4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
34 #include <linux/sched.h>
36 #include <linux/xattr.h>
37 #include <linux/selinux.h>
39 #define DEBUG_SUBSYSTEM S_LLITE
41 #include <obd_support.h>
42 #include <lustre_dlm.h>
43 #include <lustre_ver.h>
44 #include <lustre_eacl.h>
46 #include "llite_internal.h"
48 /* xattr related to IMA(Integrity Measurement Architecture) */
49 #ifndef XATTR_NAME_IMA
50 #define XATTR_NAME_IMA "security.ima"
52 #ifndef XATTR_NAME_EVM
53 #define XATTR_NAME_EVM "security.evm"
56 #define XATTR_USER_T (1)
57 #define XATTR_TRUSTED_T (2)
58 #define XATTR_SECURITY_T (3)
59 #define XATTR_ACL_ACCESS_T (4)
60 #define XATTR_ACL_DEFAULT_T (5)
61 #define XATTR_LUSTRE_T (6)
62 #define XATTR_OTHER_T (7)
65 int get_xattr_type(const char *name)
67 if (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS))
68 return XATTR_ACL_ACCESS_T;
70 if (!strcmp(name, XATTR_NAME_POSIX_ACL_DEFAULT))
71 return XATTR_ACL_DEFAULT_T;
73 if (!strncmp(name, XATTR_USER_PREFIX,
74 sizeof(XATTR_USER_PREFIX) - 1))
77 if (!strncmp(name, XATTR_TRUSTED_PREFIX,
78 sizeof(XATTR_TRUSTED_PREFIX) - 1))
79 return XATTR_TRUSTED_T;
81 if (!strncmp(name, XATTR_SECURITY_PREFIX,
82 sizeof(XATTR_SECURITY_PREFIX) - 1))
83 return XATTR_SECURITY_T;
85 if (!strncmp(name, XATTR_LUSTRE_PREFIX,
86 sizeof(XATTR_LUSTRE_PREFIX) - 1))
87 return XATTR_LUSTRE_T;
93 int xattr_type_filter(struct ll_sb_info *sbi, int xattr_type)
95 if ((xattr_type == XATTR_ACL_ACCESS_T ||
96 xattr_type == XATTR_ACL_DEFAULT_T) &&
97 !(sbi->ll_flags & LL_SBI_ACL))
100 if (xattr_type == XATTR_USER_T && !(sbi->ll_flags & LL_SBI_USER_XATTR))
102 if (xattr_type == XATTR_TRUSTED_T && !cfs_capable(CFS_CAP_SYS_ADMIN))
104 if (xattr_type == XATTR_OTHER_T)
111 int ll_setxattr_common(struct inode *inode, const char *name,
112 const void *value, size_t size,
113 int flags, __u64 valid)
115 struct ll_sb_info *sbi = ll_i2sbi(inode);
116 struct ptlrpc_request *req = NULL;
118 const char *pv = value;
121 /*FIXME: enable IMA when the conditions are ready */
122 if (strncmp(name, XATTR_NAME_IMA,
123 sizeof(XATTR_NAME_IMA)) == 0 ||
124 strncmp(name, XATTR_NAME_EVM,
125 sizeof(XATTR_NAME_EVM)) == 0)
128 xattr_type = get_xattr_type(name);
129 rc = xattr_type_filter(sbi, xattr_type);
133 if ((xattr_type == XATTR_ACL_ACCESS_T ||
134 xattr_type == XATTR_ACL_DEFAULT_T) &&
135 #ifdef HAVE_INODE_OWNER_OR_CAPABLE
136 !inode_owner_or_capable(inode))
138 !is_owner_or_cap(inode))
142 /* b10667: ignore lustre special xattr for now */
143 if (strcmp(name, XATTR_NAME_HSM) == 0 ||
144 (xattr_type == XATTR_TRUSTED_T &&
145 strcmp(name, XATTR_NAME_LOV) == 0) ||
146 (xattr_type == XATTR_LUSTRE_T &&
147 strcmp(name, "lustre.lov") == 0))
150 /* LU-549: Disable security.selinux when selinux is disabled */
151 if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
152 strcmp(name, "security.selinux") == 0)
155 /* In user.* namespace, only regular files and directories can have
156 * extended attributes. */
157 if (xattr_type == XATTR_USER_T) {
158 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
162 rc = md_setxattr(sbi->ll_md_exp, ll_inode2fid(inode), valid, name, pv,
163 size, 0, flags, ll_i2suppgid(inode), &req);
165 if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
166 LCONSOLE_INFO("Disabling user_xattr feature because "
167 "it is not supported on the server\n");
168 sbi->ll_flags &= ~LL_SBI_USER_XATTR;
173 ptlrpc_req_finished(req);
177 static int get_hsm_state(struct inode *inode, __u32 *hus_states)
179 struct md_op_data *op_data;
180 struct hsm_user_state *hus;
187 op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0,
188 LUSTRE_OPC_ANY, hus);
189 if (!IS_ERR(op_data)) {
190 rc = obd_iocontrol(LL_IOC_HSM_STATE_GET, ll_i2mdexp(inode),
191 sizeof(*op_data), op_data, NULL);
193 *hus_states = hus->hus_states;
195 CDEBUG(D_VFSTRACE, "obd_iocontrol failed. rc = %d\n",
198 ll_finish_md_op_data(op_data);
200 rc = PTR_ERR(op_data);
201 CDEBUG(D_VFSTRACE, "Could not prepare the opdata. rc = %d\n",
208 static int ll_adjust_lum(struct inode *inode, struct lov_user_md *lump)
210 struct lov_comp_md_v1 *comp_v1 = (struct lov_comp_md_v1 *)lump;
211 struct lov_user_md *v1 = lump;
212 bool release_checked = false;
213 bool need_clear_release = false;
214 __u16 entry_count = 1;
215 bool is_composite = false;
222 if (lump->lmm_magic == LOV_USER_MAGIC_COMP_V1) {
223 entry_count = comp_v1->lcm_entry_count;
227 for (i = 0; i < entry_count; i++) {
228 if (lump->lmm_magic == LOV_USER_MAGIC_COMP_V1)
229 v1 = (struct lov_user_md *)((char *)comp_v1 +
230 comp_v1->lcm_entries[i].lcme_offset);
232 /* Attributes that are saved via getxattr will always
233 * have the stripe_offset as 0. Instead, the MDS
234 * should be allowed to pick the starting OST index.
236 if (!is_composite && v1->lmm_stripe_offset == 0)
237 v1->lmm_stripe_offset = -1;
239 /* Avoid anyone directly setting the RELEASED flag. */
240 if (v1->lmm_pattern & LOV_PATTERN_F_RELEASED) {
241 if (!release_checked) {
242 __u32 state = HS_NONE;
243 rc = get_hsm_state(inode, &state);
246 if (!(state & HS_ARCHIVED))
247 need_clear_release = true;
248 release_checked = true;
250 if (need_clear_release)
251 v1->lmm_pattern ^= LOV_PATTERN_F_RELEASED;
258 int ll_setstripe_ea(struct dentry *dentry, struct lov_user_md *lump,
261 struct inode *inode = dentry->d_inode;
264 rc = ll_adjust_lum(inode, lump);
268 if (lump != NULL && S_ISREG(inode->i_mode)) {
269 __u64 it_flags = FMODE_WRITE;
272 lum_size = ll_lov_user_md_size(lump);
273 if (lum_size < 0 || size < lum_size)
276 rc = ll_lov_setstripe_ea_info(inode, dentry, it_flags, lump,
279 * b=10667: ignore -EEXIST.
280 * Silently eat error on setting trusted.lov/lustre.lov
281 * attribute for SuSE 9, it added default option to copy
282 * all attributes in 'cp' command. rsync, tar --xattrs
283 * also will try to set LOVEA for existing files.
287 } else if (S_ISDIR(inode->i_mode)) {
288 rc = ll_dir_setstripe(inode, lump, 0);
294 int ll_setxattr(struct dentry *dentry, const char *name,
295 const void *value, size_t size, int flags)
297 struct inode *inode = dentry->d_inode;
302 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
303 PFID(ll_inode2fid(inode)), inode, name);
305 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_SETXATTR, 1);
307 /* lustre/trusted.lov.xxx would be passed through xattr API */
308 if (strcmp(name, XATTR_NAME_LOV) == 0 ||
309 strcmp(name, XATTR_LUSTRE_LOV) == 0)
310 return ll_setstripe_ea(dentry, (struct lov_user_md *)value,
312 else if (strcmp(name, XATTR_NAME_LMA) == 0 ||
313 strcmp(name, XATTR_NAME_LINK) == 0)
316 return ll_setxattr_common(inode, name, value, size, flags,
320 int ll_removexattr(struct dentry *dentry, const char *name)
322 struct inode *inode = dentry->d_inode;
327 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
328 PFID(ll_inode2fid(inode)), inode, name);
330 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_REMOVEXATTR, 1);
331 return ll_setxattr_common(inode, name, NULL, 0, 0,
335 int ll_getxattr_common(struct inode *inode, const char *name,
336 void *buffer, size_t size, __u64 valid)
338 struct ll_sb_info *sbi = ll_i2sbi(inode);
339 struct ptlrpc_request *req = NULL;
340 struct mdt_body *body;
343 struct ll_inode_info *lli = ll_i2info(inode);
346 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n",
347 PFID(ll_inode2fid(inode)), inode);
349 /* listxattr have slightly different behavior from of ext3:
350 * without 'user_xattr' ext3 will list all xattr names but
351 * filtered out "^user..*"; we list them all for simplicity.
354 xattr_type = XATTR_OTHER_T;
358 xattr_type = get_xattr_type(name);
359 rc = xattr_type_filter(sbi, xattr_type);
363 /* LU-549: Disable security.selinux when selinux is disabled */
364 if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
365 strcmp(name, "security.selinux") == 0)
368 #ifdef CONFIG_FS_POSIX_ACL
369 /* posix acl is under protection of LOOKUP lock. when calling to this,
370 * we just have path resolution to the target inode, so we have great
371 * chance that cached ACL is uptodate.
373 if (xattr_type == XATTR_ACL_ACCESS_T) {
374 struct posix_acl *acl;
376 spin_lock(&lli->lli_lock);
377 acl = posix_acl_dup(lli->lli_posix_acl);
378 spin_unlock(&lli->lli_lock);
383 rc = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
384 posix_acl_release(acl);
387 if (xattr_type == XATTR_ACL_DEFAULT_T && !S_ISDIR(inode->i_mode))
392 if (sbi->ll_xattr_cache_enabled &&
393 xattr_type != XATTR_ACL_ACCESS_T &&
394 (xattr_type != XATTR_SECURITY_T ||
395 strcmp(name, "security.selinux") != 0)) {
396 rc = ll_xattr_cache_get(inode, name, buffer, size, valid);
398 goto getxattr_nocache;
402 /* Add "system.posix_acl_access" to the list */
403 if (lli->lli_posix_acl != NULL && valid & OBD_MD_FLXATTRLS) {
405 rc += sizeof(XATTR_NAME_ACL_ACCESS);
406 } else if (size - rc >= sizeof(XATTR_NAME_ACL_ACCESS)) {
407 memcpy(buffer + rc, XATTR_NAME_ACL_ACCESS,
408 sizeof(XATTR_NAME_ACL_ACCESS));
409 rc += sizeof(XATTR_NAME_ACL_ACCESS);
411 GOTO(out_xattr, rc = -ERANGE);
416 rc = md_getxattr(sbi->ll_md_exp, ll_inode2fid(inode),
417 valid, name, NULL, 0, size, 0, &req);
422 body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
425 /* only detect the xattr size */
427 GOTO(out, rc = body->mbo_eadatasize);
429 if (size < body->mbo_eadatasize) {
430 CERROR("server bug: replied size %u > %u\n",
431 body->mbo_eadatasize, (int)size);
432 GOTO(out, rc = -ERANGE);
435 if (body->mbo_eadatasize == 0)
436 GOTO(out, rc = -ENODATA);
438 /* do not need swab xattr data */
439 xdata = req_capsule_server_sized_get(&req->rq_pill, &RMF_EADATA,
440 body->mbo_eadatasize);
442 GOTO(out, rc = -EFAULT);
444 memcpy(buffer, xdata, body->mbo_eadatasize);
445 rc = body->mbo_eadatasize;
451 if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
452 LCONSOLE_INFO("%s: disabling user_xattr feature because "
453 "it is not supported on the server: rc = %d\n",
454 ll_get_fsname(inode->i_sb, NULL, 0), rc);
455 sbi->ll_flags &= ~LL_SBI_USER_XATTR;
458 ptlrpc_req_finished(req);
462 static ssize_t ll_getxattr_lov(struct inode *inode, void *buf, size_t buf_size)
466 if (S_ISREG(inode->i_mode)) {
467 struct cl_object *obj = ll_i2info(inode)->lli_clob;
469 struct cl_layout cl = {
470 .cl_buf.lb_buf = buf,
471 .cl_buf.lb_len = buf_size,
478 env = cl_env_get(&refcheck);
480 RETURN(PTR_ERR(env));
482 rc = cl_object_layout_get(env, obj, &cl);
487 GOTO(out_env, rc = -ENODATA);
494 LASSERT(buf != NULL && rc <= buf_size);
496 /* Do not return layout gen for getxattr() since
497 * otherwise it would confuse tar --xattr by
498 * recognizing layout gen as stripe offset when the
499 * file is restored. See LU-2809. */
500 if (((struct lov_mds_md *)buf)->lmm_magic == LOV_MAGIC_COMP_V1)
503 ((struct lov_mds_md *)buf)->lmm_layout_gen = 0;
505 cl_env_put(env, &refcheck);
508 } else if (S_ISDIR(inode->i_mode)) {
509 struct lov_mds_md *lmm = NULL;
511 struct ptlrpc_request *req = NULL;
513 rc = ll_dir_getstripe(inode, (void **)&lmm, &lmm_size,
519 GOTO(out_req, rc = lmm_size);
521 if (buf_size < lmm_size)
522 GOTO(out_req, rc = -ERANGE);
524 memcpy(buf, lmm, lmm_size);
525 GOTO(out_req, rc = lmm_size);
528 ptlrpc_req_finished(req);
536 ssize_t ll_getxattr(struct dentry *dentry, const char *name, void *buf,
539 struct inode *inode = dentry->d_inode;
544 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
545 PFID(ll_inode2fid(inode)), inode, name);
547 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_GETXATTR, 1);
549 if (strcmp(name, XATTR_LUSTRE_LOV) == 0 ||
550 strcmp(name, XATTR_NAME_LOV) == 0)
551 return ll_getxattr_lov(inode, buf, buf_size);
553 return ll_getxattr_common(inode, name, buf, buf_size,
557 ssize_t ll_listxattr(struct dentry *dentry, char *buf, size_t buf_size)
559 struct inode *inode = dentry->d_inode;
560 struct ll_sb_info *sbi = ll_i2sbi(inode);
565 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n",
566 PFID(ll_inode2fid(inode)), inode);
568 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_LISTXATTR, 1);
570 rc = ll_getxattr_common(inode, NULL, buf, buf_size, OBD_MD_FLXATTRLS);
574 /* If we're being called to get the size of the xattr list
575 * (buf_size == 0) then just assume that a lustre.lov xattr
578 RETURN(rc + sizeof(XATTR_LUSTRE_LOV));
584 len = strnlen(xattr_name, rem - 1) + 1;
586 if (xattr_type_filter(sbi, get_xattr_type(xattr_name)) == 0) {
587 /* Skip OK xattr type, leave it in buffer. */
592 /* Move up remaining xattrs in buffer removing the
593 * xattr that is not OK. */
594 memmove(xattr_name, xattr_name + len, rem);
598 rc2 = ll_getxattr_lov(inode, NULL, 0);
605 if (buf_size < rc + sizeof(XATTR_LUSTRE_LOV))
608 memcpy(buf + rc, XATTR_LUSTRE_LOV, sizeof(XATTR_LUSTRE_LOV));
610 RETURN(rc + sizeof(XATTR_LUSTRE_LOV));