X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmdt%2Fmdt_xattr.c;h=82d94c77309f7dba0c734ffac68c5af93f68dc5c;hp=83ecc742f93311ed2cf4672aed387919b3b21728;hb=4d8bc239c2c30a47e8833cf23db6ccd39ff61705;hpb=3db2db358a6e427bd7461c52dbd498de808a6659 diff --git a/lustre/mdt/mdt_xattr.c b/lustre/mdt/mdt_xattr.c index 83ecc74..82d94c7 100644 --- a/lustre/mdt/mdt_xattr.c +++ b/lustre/mdt/mdt_xattr.c @@ -1,387 +1,634 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: +/* + * GPL HEADER START * - * linux/mdt/mdt_xattr.c - * Lustre Metadata Target (mdt) extended attributes management. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * Copyright (C) 2002-2006 Cluster File Systems, Inc. - * Author: Peter Braam - * Author: Andreas Dilger - * Author: Phil Schwan - * Author: Huang Hua + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. * - * This file is part of the Lustre file system, http://www.lustre.org - * Lustre is a trademark of Cluster File Systems, Inc. + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). * - * You may have signed or agreed to another license before downloading - * this software. If so, you are bound by the terms and conditions - * of that agreement, and the following does not apply to you. See the - * LICENSE file included with this distribution for more information. + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.gnu.org/licenses/gpl-2.0.html * - * If you did not agree to a different license, then this copy of Lustre - * is open source software; you can redistribute it and/or modify it - * under the terms of version 2 of the GNU General Public License as - * published by the Free Software Foundation. + * GPL HEADER END + */ +/* + * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2017, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * lustre/mdt/mdt_xattr.c * - * In either case, Lustre is distributed in the hope that it will be - * useful, but WITHOUT ANY WARRANTY; without even the implied warranty - * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * license text for more details. + * Lustre Metadata Target (mdt) extended attributes management. + * + * Author: Peter Braam + * Author: Andreas Dilger + * Author: Phil Schwan + * Author: Huang Hua */ -#ifndef EXPORT_SYMTAB -# define EXPORT_SYMTAB -#endif #define DEBUG_SUBSYSTEM S_MDS -/* prerequisite for linux/xattr.h */ -#include -/* prerequisite for linux/xattr.h */ -#include -/* XATTR_{REPLACE,CREATE} */ #include - +#include +#include +#include #include "mdt_internal.h" /* return EADATA length to the caller. negative value means error */ static int mdt_getxattr_pack_reply(struct mdt_thread_info * info) { - struct req_capsule *pill = info->mti_pill ; - struct ptlrpc_request *req = mdt_info_req(info); - char *xattr_name; - __u64 valid = info->mti_body->valid; - static const char user_string[] = "user."; - int size, rc; - ENTRY; - - if (OBD_FAIL_CHECK(OBD_FAIL_MDS_GETXATTR_PACK)) - RETURN(-ENOMEM); - - /* Determine how many bytes we need */ - if ((valid & OBD_MD_FLXATTR) == OBD_MD_FLXATTR) { - xattr_name = req_capsule_client_get(pill, &RMF_NAME); - if (!xattr_name) - RETURN(-EFAULT); - - if (!(req->rq_export->exp_connect_flags & OBD_CONNECT_XATTR) && - !strncmp(xattr_name, user_string, sizeof(user_string) - 1)) - RETURN(-EOPNOTSUPP); - - size = mo_xattr_get(info->mti_env, - mdt_object_child(info->mti_object), - &LU_BUF_NULL, xattr_name); - } else if ((valid & OBD_MD_FLXATTRLS) == OBD_MD_FLXATTRLS) { - size = mo_xattr_list(info->mti_env, - mdt_object_child(info->mti_object), - &LU_BUF_NULL); - } else { - CDEBUG(D_INFO, "Valid bits: "LPX64"\n", info->mti_body->valid); - RETURN(-EINVAL); - } - - if (size < 0) { - if (size == -ENODATA) - size = 0; - else if (size != -EOPNOTSUPP) { - CDEBUG(D_INFO, "Error geting EA size: %d\n", size); - RETURN(size); - } - } - - if (info->mti_body->eadatasize != 0 && - info->mti_body->eadatasize < size) - RETURN(-ERANGE); - - req_capsule_set_size(pill, &RMF_EADATA, RCL_SERVER, - min_t(int, size, info->mti_body->eadatasize)); + struct req_capsule *pill = info->mti_pill; + struct ptlrpc_request *req = mdt_info_req(info); + const char *xattr_name; + u64 valid; + static const char user_string[] = "user."; + int size; + int rc = 0; + int rc2; + ENTRY; + + valid = info->mti_body->mbo_valid & (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS); + + /* Determine how many bytes we need */ + if (valid == OBD_MD_FLXATTR) { + xattr_name = req_capsule_client_get(pill, &RMF_NAME); + if (!xattr_name) + RETURN(-EFAULT); + + if (!(exp_connect_flags(req->rq_export) & OBD_CONNECT_XATTR) && + !strncmp(xattr_name, user_string, sizeof(user_string) - 1)) + RETURN(-EOPNOTSUPP); + + size = mo_xattr_get(info->mti_env, + mdt_object_child(info->mti_object), + &LU_BUF_NULL, xattr_name); + if (size == -ENODATA) { + /* XXX: Some client code will not handle -ENODATA + * for XATTR_NAME_LOV (trusted.lov) properly. */ + if (strcmp(xattr_name, XATTR_NAME_LOV) == 0) + rc = 0; + else + rc = -ENODATA; + + size = 0; + } + } else if (valid == OBD_MD_FLXATTRLS) { + xattr_name = "list"; + size = mo_xattr_list(info->mti_env, + mdt_object_child(info->mti_object), + &LU_BUF_NULL); + } else if (valid == OBD_MD_FLXATTRALL) { + xattr_name = "all"; + /* N.B. eadatasize = 0 is not valid for FLXATTRALL */ + /* We could calculate accurate sizes, but this would + * introduce a lot of overhead, let's do it later... */ + size = info->mti_body->mbo_eadatasize; + if (size <= 0 || size > info->mti_mdt->mdt_max_ea_size || + size & (sizeof(__u32) - 1)) { + DEBUG_REQ(D_ERROR, req, + "%s: invalid EA size(%d) for FLXATTRALL\n", + mdt_obd_name(info->mti_mdt), size); + RETURN(-EINVAL); + } + req_capsule_set_size(pill, &RMF_EAVALS, RCL_SERVER, size); + req_capsule_set_size(pill, &RMF_EAVALS_LENS, RCL_SERVER, size); + } else { + CDEBUG(D_INFO, "Valid bits: %#llx\n", + info->mti_body->mbo_valid); + RETURN(-EINVAL); + } + + if (size < 0) { + if (size != -EOPNOTSUPP && size != -ENOENT) + CERROR("%s: error geting EA size for '%s': rc = %d\n", + mdt_obd_name(info->mti_mdt), xattr_name, size); + RETURN(size); + } + + if (req_capsule_has_field(pill, &RMF_ACL, RCL_SERVER)) + req_capsule_set_size(pill, &RMF_ACL, RCL_SERVER, + LUSTRE_POSIX_ACL_MAX_SIZE_OLD); + + req_capsule_set_size(pill, &RMF_EADATA, RCL_SERVER, + info->mti_body->mbo_eadatasize == 0 ? 0 : size); + + rc2 = req_capsule_server_pack(pill); + if (rc2 < 0) + RETURN(rc2); + + if (OBD_FAIL_CHECK(OBD_FAIL_MDS_GETXATTR_PACK)) + RETURN(-ENOMEM); + + RETURN(rc < 0 ? rc : size); +} - rc = req_capsule_server_pack(pill); - if (rc) { - LASSERT(rc < 0); - RETURN(rc); - } +static int mdt_nodemap_map_acl(struct mdt_thread_info *info, void *buf, + size_t size, const char *name, + enum nodemap_tree_type tree_type) +{ + struct lu_nodemap *nodemap; + struct obd_export *exp = info->mti_exp; + int rc = size; + + ENTRY; + + if (strcmp(name, XATTR_NAME_ACL_ACCESS) == 0 || + strcmp(name, XATTR_NAME_ACL_DEFAULT) == 0) { + if (size > info->mti_mdt->mdt_max_ea_size || + (!exp_connect_large_acl(exp) && + size > LUSTRE_POSIX_ACL_MAX_SIZE_OLD)) + GOTO(out, rc = -ERANGE); + + nodemap = nodemap_get_from_exp(exp); + if (IS_ERR(nodemap)) + GOTO(out, rc = PTR_ERR(nodemap)); + + rc = nodemap_map_acl(nodemap, buf, size, tree_type); + nodemap_putref(nodemap); + if (rc < 0) + GOTO(out, rc); + } +out: + RETURN(rc); +} - RETURN(size); +static int mdt_getxattr_all(struct mdt_thread_info *info, + struct mdt_body *reqbody, struct mdt_body *repbody, + struct lu_buf *buf, struct md_object *next) +{ + const struct lu_env *env = info->mti_env; + char *v, *b, *eadatahead, *eadatatail; + __u32 *sizes; + int eadatasize, eavallen = 0, eavallens = 0, rc; + + ENTRY; + + /* + * The format of the pill is the following: + * EADATA: attr1\0attr2\0...attrn\0 + * EAVALS: val1val2...valn + * EAVALS_LENS: 4,4,...4 + */ + + eadatahead = buf->lb_buf; + + /* Fill out EADATA first */ + rc = mo_xattr_list(env, next, buf); + if (rc < 0) + GOTO(out_shrink, rc); + + eadatasize = rc; + eadatatail = eadatahead + eadatasize; + + v = req_capsule_server_get(info->mti_pill, &RMF_EAVALS); + sizes = req_capsule_server_get(info->mti_pill, &RMF_EAVALS_LENS); + + /* Fill out EAVALS and EAVALS_LENS */ + for (b = eadatahead; b < eadatatail; b += strlen(b) + 1, v += rc) { + buf->lb_buf = v; + buf->lb_len = reqbody->mbo_eadatasize - eavallen; + rc = mo_xattr_get(env, next, buf, b); + if (rc < 0) + GOTO(out_shrink, rc); + rc = mdt_nodemap_map_acl(info, buf->lb_buf, rc, b, + NODEMAP_FS_TO_CLIENT); + if (rc < 0) + GOTO(out_shrink, rc); + sizes[eavallens] = rc; + eavallens++; + eavallen += rc; + } + +out_shrink: + if (rc < 0) { + eadatasize = 0; + eavallens = 0; + eavallen = 0; + } + repbody->mbo_aclsize = eavallen; + repbody->mbo_max_mdsize = eavallens; + + req_capsule_shrink(info->mti_pill, &RMF_EAVALS, eavallen, RCL_SERVER); + req_capsule_shrink(info->mti_pill, &RMF_EAVALS_LENS, + eavallens * sizeof(__u32), RCL_SERVER); + req_capsule_shrink(info->mti_pill, &RMF_EADATA, eadatasize, RCL_SERVER); + + if (rc >= 0) + RETURN(eadatasize); + return rc; } int mdt_getxattr(struct mdt_thread_info *info) { - struct ptlrpc_request *req = mdt_info_req(info); - struct mdt_export_data *med = mdt_req2med(req); - struct md_ucred *uc = mdt_ucred(info); + struct ptlrpc_request *req = mdt_info_req(info); struct mdt_body *reqbody; struct mdt_body *repbody = NULL; struct md_object *next; struct lu_buf *buf; int easize, rc; + u64 valid; ENTRY; LASSERT(info->mti_object != NULL); - LASSERT(lu_object_assert_exists(&info->mti_object->mot_obj.mo_lu)); + LASSERT(lu_object_assert_exists(&info->mti_object->mot_obj)); + + CDEBUG(D_INODE, "getxattr "DFID"\n", PFID(&info->mti_body->mbo_fid1)); - CDEBUG(D_INODE, "getxattr "DFID"\n", PFID(&info->mti_body->fid1)); + rc = req_check_sepol(info->mti_pill); + if (rc) + RETURN(err_serious(rc)); reqbody = req_capsule_client_get(info->mti_pill, &RMF_MDT_BODY); if (reqbody == NULL) RETURN(err_serious(-EFAULT)); - rc = mdt_init_ucred(info, reqbody); + rc = mdt_init_ucred(info, reqbody); if (rc) RETURN(err_serious(rc)); next = mdt_object_child(info->mti_object); - - if (info->mti_body->valid & OBD_MD_FLRMTRGETFACL) { - __u32 perm = mdt_identity_get_perm(uc->mu_identity, - med->med_rmtclient, - req->rq_peer.nid); - - LASSERT(med->med_rmtclient); - if (!(perm & CFS_RMTACL_PERM)) - GOTO(out, rc = err_serious(-EPERM)); - - rc = mo_permission(info->mti_env, NULL, next, NULL, - MAY_RGETFACL); - if (rc) - GOTO(out, rc = err_serious(rc)); - } - easize = mdt_getxattr_pack_reply(info); - if (easize < 0) - GOTO(out, rc = err_serious(easize)); + if (easize == -ENODATA) + GOTO(out, rc = easize); + else if (easize < 0) + GOTO(out, rc = err_serious(easize)); repbody = req_capsule_server_get(info->mti_pill, &RMF_MDT_BODY); LASSERT(repbody != NULL); /* No need further getxattr. */ - if (easize == 0 || reqbody->eadatasize == 0) - GOTO(out, rc = easize); - - buf = &info->mti_buf; - buf->lb_buf = req_capsule_server_get(info->mti_pill, &RMF_EADATA); - buf->lb_len = easize; - - if (info->mti_body->valid & OBD_MD_FLXATTR) { - int flags = CFS_IC_NOTHING; - char *xattr_name = req_capsule_client_get(info->mti_pill, - &RMF_NAME); - CDEBUG(D_INODE, "getxattr %s\n", xattr_name); - - rc = mo_xattr_get(info->mti_env, next, buf, xattr_name); - if (rc < 0) { - CERROR("getxattr failed: %d\n", rc); - GOTO(out, rc); - } - - if (info->mti_body->valid & - (OBD_MD_FLRMTLSETFACL | OBD_MD_FLRMTLGETFACL)) - flags = CFS_IC_ALL; - else if (info->mti_body->valid & OBD_MD_FLRMTRGETFACL) - flags = CFS_IC_MAPPED; - - if (rc > 0 && flags != CFS_IC_NOTHING) { - int rc1; - - LASSERT(med->med_rmtclient); - rc1 = lustre_posix_acl_xattr_id2client(uc, - med->med_idmap, - (posix_acl_xattr_header *)(buf->lb_buf), - rc, flags); - if (unlikely(rc1 < 0)) - rc = rc1; - } - } else if (info->mti_body->valid & OBD_MD_FLXATTRLS) { - CDEBUG(D_INODE, "listxattr\n"); - - rc = mo_xattr_list(info->mti_env, next, buf); - if (rc < 0) - CDEBUG(D_INFO, "listxattr failed: %d\n", rc); - } else - LBUG(); - - EXIT; + if (easize == 0 || reqbody->mbo_eadatasize == 0) + GOTO(out, rc = easize); + + buf = &info->mti_buf; + buf->lb_buf = req_capsule_server_get(info->mti_pill, &RMF_EADATA); + buf->lb_len = easize; + + valid = info->mti_body->mbo_valid & (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS); + + if (valid == OBD_MD_FLXATTR) { + const char *xattr_name = req_capsule_client_get(info->mti_pill, + &RMF_NAME); + rc = mo_xattr_get(info->mti_env, next, buf, xattr_name); + if (rc < 0) + GOTO(out, rc); + + rc = mdt_nodemap_map_acl(info, buf->lb_buf, rc, xattr_name, + NODEMAP_FS_TO_CLIENT); + } else if (valid == OBD_MD_FLXATTRLS) { + CDEBUG(D_INODE, "listxattr\n"); + + rc = mo_xattr_list(info->mti_env, next, buf); + if (rc < 0) + CDEBUG(D_INFO, "listxattr failed: %d\n", rc); + } else if (valid == OBD_MD_FLXATTRALL) { + rc = mdt_getxattr_all(info, reqbody, repbody, + buf, next); + } else + LBUG(); + + EXIT; out: - if (rc >= 0) { - repbody->eadatasize = rc; - rc = 0; - } - mdt_exit_ucred(info); - return rc; + if (rc >= 0) { + mdt_counter_incr(req, LPROC_MDT_GETXATTR); + /* LU-11109: Set OBD_MD_FLXATTR on success so that + * newer clients can distinguish between nonexistent + * xattrs and zero length values. */ + repbody->mbo_valid |= OBD_MD_FLXATTR; + repbody->mbo_eadatasize = rc; + rc = 0; + } + mdt_exit_ucred(info); + return rc; } -static int mdt_rmtlsetfacl(struct mdt_thread_info *info, - struct md_object *next, - const char *xattr_name, - ext_acl_xattr_header *header, - posix_acl_xattr_header **out) +/* shrink dir layout after migration */ +static int mdt_dir_layout_shrink(struct mdt_thread_info *info) { - struct ptlrpc_request *req = mdt_info_req(info); - struct mdt_export_data *med = mdt_req2med(req); - struct md_ucred *uc = mdt_ucred(info); - struct lu_buf *buf = &info->mti_buf; - int rc; - ENTRY; - - rc = lustre_ext_acl_xattr_id2server(uc, med->med_idmap, header); - if (rc) - RETURN(rc); - - rc = mo_xattr_get(info->mti_env, next, &LU_BUF_NULL, xattr_name); - if (rc == -ENODATA) - rc = 0; - else if (rc < 0) - RETURN(rc); - - buf->lb_len = rc; - if (buf->lb_len > 0) { - OBD_ALLOC(buf->lb_buf, buf->lb_len); - if (unlikely(buf->lb_buf == NULL)) - RETURN(-ENOMEM); - - rc = mo_xattr_get(info->mti_env, next, buf, xattr_name); - if (rc < 0) { - CERROR("getxattr failed: %d\n", rc); - GOTO(_out, rc); - } - } else - buf->lb_buf = NULL; - - rc = lustre_acl_xattr_merge2posix((posix_acl_xattr_header *)(buf->lb_buf), - buf->lb_len, header, out); - EXIT; - -_out: - if (rc <= 0 && buf->lb_buf != NULL) - OBD_FREE(buf->lb_buf, buf->lb_len); - return rc; + const struct lu_env *env = info->mti_env; + struct mdt_device *mdt = info->mti_mdt; + struct lu_ucred *uc = mdt_ucred(info); + struct mdt_reint_record *rr = &info->mti_rr; + struct lmv_user_md *lmu = rr->rr_eadata; + __u32 lum_stripe_count = lmu->lum_stripe_count; + struct lu_buf *buf = &info->mti_buf; + struct lmv_mds_md_v1 *lmv; + struct md_attr *ma = &info->mti_attr; + struct ldlm_enqueue_info *einfo = &info->mti_einfo[0]; + struct mdt_object *pobj = NULL; + struct mdt_object *obj; + struct mdt_lock_handle *lhp = NULL; + struct mdt_lock_handle *lhc; + int rc; + + ENTRY; + + if (!mdt->mdt_enable_dir_migration) + RETURN(-EPERM); + + if (!md_capable(uc, CFS_CAP_SYS_ADMIN) && + uc->uc_gid != mdt->mdt_enable_remote_dir_gid && + mdt->mdt_enable_remote_dir_gid != -1) + RETURN(-EPERM); + + /* mti_big_lmm is used to save LMV, but it may be uninitialized. */ + if (unlikely(!info->mti_big_lmm)) { + info->mti_big_lmmsize = lmv_mds_md_size(64, LMV_MAGIC); + OBD_ALLOC(info->mti_big_lmm, info->mti_big_lmmsize); + if (!info->mti_big_lmm) + RETURN(-ENOMEM); + } + + obj = mdt_object_find(env, mdt, rr->rr_fid1); + if (IS_ERR(obj)) + RETURN(PTR_ERR(obj)); + + /* get parent from PFID */ + rc = mdt_attr_get_pfid(info, obj, &ma->ma_pfid); + if (rc) + GOTO(put_obj, rc); + + pobj = mdt_object_find(env, mdt, &ma->ma_pfid); + if (IS_ERR(pobj)) + GOTO(put_obj, rc = PTR_ERR(pobj)); + + /* revoke object remote LOOKUP lock */ + if (mdt_object_remote(pobj)) { + rc = mdt_revoke_remote_lookup_lock(info, pobj, obj); + if (rc) + GOTO(put_pobj, rc); + } + + /* + * lock parent if dir will be shrunk to 1 stripe, because dir will be + * converted to normal directory, as will change dir fid and update + * namespace of parent. + */ + lhp = &info->mti_lh[MDT_LH_PARENT]; + mdt_lock_reg_init(lhp, LCK_PW); + + if (le32_to_cpu(lmu->lum_stripe_count) < 2) { + rc = mdt_reint_object_lock(info, pobj, lhp, + MDS_INODELOCK_UPDATE, true); + if (rc) + GOTO(put_pobj, rc); + } + + /* lock object */ + lhc = &info->mti_lh[MDT_LH_CHILD]; + mdt_lock_reg_init(lhc, LCK_EX); + rc = mdt_reint_striped_lock(info, obj, lhc, MDS_INODELOCK_FULL, einfo, + true); + if (rc) + GOTO(unlock_pobj, rc); + + ma->ma_lmv = info->mti_big_lmm; + ma->ma_lmv_size = info->mti_big_lmmsize; + ma->ma_valid = 0; + rc = mdt_stripe_get(info, obj, ma, XATTR_NAME_LMV); + if (rc) + GOTO(unlock_obj, rc); + + /* user may run 'lfs migrate' multiple times, so it's shrunk already */ + if (!(ma->ma_valid & MA_LMV)) + GOTO(unlock_obj, rc = -EALREADY); + + lmv = &ma->ma_lmv->lmv_md_v1; + + /* ditto */ + if (!(le32_to_cpu(lmv->lmv_hash_type) & LMV_HASH_FLAG_MIGRATION)) + GOTO(unlock_obj, rc = -EALREADY); + + lum_stripe_count = lmu->lum_stripe_count; + if (!lum_stripe_count) + lum_stripe_count = cpu_to_le32(1); + + if (lmv->lmv_migrate_offset != lum_stripe_count) { + CERROR("%s: "DFID" migrate mdt count mismatch %u != %u\n", + mdt_obd_name(info->mti_mdt), PFID(rr->rr_fid1), + lmv->lmv_migrate_offset, lmu->lum_stripe_count); + GOTO(unlock_obj, rc = -EINVAL); + } + + if (lmv->lmv_master_mdt_index != lmu->lum_stripe_offset) { + CERROR("%s: "DFID" migrate mdt index mismatch %u != %u\n", + mdt_obd_name(info->mti_mdt), PFID(rr->rr_fid1), + lmv->lmv_master_mdt_index, lmu->lum_stripe_offset); + GOTO(unlock_obj, rc = -EINVAL); + } + + if (lum_stripe_count > 1 && + (lmv->lmv_hash_type & cpu_to_le32(LMV_HASH_TYPE_MASK)) != + lmu->lum_hash_type) { + CERROR("%s: "DFID" migrate mdt hash mismatch %u != %u\n", + mdt_obd_name(info->mti_mdt), PFID(rr->rr_fid1), + lmv->lmv_hash_type, lmu->lum_hash_type); + GOTO(unlock_obj, rc = -EINVAL); + } + + buf->lb_buf = rr->rr_eadata; + buf->lb_len = rr->rr_eadatalen; + rc = mo_xattr_set(env, mdt_object_child(obj), buf, XATTR_NAME_LMV, 0); + GOTO(unlock_obj, rc); + +unlock_obj: + mdt_reint_striped_unlock(info, obj, lhc, einfo, rc); +unlock_pobj: + mdt_object_unlock(info, pobj, lhp, rc); +put_pobj: + mdt_object_put(env, pobj); +put_obj: + mdt_object_put(env, obj); + + return rc; } int mdt_reint_setxattr(struct mdt_thread_info *info, struct mdt_lock_handle *unused) { - struct ptlrpc_request *req = mdt_info_req(info); - struct mdt_export_data *med = mdt_req2med(req); - struct md_ucred *uc = mdt_ucred(info); - const char user_string[] = "user."; - const char trust_string[] = "trusted."; - struct mdt_lock_handle *lh; - struct req_capsule *pill = info->mti_pill; - const struct lu_env *env = info->mti_env; - struct lu_buf *buf = &info->mti_buf; - struct mdt_reint_record *rr = &info->mti_rr; - struct lu_attr *attr = &info->mti_attr.ma_attr; - struct mdt_object *obj; - struct md_object *child; - __u64 valid = attr->la_valid; - const char *xattr_name; - int xattr_len = 0; - __u64 lockpart; - int rc; - posix_acl_xattr_header *new_xattr = NULL; - ENTRY; - - CDEBUG(D_INODE, "setxattr for "DFID"\n", PFID(rr->rr_fid1)); + struct ptlrpc_request *req = mdt_info_req(info); + struct mdt_lock_handle *lh; + const struct lu_env *env = info->mti_env; + struct lu_buf *buf = &info->mti_buf; + struct mdt_reint_record *rr = &info->mti_rr; + struct md_attr *ma = &info->mti_attr; + struct lu_attr *attr = &info->mti_attr.ma_attr; + struct mdt_object *obj; + struct md_object *child; + __u64 valid = attr->la_valid; + const char *xattr_name = rr->rr_name.ln_name; + int xattr_len = rr->rr_eadatalen; + __u64 lockpart = MDS_INODELOCK_UPDATE; + int rc; + ENTRY; + + CDEBUG(D_INODE, "setxattr for "DFID": %s %s\n", PFID(rr->rr_fid1), + valid & OBD_MD_FLXATTR ? "set" : "remove", xattr_name); + + if (info->mti_dlm_req) + ldlm_request_cancel(req, info->mti_dlm_req, 0, LATF_SKIP); if (OBD_FAIL_CHECK(OBD_FAIL_MDS_SETXATTR)) RETURN(err_serious(-ENOMEM)); - xattr_name = rr->rr_name; - - CDEBUG(D_INODE, "%s xattr %s\n", - valid & OBD_MD_FLXATTR ? "set" : "remove", xattr_name); - - rc = mdt_init_ucred_reint(info); + rc = mdt_init_ucred_reint(info); if (rc != 0) RETURN(rc); - if (valid & OBD_MD_FLRMTRSETFACL) { - __u32 perm = mdt_identity_get_perm(uc->mu_identity, - med->med_rmtclient, - req->rq_peer.nid); - - LASSERT(med->med_rmtclient); - if (!(perm & CFS_RMTACL_PERM)) - GOTO(out, rc = err_serious(-EPERM)); - } - - /* various sanity check for xattr name */ - xattr_name = req_capsule_client_get(pill, &RMF_NAME); - if (!xattr_name) - GOTO(out, rc = err_serious(-EFAULT)); - - if (strncmp(xattr_name, trust_string, sizeof(trust_string) - 1) == 0) { - if (strcmp(xattr_name + 8, XATTR_NAME_LOV) == 0) - GOTO(out, rc = -EACCES); - } - - if (!(req->rq_export->exp_connect_flags & OBD_CONNECT_XATTR) && - (strncmp(xattr_name, user_string, sizeof(user_string) - 1) == 0)) { - GOTO(out, rc = -EOPNOTSUPP); - } - - lockpart = MDS_INODELOCK_UPDATE; + if (strncmp(xattr_name, XATTR_USER_PREFIX, + sizeof(XATTR_USER_PREFIX) - 1) == 0) { + if (!(exp_connect_flags(req->rq_export) & OBD_CONNECT_XATTR)) + GOTO(out, rc = -EOPNOTSUPP); + } else if (strncmp(xattr_name, XATTR_TRUSTED_PREFIX, + sizeof(XATTR_TRUSTED_PREFIX) - 1) == 0) { + + /* setxattr(LMV) with lum is used to shrink dir layout */ + if (strcmp(xattr_name, XATTR_NAME_LMV) == 0) { + __u32 *magic = rr->rr_eadata; + + /* we don't let to remove LMV? */ + if (!rr->rr_eadata) + GOTO(out, rc = 0); + + if (le32_to_cpu(*magic) == LMV_USER_MAGIC || + le32_to_cpu(*magic) == LMV_USER_MAGIC_SPECIFIC) { + rc = mdt_dir_layout_shrink(info); + GOTO(out, rc); + } + } + + if (!md_capable(mdt_ucred(info), CFS_CAP_SYS_ADMIN)) + GOTO(out, rc = -EPERM); + + if (strcmp(xattr_name, XATTR_NAME_LOV) == 0 || + strcmp(xattr_name, XATTR_NAME_LMA) == 0 || + strcmp(xattr_name, XATTR_NAME_LMV) == 0 || + strcmp(xattr_name, XATTR_NAME_LINK) == 0 || + strcmp(xattr_name, XATTR_NAME_FID) == 0 || + strcmp(xattr_name, XATTR_NAME_VERSION) == 0 || + strcmp(xattr_name, XATTR_NAME_SOM) == 0 || + strcmp(xattr_name, XATTR_NAME_HSM) == 0 || + strcmp(xattr_name, XATTR_NAME_LFSCK_NAMESPACE) == 0) + GOTO(out, rc = 0); + } else if ((valid & OBD_MD_FLXATTR) && + (strcmp(xattr_name, XATTR_NAME_ACL_ACCESS) == 0 || + strcmp(xattr_name, XATTR_NAME_ACL_DEFAULT) == 0)) { + rc = mdt_nodemap_map_acl(info, rr->rr_eadata, xattr_len, + xattr_name, NODEMAP_CLIENT_TO_FS); + if (rc < 0) + GOTO(out, rc); + /* ACLs were mapped out, return an error so the user knows */ + if (rc != xattr_len) + GOTO(out, rc = -EPERM); + } else if ((strlen(xattr_name) > strlen(XATTR_LUSTRE_LOV) + 1) && + strncmp(xattr_name, XATTR_LUSTRE_LOV, + strlen(XATTR_LUSTRE_LOV)) == 0) { + + if (strncmp(xattr_name, XATTR_LUSTRE_LOV".add", + strlen(XATTR_LUSTRE_LOV".add")) && + strncmp(xattr_name, XATTR_LUSTRE_LOV".set", + strlen(XATTR_LUSTRE_LOV".set")) && + strncmp(xattr_name, XATTR_LUSTRE_LOV".del", + strlen(XATTR_LUSTRE_LOV".del"))) { + CERROR("%s: invalid xattr name: %s\n", + mdt_obd_name(info->mti_mdt), xattr_name); + GOTO(out, rc = -EINVAL); + } + + lockpart |= MDS_INODELOCK_LAYOUT; + } + + /* Revoke all clients' lookup lock, since the access + * permissions for this inode is changed when ACL_ACCESS is + * set. This isn't needed for ACL_DEFAULT, since that does + * not change the access permissions of this inode, nor any + * other existing inodes. It is setting the ACLs inherited + * by new directories/files at create time. */ + /* We need revoke both LOOKUP|PERM lock here, see mdt_attr_set. */ if (!strcmp(xattr_name, XATTR_NAME_ACL_ACCESS)) - lockpart |= MDS_INODELOCK_LOOKUP; + lockpart |= MDS_INODELOCK_PERM | MDS_INODELOCK_LOOKUP; + /* We need to take the lock on behalf of old clients so that newer + * clients flush their xattr caches */ + else + lockpart |= MDS_INODELOCK_XATTR; lh = &info->mti_lh[MDT_LH_PARENT]; - mdt_lock_reg_init(lh, LCK_PW); + /* ACLs were sent to clients under LCK_CR locks, so taking LCK_EX + * to cancel them. */ + mdt_lock_reg_init(lh, LCK_EX); obj = mdt_object_find_lock(info, rr->rr_fid1, lh, lockpart); - if (IS_ERR(obj)) - GOTO(out, rc = PTR_ERR(obj)); - - child = mdt_object_child(obj); - if ((valid & OBD_MD_FLXATTR) == OBD_MD_FLXATTR) { - char * xattr; - - if (!req_capsule_field_present(pill, &RMF_EADATA, RCL_CLIENT)) { - CDEBUG(D_INFO, "no xattr data supplied\n"); - GOTO(out_unlock, rc = -EFAULT); + if (IS_ERR(obj)) + GOTO(out, rc = PTR_ERR(obj)); + + tgt_vbr_obj_set(env, mdt_obj2dt(obj)); + rc = mdt_version_get_check_save(info, obj, 0); + if (rc) + GOTO(out_unlock, rc); + + if (unlikely(!(valid & OBD_MD_FLCTIME))) { + /* This isn't strictly an error, but all current clients + * should set OBD_MD_FLCTIME when setting attributes. */ + CWARN("%s: client miss to set OBD_MD_FLCTIME when " + "setxattr %s: [object "DFID"] [valid %llu]\n", + mdt_obd_name(info->mti_mdt), xattr_name, + PFID(rr->rr_fid1), valid); + attr->la_ctime = ktime_get_real_seconds(); + } + attr->la_valid = LA_CTIME; + child = mdt_object_child(obj); + if (valid & OBD_MD_FLXATTR) { + int flags = 0; + + if (attr->la_flags & XATTR_REPLACE) + flags |= LU_XATTR_REPLACE; + + if (attr->la_flags & XATTR_CREATE) + flags |= LU_XATTR_CREATE; + + mdt_fail_write(env, info->mti_mdt->mdt_bottom, + OBD_FAIL_MDS_SETXATTR_WRITE); + + buf->lb_buf = rr->rr_eadata; + buf->lb_len = xattr_len; + rc = mo_xattr_set(env, child, buf, xattr_name, flags); + /* update ctime after xattr changed */ + if (rc == 0) { + ma->ma_attr_flags |= MDS_PERM_BYPASS; + mo_attr_set(env, child, ma); + } + } else if (valid & OBD_MD_FLXATTRRM) { + rc = mo_xattr_del(env, child, xattr_name); + /* update ctime after xattr changed */ + if (rc == 0) { + ma->ma_attr_flags |= MDS_PERM_BYPASS; + mo_attr_set(env, child, ma); } + } else { + CDEBUG(D_INFO, "valid bits: %#llx\n", valid); + rc = -EINVAL; + } - xattr_len = req_capsule_get_size(pill, &RMF_EADATA, RCL_CLIENT); - if (xattr_len) { - int flags = 0; - - xattr = req_capsule_client_get(pill, &RMF_EADATA); + if (rc == 0) + mdt_counter_incr(req, LPROC_MDT_SETXATTR); - if (valid & OBD_MD_FLRMTLSETFACL) { - LASSERT(med->med_rmtclient); - xattr_len = mdt_rmtlsetfacl(info, child, - xattr_name, - (ext_acl_xattr_header *)xattr, - &new_xattr); - if (xattr_len < 0) - GOTO(out_unlock, rc = xattr_len); - - xattr = (char *)new_xattr; - } - - if (attr->la_flags & XATTR_REPLACE) - flags |= LU_XATTR_REPLACE; - - if (attr->la_flags & XATTR_CREATE) - flags |= LU_XATTR_CREATE; - - mdt_fail_write(env, info->mti_mdt->mdt_bottom, - OBD_FAIL_MDS_SETXATTR_WRITE); - - buf->lb_buf = xattr; - buf->lb_len = xattr_len; - rc = mo_xattr_set(env, child, buf, xattr_name, flags); - } - } else if ((valid & OBD_MD_FLXATTRRM) == OBD_MD_FLXATTRRM) { - rc = mo_xattr_del(env, child, xattr_name); - } else { - CDEBUG(D_INFO, "valid bits: "LPX64"\n", valid); - rc = -EINVAL; - } EXIT; out_unlock: mdt_object_unlock_put(info, obj, lh, rc); - if (unlikely(new_xattr != NULL)) - lustre_posix_acl_xattr_free(new_xattr, xattr_len); out: - mdt_exit_ucred(info); - return rc; + mdt_exit_ucred(info); + return rc; }