobj, dotdot, handle); - if (rc) + if (rc != 0) return rc; rc = mdo_declare_index_insert(env, mdd_sobj, mdo2fid(mdd_tpobj), - dotdot, handle); - if (rc) + S_IFDIR, dotdot, handle); + if (rc != 0) return rc; } - /* new target child can be directory, - * counted by target dir's nlink */ - rc = mdo_declare_ref_add(env, mdd_tpobj, handle); - if (rc) - return rc; - } + + /* new target child can be directory, + * counted by target dir's nlink */ + rc = mdo_declare_ref_add(env, mdd_tpobj, handle); + if (rc != 0) + return rc; + } la->la_valid = LA_CTIME | LA_MTIME; rc = mdo_declare_attr_set(env, mdd_spobj, la, handle); @@ -2427,21 +2921,20 @@ static int mdd_declare_rename(const struct lu_env *env, if (rc) return rc; - /* new name */ - rc = mdo_declare_index_insert(env, mdd_tpobj, mdo2fid(mdd_sobj), - tname->ln_name, handle); - if (rc) - return rc; - - /* name from target dir (old name), we declare it unconditionally - * as mdd_rename() calls delete unconditionally as well. so just - * to balance declarations vs calls to change ... */ - rc = mdo_declare_index_delete(env, mdd_tpobj, tname->ln_name, handle); - if (rc) - return rc; + /* new name */ + rc = mdo_declare_index_insert(env, mdd_tpobj, mdo2fid(mdd_sobj), + mdd_object_type(mdd_sobj), + tname->ln_name, handle); + if (rc != 0) + return rc; if (mdd_tobj && mdd_object_exists(mdd_tobj)) { /* delete target child in target parent directory */ + rc = mdo_declare_index_delete(env, mdd_tpobj, tname->ln_name, + handle); + if (rc) + return rc; + rc = mdo_declare_ref_del(env, mdd_tobj, handle); if (rc) return rc; @@ -2464,16 +2957,12 @@ static int mdd_declare_rename(const struct lu_env *env, if (rc) return rc; - rc = mdd_declare_links_del(env, mdd_tobj, handle); - if (rc) - return rc; - - rc = mdd_declare_finish_unlink(env, mdd_tobj, ma, handle); + rc = mdd_declare_finish_unlink(env, mdd_tobj, handle); if (rc) return rc; } - rc = mdd_declare_changelog_ext_store(env, mdd, tname, sname, handle); + rc = mdd_declare_changelog_store(env, mdd, tname, sname, handle); if (rc) return rc; @@ -2514,22 +3003,29 @@ static int mdd_rename(const struct lu_env *env, mdd_tobj = md2mdd_obj(tobj); mdd_sobj = mdd_object_find(env, mdd, lf); + if (IS_ERR(mdd_sobj)) + RETURN(PTR_ERR(mdd_sobj)); - rc = mdd_la_get(env, mdd_sobj, cattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_sobj, cattr); if (rc) GOTO(out_pending, rc); - rc = mdd_la_get(env, mdd_spobj, pattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_spobj, pattr); if (rc) GOTO(out_pending, rc); if (mdd_tobj) { - rc = mdd_la_get(env, mdd_tobj, tattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_tobj, tattr); if (rc) GOTO(out_pending, rc); + /* search for an existing archive. + * we should check ahead as the object + * can be destroyed in this transaction */ + if (mdd_hsm_archive_exists(env, mdd_tobj, ma)) + cl_flags |= CLF_RENAME_LAST_EXISTS; } - rc = mdd_la_get(env, mdd_tpobj, tpattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_tpobj, tpattr); if (rc) GOTO(out_pending, rc); @@ -2538,13 +3034,26 @@ static int mdd_rename(const struct lu_env *env, if (rc) GOTO(out_pending, rc); + rc = mdd_name_check(mdd, ltname); + if (rc < 0) + GOTO(out_pending, rc); + + /* FIXME: Should consider tobj and sobj too in rename_lock. */ + rc = mdd_rename_order(env, mdd, mdd_spobj, pattr, mdd_tpobj); + if (rc < 0) + GOTO(out_pending, rc); + handle = mdd_trans_create(env, mdd); if (IS_ERR(handle)) GOTO(out_pending, rc = PTR_ERR(handle)); memset(ldata, 0, sizeof(*ldata)); - mdd_linkea_prepare(env, mdd_sobj, NULL, NULL, mdd_object_fid(mdd_tpobj), - ltname, 1, 0, ldata); + rc = mdd_linkea_prepare(env, mdd_sobj, mdd_object_fid(mdd_spobj), + lsname, mdd_object_fid(mdd_tpobj), ltname, + 1, 0, ldata); + if (rc) + GOTO(stop, rc); + rc = mdd_declare_rename(env, mdd, mdd_spobj, mdd_tpobj, mdd_sobj, mdd_tobj, lsname, ltname, ma, ldata, handle); if (rc) @@ -2554,63 +3063,60 @@ static int mdd_rename(const struct lu_env *env, if (rc) GOTO(stop, rc); - /* FIXME: Should consider tobj and sobj too in rename_lock. */ - rc = mdd_rename_order(env, mdd, mdd_spobj, pattr, mdd_tpobj); - if (rc < 0) - GOTO(cleanup_unlocked, rc); - is_dir = S_ISDIR(cattr->la_mode); - /* Remove source name from source directory */ - rc = __mdd_index_delete(env, mdd_spobj, sname, is_dir, handle, - mdd_object_capa(env, mdd_spobj)); - if (rc) - GOTO(cleanup, rc); + /* Remove source name from source directory */ + rc = __mdd_index_delete(env, mdd_spobj, sname, is_dir, handle); + if (rc != 0) + GOTO(stop, rc); - /* "mv dir1 dir2" needs "dir1/.." link update */ - if (is_dir && mdd_sobj && !lu_fid_eq(spobj_fid, tpobj_fid)) { - rc = __mdd_index_delete_only(env, mdd_sobj, dotdot, handle, - mdd_object_capa(env, mdd_sobj)); - if (rc) - GOTO(fixup_spobj2, rc); + /* "mv dir1 dir2" needs "dir1/.." link update */ + if (is_dir && !lu_fid_eq(spobj_fid, tpobj_fid)) { + rc = __mdd_index_delete_only(env, mdd_sobj, dotdot, handle); + if (rc != 0) + GOTO(fixup_spobj2, rc); - rc = __mdd_index_insert_only(env, mdd_sobj, tpobj_fid, dotdot, - handle, mdd_object_capa(env, mdd_sobj)); - if (rc) + rc = __mdd_index_insert_only(env, mdd_sobj, tpobj_fid, S_IFDIR, + dotdot, handle); + if (rc != 0) GOTO(fixup_spobj, rc); } - /* Remove target name from target directory - * 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) { - if (mdd_tobj) { - /* tname might been renamed to something else */ - GOTO(fixup_spobj, rc); - } - if (rc != -ENOENT) - GOTO(fixup_spobj, rc); - } + if (mdd_tobj != NULL && mdd_object_exists(mdd_tobj)) { + rc = __mdd_index_delete(env, mdd_tpobj, tname, is_dir, handle); + if (rc != 0) + /* tname might been renamed to something else */ + GOTO(fixup_spobj, rc); + } /* Insert new fid with target name into target dir */ - rc = __mdd_index_insert(env, mdd_tpobj, lf, tname, is_dir, handle, - mdd_object_capa(env, mdd_tpobj)); - if (rc) + rc = __mdd_index_insert(env, mdd_tpobj, lf, cattr->la_mode, + tname, handle); + if (rc != 0) GOTO(fixup_tpobj, rc); LASSERT(ma->ma_attr.la_valid & LA_CTIME); la->la_ctime = la->la_mtime = ma->ma_attr.la_ctime; - /* XXX: mdd_sobj must be local one if it is NOT NULL. */ - if (mdd_sobj) { - la->la_valid = LA_CTIME; - rc = mdd_update_time(env, mdd_sobj, cattr, la, handle); - if (rc) - GOTO(fixup_tpobj, rc); - } + /* XXX: mdd_sobj must be local one if it is NOT NULL. */ + la->la_valid = LA_CTIME; + rc = mdd_update_time(env, mdd_sobj, cattr, la, handle); + if (rc) + GOTO(fixup_tpobj, rc); + + /* Update the linkEA for the source object */ + mdd_write_lock(env, mdd_sobj, MOR_SRC_CHILD); + rc = mdd_links_rename(env, mdd_sobj, mdo2fid(mdd_spobj), lsname, + mdo2fid(mdd_tpobj), ltname, handle, ldata, + 0, 0); + if (rc == -ENOENT) + /* Old files might not have EA entry */ + mdd_links_add(env, mdd_sobj, mdo2fid(mdd_spobj), + lsname, handle, NULL, 0); + mdd_write_unlock(env, mdd_sobj); + /* We don't fail the transaction if the link ea can't be + updated -- fid2path will use alternate lookup method. */ + rc = 0; /* Remove old target object * For tobj is remote case cmm layer has processed @@ -2634,7 +3140,7 @@ static int mdd_rename(const struct lu_env *env, tobj_ref = 1; /* fetch updated nlink */ - rc = mdd_la_get(env, mdd_tobj, tattr, BYPASS_CAPA); + rc = mdd_la_get(env, mdd_tobj, tattr); if (rc != 0) { CERROR("%s: Failed to get nlink for tobj " DFID": rc = %d\n", @@ -2656,7 +3162,8 @@ static int mdd_rename(const struct lu_env *env, /* XXX: this transfer to ma will be removed with LOD/OSP */ ma->ma_attr = *tattr; ma->ma_valid |= MA_INODE; - rc = mdd_finish_unlink(env, mdd_tobj, ma, handle); + rc = mdd_finish_unlink(env, mdd_tobj, ma, mdd_tpobj, ltname, + handle); if (rc != 0) { CERROR("%s: Failed to unlink tobj " DFID": rc = %d\n", @@ -2666,8 +3173,13 @@ static int mdd_rename(const struct lu_env *env, } /* fetch updated nlink */ - rc = mdd_la_get(env, mdd_tobj, tattr, BYPASS_CAPA); - if (rc != 0) { + rc = mdd_la_get(env, mdd_tobj, tattr); + if (rc == -ENOENT) { + /* the object got removed, let's + * return the latest known attributes */ + tattr->la_nlink = 0; + rc = 0; + } else if (rc != 0) { CERROR("%s: Failed to get nlink for tobj " DFID": rc = %d\n", mdd2obd_dev(mdd)->obd_name, @@ -2678,11 +3190,10 @@ static int mdd_rename(const struct lu_env *env, ma->ma_attr = *tattr; ma->ma_valid |= MA_INODE; - if (tattr->la_nlink == 0) { + if (tattr->la_nlink == 0) cl_flags |= CLF_RENAME_LAST; - if (mdd_hsm_archive_exists(env, mdd_tobj, ma)) - cl_flags |= CLF_RENAME_LAST_EXISTS; - } + else + cl_flags &= ~CLF_RENAME_LAST_EXISTS; } la->la_valid = LA_CTIME | LA_MTIME; @@ -2693,29 +3204,15 @@ static int mdd_rename(const struct lu_env *env, if (mdd_spobj != mdd_tpobj) { la->la_valid = LA_CTIME | LA_MTIME; rc = mdd_update_time(env, mdd_tpobj, tpattr, la, handle); + if (rc != 0) + GOTO(fixup_tpobj, rc); } - if (rc == 0 && mdd_sobj) { - mdd_write_lock(env, mdd_sobj, MOR_SRC_CHILD); - rc = mdd_links_rename(env, mdd_sobj, mdo2fid(mdd_spobj), lsname, - mdo2fid(mdd_tpobj), ltname, handle, NULL, - 0, 0); - if (rc == -ENOENT) - /* Old files might not have EA entry */ - mdd_links_add(env, mdd_sobj, mdo2fid(mdd_spobj), - lsname, handle, NULL, 0); - mdd_write_unlock(env, mdd_sobj); - /* We don't fail the transaction if the link ea can't be - updated -- fid2path will use alternate lookup method. */ - rc = 0; - } - EXIT; fixup_tpobj: if (rc) { - rc2 = __mdd_index_delete(env, mdd_tpobj, tname, is_dir, handle, - BYPASS_CAPA); + rc2 = __mdd_index_delete(env, mdd_tpobj, tname, is_dir, handle); if (rc2) CWARN("tp obj fix error %d\n",rc2); @@ -2727,63 +3224,1309 @@ fixup_tpobj: mdo_ref_add(env, mdd_tobj, handle); } - rc2 = __mdd_index_insert(env, mdd_tpobj, - mdo2fid(mdd_tobj), tname, - is_dir, handle, - BYPASS_CAPA); - - if (rc2) - CWARN("tp obj fix error %d\n",rc2); - } - } + rc2 = __mdd_index_insert(env, mdd_tpobj, + mdo2fid(mdd_tobj), + mdd_object_type(mdd_tobj), + tname, handle); + if (rc2 != 0) + CWARN("tp obj fix error: rc = %d\n", rc2); + } + } fixup_spobj: if (rc && is_dir && mdd_sobj && mdd_spobj != mdd_tpobj) { - rc2 = __mdd_index_delete_only(env, mdd_sobj, dotdot, handle, - BYPASS_CAPA); - + rc2 = __mdd_index_delete_only(env, mdd_sobj, dotdot, handle); if (rc2) CWARN("%s: sp obj dotdot delete error: rc = %d\n", mdd2obd_dev(mdd)->obd_name, rc2); - rc2 = __mdd_index_insert_only(env, mdd_sobj, spobj_fid, - dotdot, handle, BYPASS_CAPA); - if (rc2) + rc2 = __mdd_index_insert_only(env, mdd_sobj, spobj_fid, S_IFDIR, + dotdot, handle); + if (rc2 != 0) CWARN("%s: sp obj dotdot insert error: rc = %d\n", mdd2obd_dev(mdd)->obd_name, rc2); } fixup_spobj2: - if (rc) { - rc2 = __mdd_index_insert(env, mdd_spobj, - lf, sname, is_dir, handle, BYPASS_CAPA); - if (rc2) - CWARN("sp obj fix error %d\n",rc2); - } + if (rc != 0) { + rc2 = __mdd_index_insert(env, mdd_spobj, lf, + mdd_object_type(mdd_sobj), sname, + handle); + if (rc2 != 0) + CWARN("sp obj fix error: rc = %d\n", rc2); + } + cleanup: if (tobj_locked) mdd_write_unlock(env, mdd_tobj); -cleanup_unlocked: - if (rc == 0) - rc = mdd_changelog_ext_ns_store(env, mdd, CL_RENAME, cl_flags, - mdd_tobj, tpobj_fid, lf, - spobj_fid, ltname, lsname, - handle); + + if (rc == 0) + rc = mdd_changelog_ns_store(env, mdd, CL_RENAME, cl_flags, + mdd_tobj, tpobj_fid, lf, spobj_fid, + ltname, lsname, handle); stop: - mdd_trans_stop(env, mdd, rc, handle); + rc = mdd_trans_stop(env, mdd, rc, handle); + out_pending: mdd_object_put(env, mdd_sobj); return rc; } +/** + * During migration once the parent FID has been changed, + * we need update the parent FID in linkea. + **/ +static int mdd_linkea_update_child_internal(const struct lu_env *env, + struct mdd_object *parent, + struct mdd_object *newparent, + struct mdd_object *child, + const char *name, int namelen, + struct thandle *handle, + bool declare) +{ + struct mdd_thread_info *info = mdd_env_info(env); + struct linkea_data ldata = { NULL }; + struct lu_buf *buf = &info->mti_link_buf; + int count; + int rc = 0; + + ENTRY; + + buf = lu_buf_check_and_alloc(buf, PATH_MAX); + if (buf->lb_buf == NULL) + RETURN(-ENOMEM); + + ldata.ld_buf = buf; + rc = mdd_links_read(env, child, &ldata); + if (rc != 0) { + if (rc == -ENOENT || rc == -ENODATA) + rc = 0; + RETURN(rc); + } + + LASSERT(ldata.ld_leh != NULL); + ldata.ld_lee = (struct link_ea_entry *)(ldata.ld_leh + 1); + for (count = 0; count < ldata.ld_leh->leh_reccount; count++) { + struct mdd_device *mdd = mdo2mdd(&child->mod_obj); + struct lu_name lname; + struct lu_fid fid; + + linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, + &lname, &fid); + + if (strncmp(lname.ln_name, name, namelen) != 0 || + !lu_fid_eq(&fid, mdd_object_fid(parent))) { + ldata.ld_lee = (struct link_ea_entry *) + ((char *)ldata.ld_lee + + ldata.ld_reclen); + continue; + } + + CDEBUG(D_INFO, "%s: update "DFID" with %.*s:"DFID"\n", + mdd2obd_dev(mdd)->obd_name, PFID(mdd_object_fid(child)), + lname.ln_namelen, lname.ln_name, + PFID(mdd_object_fid(newparent))); + /* update to the new parent fid */ + linkea_entry_pack(ldata.ld_lee, &lname, + mdd_object_fid(newparent)); + if (declare) + rc = mdd_declare_links_add(env, child, handle, &ldata); + else + rc = mdd_links_write(env, child, &ldata, handle); + break; + } + RETURN(rc); +} + +static int mdd_linkea_declare_update_child(const struct lu_env *env, + struct mdd_object *parent, + struct mdd_object *newparent, + struct mdd_object *child, + const char *name, int namelen, + struct thandle *handle) +{ + return mdd_linkea_update_child_internal(env, parent, newparent, + child, name, + namelen, handle, true); +} + +static int mdd_linkea_update_child(const struct lu_env *env, + struct mdd_object *parent, + struct mdd_object *newparent, + struct mdd_object *child, + const char *name, int namelen, + struct thandle *handle) +{ + return mdd_linkea_update_child_internal(env, parent, newparent, + child, name, + namelen, handle, false); +} + +static int mdd_update_linkea_internal(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + const struct lu_name *child_name, + struct linkea_data *ldata, + struct thandle *handle, + int declare) +{ + struct mdd_thread_info *info = mdd_env_info(env); + int count; + int rc = 0; + ENTRY; + + LASSERT(ldata->ld_buf != NULL); + LASSERT(ldata->ld_leh != NULL); + + /* If it is mulitple links file, we need update the name entry for + * all parent */ + ldata->ld_lee = (struct link_ea_entry *)(ldata->ld_leh + 1); + for (count = 0; count < ldata->ld_leh->leh_reccount; count++) { + struct mdd_device *mdd = mdo2mdd(&mdd_sobj->mod_obj); + struct mdd_object *pobj; + struct lu_name lname; + struct lu_fid fid; + + linkea_entry_unpack(ldata->ld_lee, &ldata->ld_reclen, + &lname, &fid); + pobj = mdd_object_find(env, mdd, &fid); + if (IS_ERR(pobj)) { + CWARN("%s: cannot find obj "DFID": rc = %ld\n", + mdd2obd_dev(mdd)->obd_name, PFID(&fid), + PTR_ERR(pobj)); + continue; + } + + if (!mdd_object_exists(pobj)) { + CDEBUG(D_INFO, "%s: obj "DFID" does not exist\n", + mdd2obd_dev(mdd)->obd_name, PFID(&fid)); + goto next_put; + } + + if (pobj == mdd_pobj && + lname.ln_namelen == child_name->ln_namelen && + strncmp(lname.ln_name, child_name->ln_name, + lname.ln_namelen) == 0) { + CDEBUG(D_INFO, "%s: skip its own %s: "DFID"\n", + mdd2obd_dev(mdd)->obd_name, child_name->ln_name, + PFID(&fid)); + goto next_put; + } + + CDEBUG(D_INFO, "%s: update "DFID" with "DNAME":"DFID"\n", + mdd2obd_dev(mdd)->obd_name, PFID(mdd_object_fid(pobj)), + PNAME(&lname), PFID(mdd_object_fid(mdd_tobj))); + + if (declare) { + /* Remove source name from source directory */ + /* Insert new fid with target name into target dir */ + rc = mdo_declare_index_delete(env, pobj, lname.ln_name, + handle); + if (rc != 0) + GOTO(next_put, rc); + + rc = mdo_declare_index_insert(env, pobj, + mdd_object_fid(mdd_tobj), + mdd_object_type(mdd_tobj), + lname.ln_name, handle); + if (rc != 0) + GOTO(next_put, rc); + + rc = mdo_declare_ref_add(env, mdd_tobj, handle); + if (rc) + GOTO(next_put, rc); + + rc = mdo_declare_ref_del(env, mdd_sobj, handle); + if (rc) + GOTO(next_put, rc); + } else { + char *tmp_name = info->mti_key; + + if (lname.ln_namelen >= sizeof(info->mti_key)) { + /* lnamelen is too big(> NAME_MAX + 16), + * something wrong about this linkea, let's + * skip it */ + CWARN("%s: the name %.*s is too long under " + DFID"\n", mdd2obd_dev(mdd)->obd_name, + lname.ln_namelen, lname.ln_name, + PFID(&fid)); + goto next_put; + } + + /* Note: lname might be without \0 at the end, see + * linkea_entry_unpack(), let's add extra \0 by + * snprintf */ + snprintf(tmp_name, sizeof(info->mti_key), "%.*s", + lname.ln_namelen, lname.ln_name); + lname.ln_name = tmp_name; + + /* Let's check if this linkEA still valid, before + * it might be packed into the RPC buffer. */ + rc = mdd_lookup(env, &pobj->mod_obj, &lname, + &info->mti_fid, NULL); + if (rc < 0 || !lu_fid_eq(&info->mti_fid, + mdd_object_fid(mdd_sobj))) + GOTO(next_put, rc == -ENOENT ? 0 : rc); + + rc = __mdd_index_delete(env, pobj, tmp_name, 0, handle); + if (rc != 0) + GOTO(next_put, rc); + + rc = __mdd_index_insert(env, pobj, + mdd_object_fid(mdd_tobj), + mdd_object_type(mdd_tobj), + tmp_name, handle); + if (rc != 0) + GOTO(next_put, rc); + + mdd_write_lock(env, mdd_tobj, MOR_SRC_CHILD); + rc = mdo_ref_add(env, mdd_tobj, handle); + mdd_write_unlock(env, mdd_tobj); + if (rc) + GOTO(next_put, rc); + + mdd_write_lock(env, mdd_sobj, MOR_TGT_CHILD); + mdo_ref_del(env, mdd_sobj, handle); + mdd_write_unlock(env, mdd_sobj); + } +next_put: + mdd_object_put(env, pobj); + if (rc != 0) + break; + + ldata->ld_lee = (struct link_ea_entry *)((char *)ldata->ld_lee + + ldata->ld_reclen); + } + + RETURN(rc); +} + +static int mdd_migrate_xattrs(const struct lu_env *env, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj) +{ + struct mdd_thread_info *info = mdd_env_info(env); + struct mdd_device *mdd = mdo2mdd(&mdd_sobj->mod_obj); + char *xname; + struct thandle *handle; + struct lu_buf xbuf; + int xlen; + int rem; + int xsize; + int list_xsize; + struct lu_buf list_xbuf; + int rc; + + /* retrieve xattr list from the old object */ + list_xsize = mdo_xattr_list(env, mdd_sobj, &LU_BUF_NULL); + if (list_xsize == -ENODATA) + return 0; + + if (list_xsize < 0) + return list_xsize; + + lu_buf_check_and_alloc(&info->mti_big_buf, list_xsize); + if (info->mti_big_buf.lb_buf == NULL) + return -ENOMEM; + + list_xbuf.lb_buf = info->mti_big_buf.lb_buf; + list_xbuf.lb_len = list_xsize; + rc = mdo_xattr_list(env, mdd_sobj, &list_xbuf); + if (rc < 0) + return rc; + rc = 0; + rem = list_xsize; + xname = list_xbuf.lb_buf; + while (rem > 0) { + xlen = strnlen(xname, rem - 1) + 1; + if (strcmp(XATTR_NAME_LMA, xname) == 0 || + strcmp(XATTR_NAME_LMV, xname) == 0) + goto next; + + /* For directory, if there are default layout, migrate here */ + if (strcmp(XATTR_NAME_LOV, xname) == 0 && + !S_ISDIR(lu_object_attr(&mdd_sobj->mod_obj.mo_lu))) + goto next; + + xsize = mdo_xattr_get(env, mdd_sobj, &LU_BUF_NULL, xname); + if (xsize == -ENODATA) + goto next; + if (xsize < 0) + GOTO(out, rc); + + lu_buf_check_and_alloc(&info->mti_link_buf, xsize); + if (info->mti_link_buf.lb_buf == NULL) + GOTO(out, rc = -ENOMEM); + + xbuf.lb_len = xsize; + xbuf.lb_buf = info->mti_link_buf.lb_buf; + rc = mdo_xattr_get(env, mdd_sobj, &xbuf, xname); + if (rc == -ENODATA) + goto next; + if (rc < 0) + GOTO(out, rc); + + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + GOTO(out, rc = PTR_ERR(handle)); + + rc = mdo_declare_xattr_set(env, mdd_tobj, &xbuf, xname, 0, + handle); + if (rc != 0) + GOTO(stop_trans, rc); + /* Note: this transaction is part of migration, and it is not + * the last step of migration, so we set th_local = 1 to avoid + * update last rcvd for this transaction */ + handle->th_local = 1; + rc = mdd_trans_start(env, mdd, handle); + if (rc != 0) + GOTO(stop_trans, rc); + +again: + rc = mdo_xattr_set(env, mdd_tobj, &xbuf, xname, 0, handle); + if (rc == -EEXIST) + GOTO(stop_trans, rc = 0); + + if (unlikely(rc == -ENOSPC && + strcmp(xname, XATTR_NAME_LINK) == 0)) { + rc = linkea_overflow_shrink( + (struct linkea_data *)(xbuf.lb_buf)); + if (likely(rc > 0)) { + xbuf.lb_len = rc; + goto again; + } + } + + if (rc != 0) + GOTO(stop_trans, rc); +stop_trans: + rc = mdd_trans_stop(env, mdd, rc, handle); + if (rc != 0) + GOTO(out, rc); +next: + rem -= xlen; + memmove(xname, xname + xlen, rem); + } +out: + return rc; +} + +static int mdd_declare_migrate_create(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + struct md_op_spec *spec, + struct lu_attr *la, + union lmv_mds_md *mgr_ea, + struct linkea_data *ldata, + struct thandle *handle) +{ + struct lu_attr *la_flag = MDD_ENV_VAR(env, la_for_fix); + const struct lu_buf *buf; + int rc; + int mgr_easize; + + rc = mdd_declare_create_object_internal(env, mdd_pobj, mdd_tobj, la, + handle, spec, NULL); + if (rc != 0) + return rc; + + rc = mdd_declare_object_initialize(env, mdd_pobj, mdd_tobj, la, + handle); + if (rc != 0) + return rc; + + if (S_ISLNK(la->la_mode)) { + const char *target_name = spec->u.sp_symname; + int sym_len = strlen(target_name); + const struct lu_buf *buf; + + buf = mdd_buf_get_const(env, target_name, sym_len); + rc = dt_declare_record_write(env, mdd_object_child(mdd_tobj), + buf, 0, handle); + if (rc != 0) + return rc; + } else if (S_ISDIR(la->la_mode) && ldata != NULL) { + rc = mdd_declare_links_add(env, mdd_tobj, handle, ldata); + if (rc != 0) + return rc; + } + + if (spec->u.sp_ea.eadata != NULL && spec->u.sp_ea.eadatalen != 0) { + buf = mdd_buf_get_const(env, spec->u.sp_ea.eadata, + spec->u.sp_ea.eadatalen); + rc = mdo_declare_xattr_set(env, mdd_tobj, buf, XATTR_NAME_LOV, + 0, handle); + if (rc) + return rc; + } + + mgr_easize = lmv_mds_md_size(2, LMV_MAGIC_V1); + buf = mdd_buf_get_const(env, mgr_ea, mgr_easize); + rc = mdo_declare_xattr_set(env, mdd_sobj, buf, XATTR_NAME_LMV, + 0, handle); + if (rc) + return rc; + + la_flag->la_valid = LA_FLAGS; + la_flag->la_flags = la->la_flags | LUSTRE_IMMUTABLE_FL; + rc = mdo_declare_attr_set(env, mdd_sobj, la_flag, handle); + + return rc; +} + +static int mdd_migrate_create(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + const struct lu_name *lname, + struct lu_attr *la) +{ + struct mdd_thread_info *info = mdd_env_info(env); + struct mdd_device *mdd = mdo2mdd(&mdd_sobj->mod_obj); + struct md_op_spec *spec = &info->mti_spec; + struct lu_buf lmm_buf = { NULL }; + struct lu_buf link_buf = { NULL }; + struct lu_buf mgr_buf; + struct thandle *handle; + struct lmv_mds_md_v1 *mgr_ea; + struct lu_attr *la_flag = MDD_ENV_VAR(env, la_for_fix); + struct dt_allocation_hint *hint = &mdd_env_info(env)->mti_hint; + int mgr_easize; + struct linkea_data *ldata = &mdd_env_info(env)->mti_link_data; + int rc; + ENTRY; + + /* prepare spec for create */ + memset(spec, 0, sizeof(*spec)); + spec->sp_cr_lookup = 0; + spec->sp_feat = &dt_directory_features; + if (S_ISLNK(la->la_mode)) { + const struct lu_buf *buf; + + buf = lu_buf_check_and_alloc( + &mdd_env_info(env)->mti_big_buf, + la->la_size + 1); + link_buf = *buf; + link_buf.lb_len = la->la_size + 1; + memset(link_buf.lb_buf, 0, link_buf.lb_len); + rc = mdd_readlink(env, &mdd_sobj->mod_obj, &link_buf); + if (rc <= 0) { + rc = rc != 0 ? rc : -EFAULT; + CERROR("%s: "DFID" readlink failed: rc = %d\n", + mdd2obd_dev(mdd)->obd_name, + PFID(mdd_object_fid(mdd_sobj)), rc); + RETURN(rc); + } + spec->u.sp_symname = link_buf.lb_buf; + } else if (S_ISREG(la->la_mode)) { + /* retrieve lov of the old object */ + rc = mdd_get_lov_ea(env, mdd_sobj, &lmm_buf); + if (rc != 0 && rc != -ENODATA) + RETURN(rc); + if (lmm_buf.lb_buf != NULL && lmm_buf.lb_len != 0) { + spec->u.sp_ea.eadata = lmm_buf.lb_buf; + spec->u.sp_ea.eadatalen = lmm_buf.lb_len; + spec->sp_cr_flags |= MDS_OPEN_HAS_EA; + } + } else if (S_ISDIR(la->la_mode)) { + rc = mdd_links_read_with_rec(env, mdd_sobj, ldata); + if (rc == -ENODATA) { + /* ignore the non-linkEA error */ + ldata = NULL; + rc = 0; + } + if (rc < 0) + RETURN(rc); + } + + mgr_easize = lmv_mds_md_size(2, LMV_MAGIC_V1); + lu_buf_check_and_alloc(&info->mti_xattr_buf, mgr_easize); + mgr_buf.lb_buf = info->mti_xattr_buf.lb_buf; + mgr_buf.lb_len = mgr_easize; + mgr_ea = mgr_buf.lb_buf; + memset(mgr_ea, 0, sizeof(*mgr_ea)); + mgr_ea->lmv_magic = cpu_to_le32(LMV_MAGIC_V1); + mgr_ea->lmv_stripe_count = cpu_to_le32(2); + mgr_ea->lmv_master_mdt_index = mdd_seq_site(mdd)->ss_node_id; + mgr_ea->lmv_hash_type = cpu_to_le32(LMV_HASH_FLAG_MIGRATION); + fid_cpu_to_le(&mgr_ea->lmv_stripe_fids[0], mdd_object_fid(mdd_sobj)); + fid_cpu_to_le(&mgr_ea->lmv_stripe_fids[1], mdd_object_fid(mdd_tobj)); + + mdd_object_make_hint(env, mdd_pobj, mdd_tobj, la, spec, hint); + + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + GOTO(out_free, rc = PTR_ERR(handle)); + + /* Note: this transaction is part of migration, and it is not + * the last step of migration, so we set th_local = 1 to avoid + * update last rcvd for this transaction */ + handle->th_local = 1; + rc = mdd_declare_migrate_create(env, mdd_pobj, mdd_sobj, mdd_tobj, spec, + la, mgr_buf.lb_buf, ldata, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + rc = mdd_trans_start(env, mdd, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + /* don't set nlink from the original object */ + la->la_valid &= ~LA_NLINK; + + /* create the target object */ + rc = mdd_create_object(env, mdd_pobj, mdd_tobj, la, spec, NULL, NULL, + hint, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + if (S_ISDIR(la->la_mode) && ldata != NULL) { + rc = mdd_links_write(env, mdd_tobj, ldata, handle); + if (rc != 0) + GOTO(stop_trans, rc); + } + + /* Set MIGRATE EA on the source inode, so once the migration needs + * to be re-done during failover, the re-do process can locate the + * target object which is already being created. */ + rc = mdo_xattr_set(env, mdd_sobj, &mgr_buf, XATTR_NAME_LMV, 0, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + /* Set immutable flag, so any modification is disabled until + * the migration is done. Once the migration is interrupted, + * if the resume process find the migrating object has both + * IMMUTALBE flag and MIGRATE EA, it need to clear IMMUTABLE + * flag and approve the migration */ + la_flag->la_valid = LA_FLAGS; + la_flag->la_flags = la->la_flags | LUSTRE_IMMUTABLE_FL; + rc = mdo_attr_set(env, mdd_sobj, la_flag, handle); +stop_trans: + if (handle != NULL) + rc = mdd_trans_stop(env, mdd, rc, handle); +out_free: + if (lmm_buf.lb_buf != NULL) + OBD_FREE(lmm_buf.lb_buf, lmm_buf.lb_len); + RETURN(rc); +} + +static int mdd_migrate_entries(const struct lu_env *env, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj) +{ + struct dt_object *next = mdd_object_child(mdd_sobj); + struct mdd_device *mdd = mdo2mdd(&mdd_sobj->mod_obj); + struct dt_object *dt_tobj = mdd_object_child(mdd_tobj); + struct thandle *handle; + struct dt_it *it; + const struct dt_it_ops *iops; + int result; + struct lu_dirent *ent; + int rc; + ENTRY; + + OBD_ALLOC(ent, NAME_MAX + sizeof(*ent) + 1); + if (ent == NULL) + RETURN(-ENOMEM); + + if (!dt_try_as_dir(env, next)) + GOTO(out_ent, rc = -ENOTDIR); + /* + * iterate directories + */ + iops = &next->do_index_ops->dio_it; + it = iops->init(env, next, LUDA_FID | LUDA_TYPE); + if (IS_ERR(it)) + GOTO(out_ent, rc = PTR_ERR(it)); + + rc = iops->load(env, it, 0); + if (rc == 0) + rc = iops->next(env, it); + else if (rc > 0) + rc = 0; + /* + * At this point and across for-loop: + * + * rc == 0 -> ok, proceed. + * rc > 0 -> end of directory. + * rc < 0 -> error. + */ + do { + struct mdd_object *child; + char *name = mdd_env_info(env)->mti_key; + int len; + int is_dir; + bool target_exist = false; + + len = iops->key_size(env, it); + if (len == 0) + goto next; + + result = iops->rec(env, it, (struct dt_rec *)ent, + LUDA_FID | LUDA_TYPE); + if (result == -ESTALE) + goto next; + if (result != 0) { + rc = result; + goto out; + } + + fid_le_to_cpu(&ent->lde_fid, &ent->lde_fid); + + /* Insert new fid with target name into target dir */ + if ((ent->lde_namelen == 1 && ent->lde_name[0] == '.') || + (ent->lde_namelen == 2 && ent->lde_name[0] == '.' && + ent->lde_name[1] == '.')) + goto next; + + child = mdd_object_find(env, mdd, &ent->lde_fid); + if (IS_ERR(child)) + GOTO(out, rc = PTR_ERR(child)); + + mdd_write_lock(env, child, MOR_SRC_CHILD); + is_dir = S_ISDIR(mdd_object_type(child)); + + snprintf(name, ent->lde_namelen + 1, "%s", ent->lde_name); + + /* Check whether the name has been inserted to the target */ + if (dt_try_as_dir(env, dt_tobj)) { + struct lu_fid *fid = &mdd_env_info(env)->mti_fid2; + + rc = dt_lookup(env, dt_tobj, (struct dt_rec *)fid, + (struct dt_key *)name); + if (unlikely(rc == 0)) + target_exist = true; + } + + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + GOTO(out_put, rc = PTR_ERR(handle)); + + /* Note: this transaction is part of migration, and it is not + * the last step of migration, so we set th_local = 1 to avoid + * updating last rcvd for this transaction */ + handle->th_local = 1; + if (likely(!target_exist)) { + rc = mdo_declare_index_insert(env, mdd_tobj, + &ent->lde_fid, + mdd_object_type(child), + name, handle); + if (rc != 0) + GOTO(out_put, rc); + + if (is_dir) { + rc = mdo_declare_ref_add(env, mdd_tobj, handle); + if (rc != 0) + GOTO(out_put, rc); + } + } + + rc = mdo_declare_index_delete(env, mdd_sobj, name, handle); + if (rc != 0) + GOTO(out_put, rc); + + if (is_dir) { + rc = mdo_declare_ref_del(env, mdd_sobj, handle); + if (rc != 0) + GOTO(out_put, rc); + + /* Update .. for child */ + rc = mdo_declare_index_delete(env, child, dotdot, + handle); + if (rc != 0) + GOTO(out_put, rc); + + rc = mdo_declare_index_insert(env, child, + mdd_object_fid(mdd_tobj), + S_IFDIR, dotdot, handle); + if (rc != 0) + GOTO(out_put, rc); + } + + rc = mdd_linkea_declare_update_child(env, mdd_sobj,mdd_tobj, + child, name, + strlen(name), + handle); + if (rc != 0) + GOTO(out_put, rc); + + rc = mdd_trans_start(env, mdd, handle); + if (rc != 0) { + CERROR("%s: transaction start failed: rc = %d\n", + mdd2obd_dev(mdd)->obd_name, rc); + GOTO(out_put, rc); + } + + if (likely(!target_exist)) { + rc = __mdd_index_insert(env, mdd_tobj, &ent->lde_fid, + mdd_object_type(child), + name, handle); + if (rc != 0) + GOTO(out_put, rc); + } + + rc = __mdd_index_delete(env, mdd_sobj, name, is_dir, handle); + if (rc != 0) + GOTO(out_put, rc); + + if (is_dir) { + rc = __mdd_index_delete_only(env, child, dotdot, + handle); + if (rc != 0) + GOTO(out_put, rc); + + rc = __mdd_index_insert_only(env, child, + mdd_object_fid(mdd_tobj), S_IFDIR, + dotdot, handle); + if (rc != 0) + GOTO(out_put, rc); + } + + rc = mdd_linkea_update_child(env, mdd_sobj, mdd_tobj, + child, name, + strlen(name), handle); + +out_put: + mdd_write_unlock(env, child); + mdd_object_put(env, child); + rc = mdd_trans_stop(env, mdd, rc, handle); + if (rc != 0) + GOTO(out, rc); +next: + result = iops->next(env, it); + if (OBD_FAIL_CHECK(OBD_FAIL_MIGRATE_ENTRIES)) + GOTO(out, rc = -EINTR); + + if (result == -ESTALE) + goto next; + } while (result == 0); +out: + iops->put(env, it); + iops->fini(env, it); +out_ent: + OBD_FREE(ent, NAME_MAX + sizeof(*ent) + 1); + RETURN(rc); +} + +static int mdd_declare_update_linkea(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + const struct lu_name *child_name, + struct linkea_data *ldata, + struct thandle *handle) +{ + return mdd_update_linkea_internal(env, mdd_pobj, mdd_sobj, mdd_tobj, + child_name, ldata, handle, 1); +} + +static int mdd_update_linkea(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + const struct lu_name *child_name, + struct linkea_data *ldata, + struct thandle *handle) +{ + return mdd_update_linkea_internal(env, mdd_pobj, mdd_sobj, mdd_tobj, + child_name, ldata, handle, 0); +} + +static int mdd_declare_migrate_update_name(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + const struct lu_name *lname, + struct lu_attr *la, + struct lu_attr *parent_la, + struct linkea_data *ldata, + struct thandle *handle) +{ + struct mdd_device *mdd = mdo2mdd(&mdd_sobj->mod_obj); + struct lu_attr *la_flag = MDD_ENV_VAR(env, tattr); + int rc; + + /* Revert IMMUTABLE flag */ + la_flag->la_valid = LA_FLAGS; + la_flag->la_flags = la->la_flags & ~LUSTRE_IMMUTABLE_FL; + rc = mdo_declare_attr_set(env, mdd_sobj, la_flag, handle); + if (rc != 0) + return rc; + + /* delete entry from source dir */ + rc = mdo_declare_index_delete(env, mdd_pobj, lname->ln_name, handle); + if (rc != 0) + return rc; + + if (ldata->ld_buf != NULL) { + rc = mdd_declare_update_linkea(env, mdd_pobj, mdd_sobj, + mdd_tobj, lname, ldata, handle); + if (rc != 0) + return rc; + } + + if (S_ISREG(mdd_object_type(mdd_sobj))) { + rc = mdo_declare_xattr_del(env, mdd_sobj, XATTR_NAME_LOV, + handle); + if (rc != 0) + return rc; + + handle->th_complex = 1; + rc = mdo_declare_xattr_set(env, mdd_tobj, NULL, + XATTR_NAME_FID, + LU_XATTR_REPLACE, handle); + if (rc < 0) + return rc; + } + + if (S_ISDIR(mdd_object_type(mdd_sobj))) { + rc = mdo_declare_ref_del(env, mdd_pobj, handle); + if (rc != 0) + return rc; + } + + /* new name */ + rc = mdo_declare_index_insert(env, mdd_pobj, mdo2fid(mdd_tobj), + mdd_object_type(mdd_tobj), + lname->ln_name, handle); + if (rc != 0) + return rc; + + rc = mdd_declare_links_add(env, mdd_tobj, handle, NULL); + if (rc != 0) + return rc; + + if (S_ISDIR(mdd_object_type(mdd_sobj))) { + rc = mdo_declare_ref_add(env, mdd_pobj, handle); + if (rc != 0) + return rc; + } + + /* delete old object */ + rc = mdo_declare_ref_del(env, mdd_sobj, handle); + if (rc != 0) + return rc; + + if (S_ISDIR(mdd_object_type(mdd_sobj))) { + /* delete old object */ + rc = mdo_declare_ref_del(env, mdd_sobj, handle); + if (rc != 0) + return rc; + /* set nlink to 0 */ + rc = mdo_declare_attr_set(env, mdd_sobj, la, handle); + if (rc != 0) + return rc; + } + + rc = mdd_declare_finish_unlink(env, mdd_sobj, handle); + if (rc) + return rc; + + rc = mdo_declare_attr_set(env, mdd_pobj, parent_la, handle); + if (rc != 0) + return rc; + + rc = mdd_declare_changelog_store(env, mdd, lname, NULL, handle); + + return rc; +} + +static int mdd_migrate_update_name(const struct lu_env *env, + struct mdd_object *mdd_pobj, + struct mdd_object *mdd_sobj, + struct mdd_object *mdd_tobj, + const struct lu_name *lname, + struct md_attr *ma) +{ + struct lu_attr *p_la = MDD_ENV_VAR(env, la_for_fix); + struct lu_attr *so_attr = MDD_ENV_VAR(env, cattr); + struct lu_attr *la_flag = MDD_ENV_VAR(env, tattr); + struct mdd_device *mdd = mdo2mdd(&mdd_sobj->mod_obj); + struct linkea_data *ldata = &mdd_env_info(env)->mti_link_data; + struct thandle *handle; + int is_dir = S_ISDIR(mdd_object_type(mdd_sobj)); + const char *name = lname->ln_name; + int rc; + ENTRY; + + /* update time for parent */ + LASSERT(ma->ma_attr.la_valid & LA_CTIME); + p_la->la_ctime = p_la->la_mtime = ma->ma_attr.la_ctime; + p_la->la_valid = LA_CTIME; + + rc = mdd_la_get(env, mdd_sobj, so_attr); + if (rc != 0) + RETURN(rc); + + ldata->ld_buf = NULL; + rc = mdd_links_read(env, mdd_sobj, ldata); + if (rc != 0 && rc != -ENOENT && rc != -ENODATA) + RETURN(rc); + + handle = mdd_trans_create(env, mdd); + if (IS_ERR(handle)) + RETURN(PTR_ERR(handle)); + + rc = mdd_declare_migrate_update_name(env, mdd_pobj, mdd_sobj, mdd_tobj, + lname, so_attr, p_la, ldata, + handle); + if (rc != 0) { + /* If the migration can not be fit in one transaction, just + * leave it in the original MDT */ + if (rc == -E2BIG) + GOTO(stop_trans, rc = 0); + else + GOTO(stop_trans, rc); + } + + CDEBUG(D_INFO, "%s: update "DFID"/"DFID" with %s:"DFID"\n", + mdd2obd_dev(mdd)->obd_name, PFID(mdd_object_fid(mdd_pobj)), + PFID(mdd_object_fid(mdd_sobj)), lname->ln_name, + PFID(mdd_object_fid(mdd_tobj))); + + rc = mdd_trans_start(env, mdd, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + /* Revert IMMUTABLE flag */ + la_flag->la_valid = LA_FLAGS; + la_flag->la_flags = so_attr->la_flags & ~LUSTRE_IMMUTABLE_FL; + rc = mdo_attr_set(env, mdd_sobj, la_flag, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + /* Remove source name from source directory */ + rc = __mdd_index_delete(env, mdd_pobj, name, is_dir, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + if (ldata->ld_buf != NULL) { + rc = mdd_update_linkea(env, mdd_pobj, mdd_sobj, mdd_tobj, + lname, ldata, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + /* linkea update might decrease the source object + * nlink, let's get the attr again after ref_del */ + rc = mdd_la_get(env, mdd_sobj, so_attr); + if (rc != 0) + GOTO(stop_trans, rc); + } + + if (S_ISREG(so_attr->la_mode)) { + if (so_attr->la_nlink == 1) { + rc = mdo_xattr_del(env, mdd_sobj, XATTR_NAME_LOV, + handle); + if (rc != 0 && rc != -ENODATA) + GOTO(stop_trans, rc); + + rc = mdo_xattr_set(env, mdd_tobj, NULL, + XATTR_NAME_FID, + LU_XATTR_REPLACE, handle); + if (rc < 0) + GOTO(stop_trans, rc); + } + } + + /* Insert new fid with target name into target dir */ + rc = __mdd_index_insert(env, mdd_pobj, mdd_object_fid(mdd_tobj), + mdd_object_type(mdd_tobj), name, handle); + if (rc != 0) + GOTO(stop_trans, rc); + + mdd_write_lock(env, mdd_sobj, MOR_TGT_CHILD); + + mdd_sobj->mod_flags |= DEAD_OBJ; + rc = mdd_mark_orphan_object(env, mdd_sobj, handle, false); + if (rc != 0) + GOTO(out_unlock, rc); + + rc = __mdd_orphan_add(env, mdd_sobj, handle); + if (rc != 0) + GOTO(out_unlock, rc); + + mdo_ref_del(env, mdd_sobj, handle); + if (is_dir) + mdo_ref_del(env, mdd_sobj, handle); + + /* Get the attr again after ref_del */ + rc = mdd_la_get(env, mdd_sobj, so_attr); + if (rc != 0) + GOTO(out_unlock, rc); + + ma->ma_attr = *so_attr; + ma->ma_valid |= MA_INODE; + + rc = mdd_attr_set_internal(env, mdd_pobj, p_la, handle, 0); + if (rc != 0) + GOTO(out_unlock, rc); + + rc = mdd_changelog_ns_store(env, mdd, CL_MIGRATE, 0, mdd_tobj, + mdo2fid(mdd_pobj), mdo2fid(mdd_sobj), + mdo2fid(mdd_pobj), lname, lname, handle); + if (rc != 0) { + CWARN("%s: changelog for migrate %s "DFID + "under "DFID" failed: rc = %d\n", + mdd2obd_dev(mdd)->obd_name, lname->ln_name, + PFID(mdd_object_fid(mdd_sobj)), + PFID(mdd_object_fid(mdd_pobj)), rc); + /* Sigh, there are no easy way to migrate back the object, so + * let's reset the result to 0 for now XXX */ + rc = 0; + } +out_unlock: + mdd_write_unlock(env, mdd_sobj); + +stop_trans: + rc = mdd_trans_stop(env, mdd, rc, handle); + + RETURN(rc); +} + +static int mdd_fld_lookup(const struct lu_env *env, struct mdd_device *mdd, + const struct lu_fid *fid, __u32 *mdt_index) +{ + struct lu_seq_range *range = &mdd_env_info(env)->mti_range; + struct seq_server_site *ss; + int rc; + + ss = mdd->mdd_md_dev.md_lu_dev.ld_site->ld_seq_site; + + range->lsr_flags = LU_SEQ_RANGE_MDT; + rc = fld_server_lookup(env, ss->ss_server_fld, fid->f_seq, range); + if (rc != 0) + return rc; + + *mdt_index = range->lsr_index; + + return 0; +} +/** + * Check whether we should migrate the file/dir + * return val + * < 0 permission check failed or other error. + * = 0 the file can be migrated. + * > 0 the file does not need to be migrated, mostly + * for multiple link file + **/ +static int mdd_migrate_sanity_check(const struct lu_env *env, + struct mdd_object *pobj, + const struct lu_attr *pattr, + struct mdd_object *sobj, + struct lu_attr *sattr) +{ + struct mdd_thread_info *info = mdd_env_info(env); + struct linkea_data *ldata = &info->mti_link_data; + struct mdd_device *mdd = mdo2mdd(&pobj->mod_obj); + int mgr_easize; + struct lu_buf *mgr_buf; + int count; + int rc; + __u64 mdt_index; + ENTRY; + + mgr_easize = lmv_mds_md_size(2, LMV_MAGIC_V1); + mgr_buf = lu_buf_check_and_alloc(&info->mti_big_buf, mgr_easize); + if (mgr_buf->lb_buf == NULL) + RETURN(-ENOMEM); + + rc = mdo_xattr_get(env, sobj, mgr_buf, XATTR_NAME_LMV); + if (rc > 0) { + union lmv_mds_md *lmm = mgr_buf->lb_buf; + + /* If the object has migrateEA, it means IMMUTE flag + * is being set by previous migration process, so it + * needs to override the IMMUTE flag, otherwise the + * following sanity check will fail */ + if (le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type) & + LMV_HASH_FLAG_MIGRATION) { + struct mdd_device *mdd = mdo2mdd(&sobj->mod_obj); + + sattr->la_flags &= ~LUSTRE_IMMUTABLE_FL; + CDEBUG(D_HA, "%s: "DFID" override IMMUTE FLAG\n", + mdd2obd_dev(mdd)->obd_name, + PFID(mdd_object_fid(sobj))); + } + } + + rc = mdd_rename_sanity_check(env, pobj, pattr, pobj, pattr, + sobj, sattr, NULL, NULL); + if (rc != 0) + RETURN(rc); + + /* Then it will check if the file should be migrated. If the file + * has mulitple links, we only need migrate the file if all of its + * entries has been migrated to the remote MDT */ + if (!S_ISREG(sattr->la_mode) || sattr->la_nlink < 2) + RETURN(0); + + rc = mdd_links_read(env, sobj, ldata); + if (rc != 0) { + /* For multiple links files, if there are no linkEA data at all, + * means the file might be created before linkEA is enabled, and + * all of its links should not be migrated yet, otherwise it + * should have some linkEA there */ + if (rc == -ENOENT || rc == -ENODATA) + RETURN(1); + RETURN(rc); + } + + mdt_index = mdd->mdd_md_dev.md_lu_dev.ld_site->ld_seq_site->ss_node_id; + /* If there are still links locally, then the file will not be + * migrated. */ + LASSERT(ldata->ld_leh != NULL); + + /* If the linkEA is overflow, then means there are some unknown name + * entries under unknown parents, that will prevent the migration. */ + if (unlikely(ldata->ld_leh->leh_overflow_time)) + RETURN(1); + + ldata->ld_lee = (struct link_ea_entry *)(ldata->ld_leh + 1); + for (count = 0; count < ldata->ld_leh->leh_reccount; count++) { + struct lu_name lname; + struct lu_fid fid; + __u32 parent_mdt_index; + + linkea_entry_unpack(ldata->ld_lee, &ldata->ld_reclen, + &lname, &fid); + ldata->ld_lee = (struct link_ea_entry *)((char *)ldata->ld_lee + + ldata->ld_reclen); + + rc = mdd_fld_lookup(env, mdd, &fid, &parent_mdt_index); + if (rc != 0) + RETURN(rc); + + /* Migrate the object only if none of its parents are on the + * current MDT. */ + if (parent_mdt_index != mdt_index) + continue; + + CDEBUG(D_INFO, DFID"still has local entry %.*s "DFID"\n", + PFID(mdd_object_fid(sobj)), lname.ln_namelen, + lname.ln_name, PFID(&fid)); + rc = 1; + break; + } + + RETURN(rc); +} + +static int mdd_migrate(const struct lu_env *env, struct md_object *pobj, + struct md_object *sobj, const struct lu_name *lname, + struct md_object *tobj, struct md_attr *ma) +{ + struct mdd_object *mdd_pobj = md2mdd_obj(pobj); + struct mdd_device *mdd = mdo2mdd(pobj); + struct mdd_object *mdd_sobj = md2mdd_obj(sobj); + struct mdd_object *mdd_tobj = md2mdd_obj(tobj); + struct lu_attr *so_attr = MDD_ENV_VAR(env, cattr); + struct lu_attr *pattr = MDD_ENV_VAR(env, pattr); + bool created = false; + int rc; + + ENTRY; + /* If the file will being migrated, it will check whether + * the file is being opened by someone else right now */ + mdd_read_lock(env, mdd_sobj, MOR_SRC_CHILD); + if (mdd_sobj->mod_count > 0) { + CDEBUG(D_OTHER, + "%s: "DFID"%s is already opened count %d: rc = %d\n", + mdd2obd_dev(mdd)->obd_name, + PFID(mdd_object_fid(mdd_sobj)), lname->ln_name, + mdd_sobj->mod_count, -EBUSY); + mdd_read_unlock(env, mdd_sobj); + GOTO(put, rc = -EBUSY); + } + mdd_read_unlock(env, mdd_sobj); + + rc = mdd_la_get(env, mdd_sobj, so_attr); + if (rc != 0) + GOTO(put, rc); + + rc = mdd_la_get(env, mdd_pobj, pattr); + if (rc != 0) + GOTO(put, rc); + + rc = mdd_migrate_sanity_check(env, mdd_pobj, pattr, mdd_sobj, so_attr); + if (rc != 0) { + if (rc > 0) + rc = 0; + GOTO(put, rc); + } + + /* Sigh, it is impossible to finish all of migration in a single + * transaction, for example migrating big directory entries to the + * new MDT, it needs insert all of name entries of children in the + * new directory. + * + * So migration will be done in multiple steps and transactions. + * + * 1. create an orphan object on the remote MDT in one transaction. + * 2. migrate extend attributes to the new target file/directory. + * 3. For directory, migrate the entries to the new MDT and update + * linkEA of each children. Because we can not migrate all entries + * in a single transaction, so the migrating directory will become + * a striped directory during migration, so once the process is + * interrupted, the directory is still accessible. (During lookup, + * client will locate the name by searching both original and target + * object). + * 4. Finally, update the name/FID to point to the new file/directory + * in a separate transaction. + */ + + /* step 1: Check whether the orphan object has been created, and create + * orphan object on the remote MDT if needed */ + if (!mdd_object_exists(mdd_tobj)) { + rc = mdd_migrate_create(env, mdd_pobj, mdd_sobj, mdd_tobj, + lname, so_attr); + if (rc != 0) + GOTO(put, rc); + created = true; + } + + LASSERT(mdd_object_exists(mdd_tobj)); + /* step 2: migrate xattr */ + rc = mdd_migrate_xattrs(env, mdd_sobj, mdd_tobj); + if (rc != 0) + GOTO(put, rc); + + /* step 3: migrate name entries to the orphan object */ + if (S_ISDIR(lu_object_attr(&mdd_sobj->mod_obj.mo_lu))) { + rc = mdd_migrate_entries(env, mdd_sobj, mdd_tobj); + if (rc != 0) + GOTO(put, rc); + if (unlikely(OBD_FAIL_CHECK_RESET(OBD_FAIL_MIGRATE_NET_REP, + OBD_FAIL_MDS_REINT_NET_REP))) + GOTO(put, rc = 0); + } else { + OBD_FAIL_TIMEOUT(OBD_FAIL_MIGRATE_DELAY, cfs_fail_val); + } + + LASSERT(mdd_object_exists(mdd_tobj)); + /* step 4: update name entry to the new object */ + rc = mdd_migrate_update_name(env, mdd_pobj, mdd_sobj, mdd_tobj, lname, + ma); + if (rc != 0) + GOTO(put, rc); + + /* newly created target was not locked, don't cache its attributes */ + if (created) + mdd_invalidate(env, tobj); +put: + RETURN(rc); +} + const 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_create_data = mdd_create_data, + .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_create_data = mdd_create_data, + .mdo_migrate = mdd_migrate, };