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.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
38 #include <linux/sched.h>
40 #include <linux/selinux.h>
42 #define DEBUG_SUBSYSTEM S_LLITE
44 #include <obd_support.h>
45 #include <lustre_lite.h>
46 #include <lustre_dlm.h>
47 #include <lustre_ver.h>
48 #include <lustre_eacl.h>
50 #include "llite_internal.h"
52 #define XATTR_USER_T (1)
53 #define XATTR_TRUSTED_T (2)
54 #define XATTR_SECURITY_T (3)
55 #define XATTR_ACL_ACCESS_T (4)
56 #define XATTR_ACL_DEFAULT_T (5)
57 #define XATTR_LUSTRE_T (6)
58 #define XATTR_OTHER_T (7)
61 int get_xattr_type(const char *name)
63 if (!strcmp(name, POSIX_ACL_XATTR_ACCESS))
64 return XATTR_ACL_ACCESS_T;
66 if (!strcmp(name, POSIX_ACL_XATTR_DEFAULT))
67 return XATTR_ACL_DEFAULT_T;
69 if (!strncmp(name, XATTR_USER_PREFIX,
70 sizeof(XATTR_USER_PREFIX) - 1))
73 if (!strncmp(name, XATTR_TRUSTED_PREFIX,
74 sizeof(XATTR_TRUSTED_PREFIX) - 1))
75 return XATTR_TRUSTED_T;
77 if (!strncmp(name, XATTR_SECURITY_PREFIX,
78 sizeof(XATTR_SECURITY_PREFIX) - 1))
79 return XATTR_SECURITY_T;
81 if (!strncmp(name, XATTR_LUSTRE_PREFIX,
82 sizeof(XATTR_LUSTRE_PREFIX) - 1))
83 return XATTR_LUSTRE_T;
89 int xattr_type_filter(struct ll_sb_info *sbi, int xattr_type)
91 if ((xattr_type == XATTR_ACL_ACCESS_T ||
92 xattr_type == XATTR_ACL_DEFAULT_T) &&
93 !(sbi->ll_flags & LL_SBI_ACL))
96 if (xattr_type == XATTR_USER_T && !(sbi->ll_flags & LL_SBI_USER_XATTR))
98 if (xattr_type == XATTR_TRUSTED_T && !cfs_capable(CFS_CAP_SYS_ADMIN))
100 if (xattr_type == XATTR_OTHER_T)
107 int ll_setxattr_common(struct inode *inode, const char *name,
108 const void *value, size_t size,
109 int flags, __u64 valid)
111 struct ll_sb_info *sbi = ll_i2sbi(inode);
112 struct ptlrpc_request *req = NULL;
115 posix_acl_xattr_header *new_value = NULL;
116 struct rmtacl_ctl_entry *rce = NULL;
117 ext_acl_xattr_header *acl = NULL;
118 const char *pv = value;
121 xattr_type = get_xattr_type(name);
122 rc = xattr_type_filter(sbi, xattr_type);
126 if ((xattr_type == XATTR_ACL_ACCESS_T ||
127 xattr_type == XATTR_ACL_DEFAULT_T) &&
128 #ifdef HAVE_INODE_OWNER_OR_CAPABLE
129 !inode_owner_or_capable(inode))
131 !is_owner_or_cap(inode))
135 /* b10667: ignore lustre special xattr for now */
136 if ((xattr_type == XATTR_TRUSTED_T && strcmp(name, "trusted.lov") == 0) ||
137 (xattr_type == XATTR_LUSTRE_T && strcmp(name, "lustre.lov") == 0))
140 /* b15587: ignore security.capability xattr for now */
141 if ((xattr_type == XATTR_SECURITY_T &&
142 strcmp(name, "security.capability") == 0))
145 /* LU-549: Disable security.selinux when selinux is disabled */
146 if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
147 strcmp(name, "security.selinux") == 0)
150 #ifdef CONFIG_FS_POSIX_ACL
151 if (sbi->ll_flags & LL_SBI_RMT_CLIENT &&
152 (xattr_type == XATTR_ACL_ACCESS_T ||
153 xattr_type == XATTR_ACL_DEFAULT_T)) {
154 rce = rct_search(&sbi->ll_rct, current_pid());
156 (rce->rce_ops != RMT_LSETFACL &&
157 rce->rce_ops != RMT_RSETFACL))
160 if (rce->rce_ops == RMT_LSETFACL) {
161 struct eacl_entry *ee;
163 ee = et_search_del(&sbi->ll_et, current_pid(),
164 ll_inode2fid(inode), xattr_type);
166 if (valid & OBD_MD_FLXATTR) {
167 acl = lustre_acl_xattr_merge2ext(
168 (posix_acl_xattr_header *)value,
172 RETURN(PTR_ERR(acl));
174 size = CFS_ACL_XATTR_SIZE(\
175 le32_to_cpu(acl->a_count), \
177 pv = (const char *)acl;
180 } else if (rce->rce_ops == RMT_RSETFACL) {
181 int acl_size = lustre_posix_acl_xattr_filter(
182 (posix_acl_xattr_header *)value,
184 if (unlikely(acl_size < 0))
188 pv = (const char *)new_value;
192 valid |= rce_ops2valid(rce->rce_ops);
195 oc = ll_mdscapa_get(inode);
196 rc = md_setxattr(sbi->ll_md_exp, ll_inode2fid(inode), oc,
197 valid, name, pv, size, 0, flags,
198 ll_i2suppgid(inode), &req);
200 #ifdef CONFIG_FS_POSIX_ACL
201 if (new_value != NULL)
202 lustre_posix_acl_xattr_free(new_value, size);
204 lustre_ext_acl_xattr_free(acl);
207 if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
208 LCONSOLE_INFO("Disabling user_xattr feature because "
209 "it is not supported on the server\n");
210 sbi->ll_flags &= ~LL_SBI_USER_XATTR;
215 ptlrpc_req_finished(req);
219 int ll_setxattr(struct dentry *dentry, const char *name,
220 const void *value, size_t size, int flags)
222 struct inode *inode = dentry->d_inode;
227 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
228 PFID(ll_inode2fid(inode)), inode, name);
230 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_SETXATTR, 1);
232 if ((strncmp(name, XATTR_TRUSTED_PREFIX,
233 sizeof(XATTR_TRUSTED_PREFIX) - 1) == 0 &&
234 strcmp(name + sizeof(XATTR_TRUSTED_PREFIX) - 1, "lov") == 0) ||
235 (strncmp(name, XATTR_LUSTRE_PREFIX,
236 sizeof(XATTR_LUSTRE_PREFIX) - 1) == 0 &&
237 strcmp(name + sizeof(XATTR_LUSTRE_PREFIX) - 1, "lov") == 0)) {
238 struct lov_user_md *lump = (struct lov_user_md *)value;
241 /* Attributes that are saved via getxattr will always have
242 * the stripe_offset as 0. Instead, the MDS should be
243 * allowed to pick the starting OST index. b=17846 */
244 if (lump != NULL && lump->lmm_stripe_offset == 0)
245 lump->lmm_stripe_offset = -1;
247 if (lump != NULL && S_ISREG(inode->i_mode)) {
249 __u64 it_flags = FMODE_WRITE;
250 int lum_size = (lump->lmm_magic == LOV_USER_MAGIC_V1) ?
251 sizeof(*lump) : sizeof(struct lov_user_md_v3);
254 rc = ll_lov_setstripe_ea_info(inode, &f, it_flags, lump,
256 /* b10667: rc always be 0 here for now */
258 } else if (S_ISDIR(inode->i_mode)) {
259 rc = ll_dir_setstripe(inode, lump, 0);
264 } else if (strcmp(name, XATTR_NAME_LMA) == 0 ||
265 strcmp(name, XATTR_NAME_LINK) == 0)
268 return ll_setxattr_common(inode, name, value, size, flags,
272 int ll_removexattr(struct dentry *dentry, const char *name)
274 struct inode *inode = dentry->d_inode;
279 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
280 PFID(ll_inode2fid(inode)), inode, name);
282 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_REMOVEXATTR, 1);
283 return ll_setxattr_common(inode, name, NULL, 0, 0,
288 int ll_getxattr_common(struct inode *inode, const char *name,
289 void *buffer, size_t size, __u64 valid)
291 struct ll_sb_info *sbi = ll_i2sbi(inode);
292 struct ptlrpc_request *req = NULL;
293 struct mdt_body *body;
297 struct rmtacl_ctl_entry *rce = NULL;
298 struct ll_inode_info *lli = ll_i2info(inode);
301 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n",
302 PFID(ll_inode2fid(inode)), inode);
304 /* listxattr have slightly different behavior from of ext3:
305 * without 'user_xattr' ext3 will list all xattr names but
306 * filtered out "^user..*"; we list them all for simplicity.
309 xattr_type = XATTR_OTHER_T;
313 xattr_type = get_xattr_type(name);
314 rc = xattr_type_filter(sbi, xattr_type);
318 /* b15587: ignore security.capability xattr for now */
319 if ((xattr_type == XATTR_SECURITY_T &&
320 strcmp(name, "security.capability") == 0))
323 /* LU-549: Disable security.selinux when selinux is disabled */
324 if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
325 strcmp(name, "security.selinux") == 0)
328 #ifdef CONFIG_FS_POSIX_ACL
329 if (sbi->ll_flags & LL_SBI_RMT_CLIENT &&
330 (xattr_type == XATTR_ACL_ACCESS_T ||
331 xattr_type == XATTR_ACL_DEFAULT_T)) {
332 rce = rct_search(&sbi->ll_rct, current_pid());
334 (rce->rce_ops != RMT_LSETFACL &&
335 rce->rce_ops != RMT_LGETFACL &&
336 rce->rce_ops != RMT_RSETFACL &&
337 rce->rce_ops != RMT_RGETFACL))
341 /* posix acl is under protection of LOOKUP lock. when calling to this,
342 * we just have path resolution to the target inode, so we have great
343 * chance that cached ACL is uptodate.
345 if (xattr_type == XATTR_ACL_ACCESS_T &&
346 !(sbi->ll_flags & LL_SBI_RMT_CLIENT)) {
348 struct posix_acl *acl;
350 spin_lock(&lli->lli_lock);
351 acl = posix_acl_dup(lli->lli_posix_acl);
352 spin_unlock(&lli->lli_lock);
357 rc = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
358 posix_acl_release(acl);
361 if (xattr_type == XATTR_ACL_DEFAULT_T && !S_ISDIR(inode->i_mode))
366 if (sbi->ll_xattr_cache_enabled && xattr_type != XATTR_ACL_ACCESS_T) {
367 rc = ll_xattr_cache_get(inode, name, buffer, size, valid);
369 goto getxattr_nocache;
373 /* Add "system.posix_acl_access" to the list */
374 if (lli->lli_posix_acl != NULL && valid & OBD_MD_FLXATTRLS) {
376 rc += sizeof(XATTR_NAME_ACL_ACCESS);
377 } else if (size - rc >= sizeof(XATTR_NAME_ACL_ACCESS)) {
378 memcpy(buffer + rc, XATTR_NAME_ACL_ACCESS,
379 sizeof(XATTR_NAME_ACL_ACCESS));
380 rc += sizeof(XATTR_NAME_ACL_ACCESS);
382 GOTO(out_xattr, rc = -ERANGE);
387 oc = ll_mdscapa_get(inode);
388 rc = md_getxattr(sbi->ll_md_exp, ll_inode2fid(inode), oc,
389 valid | (rce ? rce_ops2valid(rce->rce_ops) : 0),
390 name, NULL, 0, size, 0, &req);
396 body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
399 /* only detect the xattr size */
401 GOTO(out, rc = body->eadatasize);
403 if (size < body->eadatasize) {
404 CERROR("server bug: replied size %u > %u\n",
405 body->eadatasize, (int)size);
406 GOTO(out, rc = -ERANGE);
409 if (body->eadatasize == 0)
410 GOTO(out, rc = -ENODATA);
412 /* do not need swab xattr data */
413 xdata = req_capsule_server_sized_get(&req->rq_pill, &RMF_EADATA,
416 GOTO(out, rc = -EFAULT);
418 memcpy(buffer, xdata, body->eadatasize);
419 rc = body->eadatasize;
422 #ifdef CONFIG_FS_POSIX_ACL
423 if (rce != NULL && rce->rce_ops == RMT_LSETFACL) {
424 ext_acl_xattr_header *acl;
426 acl = lustre_posix_acl_xattr_2ext(buffer, rc);
428 GOTO(out, rc = PTR_ERR(acl));
430 rc = ee_add(&sbi->ll_et, current_pid(), ll_inode2fid(inode),
432 if (unlikely(rc < 0)) {
433 lustre_ext_acl_xattr_free(acl);
441 if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
442 LCONSOLE_INFO("%s: disabling user_xattr feature because "
443 "it is not supported on the server: rc = %d\n",
444 ll_get_fsname(inode->i_sb, NULL, 0), rc);
445 sbi->ll_flags &= ~LL_SBI_USER_XATTR;
448 ptlrpc_req_finished(req);
452 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
453 void *buffer, size_t size)
455 struct inode *inode = dentry->d_inode;
460 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), xattr %s\n",
461 PFID(ll_inode2fid(inode)), inode, name);
463 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_GETXATTR, 1);
465 if ((strncmp(name, XATTR_TRUSTED_PREFIX,
466 sizeof(XATTR_TRUSTED_PREFIX) - 1) == 0 &&
467 strcmp(name + sizeof(XATTR_TRUSTED_PREFIX) - 1, "lov") == 0) ||
468 (strncmp(name, XATTR_LUSTRE_PREFIX,
469 sizeof(XATTR_LUSTRE_PREFIX) - 1) == 0 &&
470 strcmp(name + sizeof(XATTR_LUSTRE_PREFIX) - 1, "lov") == 0)) {
471 struct lov_stripe_md *lsm;
472 struct lov_user_md *lump;
473 struct lov_mds_md *lmm = NULL;
474 struct ptlrpc_request *request = NULL;
475 int rc = 0, lmmsize = 0;
477 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
480 if (size == 0 && S_ISDIR(inode->i_mode)) {
481 /* XXX directory EA is fix for now, optimize to save
483 GOTO(out, rc = sizeof(struct lov_user_md));
486 lsm = ccc_inode_lsm_get(inode);
488 if (S_ISDIR(inode->i_mode)) {
489 rc = ll_dir_getstripe(inode, (void **)&lmm,
490 &lmmsize, &request, 0);
495 /* LSM is present already after lookup/getattr call.
496 * we need to grab layout lock once it is implemented */
497 rc = obd_packmd(ll_i2dtexp(inode), &lmm, lsm);
500 ccc_inode_lsm_put(inode, lsm);
506 /* used to call ll_get_max_mdsize() forward to get
507 * the maximum buffer size, while some apps (such as
508 * rsync 3.0.x) care much about the exact xattr value
514 if (size < lmmsize) {
515 CERROR("server bug: replied size %d > %d for %s (%s)\n",
516 lmmsize, (int)size, dentry->d_name.name, name);
517 GOTO(out, rc = -ERANGE);
520 lump = (struct lov_user_md *)buffer;
521 memcpy(lump, lmm, lmmsize);
522 /* do not return layout gen for getxattr otherwise it would
523 * confuse tar --xattr by recognizing layout gen as stripe
524 * offset when the file is restored. See LU-2809. */
525 lump->lmm_layout_gen = 0;
530 ptlrpc_req_finished(request);
532 obd_free_diskmd(ll_i2dtexp(inode), &lmm);
536 return ll_getxattr_common(inode, name, buffer, size, OBD_MD_FLXATTR);
539 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size)
541 struct inode *inode = dentry->d_inode;
543 struct lov_mds_md *lmm = NULL;
544 struct ptlrpc_request *request = NULL;
549 CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n",
550 PFID(ll_inode2fid(inode)), inode);
552 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_LISTXATTR, 1);
554 rc = ll_getxattr_common(inode, NULL, buffer, size, OBD_MD_FLXATTRLS);
558 if (buffer != NULL) {
559 struct ll_sb_info *sbi = ll_i2sbi(inode);
560 char *xattr_name = buffer;
564 xlen = strnlen(xattr_name, rem - 1) + 1;
566 if (xattr_type_filter(sbi,
567 get_xattr_type(xattr_name)) == 0) {
568 /* skip OK xattr type
574 /* move up remaining xattrs in buffer
575 * removing the xattr that is not OK
577 memmove(xattr_name, xattr_name + xlen, rem);
581 if (S_ISREG(inode->i_mode)) {
582 if (!ll_i2info(inode)->lli_has_smd)
584 } else if (S_ISDIR(inode->i_mode)) {
585 rc2 = ll_dir_getstripe(inode, (void **)&lmm, &lmmsize, &request,
591 } else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) {
592 const int prefix_len = sizeof(XATTR_LUSTRE_PREFIX) - 1;
593 const size_t name_len = sizeof("lov") - 1;
594 const size_t total_len = prefix_len + name_len + 1;
596 if (((rc + total_len) > size) && (buffer != NULL)) {
597 ptlrpc_req_finished(request);
601 if (buffer != NULL) {
603 memcpy(buffer, XATTR_LUSTRE_PREFIX, prefix_len);
604 memcpy(buffer + prefix_len, "lov", name_len);
605 buffer[prefix_len + name_len] = '\0';
610 ptlrpc_req_finished(request);