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, 2012, 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 "../include/obd_support.h"
45 #include "../include/lustre_lite.h"
46 #include "../include/lustre_dlm.h"
47 #include "../include/lustre_ver.h"
48 #include "../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 && !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 #ifdef CONFIG_FS_POSIX_ACL
116 struct rmtacl_ctl_entry *rce = NULL;
117 posix_acl_xattr_header *new_value = NULL;
118 ext_acl_xattr_header *acl = NULL;
120 const char *pv = value;
122 xattr_type = get_xattr_type(name);
123 rc = xattr_type_filter(sbi, xattr_type);
127 if ((xattr_type == XATTR_ACL_ACCESS_T ||
128 xattr_type == XATTR_ACL_DEFAULT_T) &&
129 !inode_owner_or_capable(inode))
132 /* b10667: ignore lustre special xattr for now */
133 if ((xattr_type == XATTR_TRUSTED_T && strcmp(name, "trusted.lov") == 0) ||
134 (xattr_type == XATTR_LUSTRE_T && strcmp(name, "lustre.lov") == 0))
137 /* b15587: ignore security.capability xattr for now */
138 if ((xattr_type == XATTR_SECURITY_T &&
139 strcmp(name, "security.capability") == 0))
142 /* LU-549: Disable security.selinux when selinux is disabled */
143 if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
144 strcmp(name, "security.selinux") == 0)
147 #ifdef CONFIG_FS_POSIX_ACL
148 if (sbi->ll_flags & LL_SBI_RMT_CLIENT &&
149 (xattr_type == XATTR_ACL_ACCESS_T ||
150 xattr_type == XATTR_ACL_DEFAULT_T)) {
151 rce = rct_search(&sbi->ll_rct, current_pid());
153 (rce->rce_ops != RMT_LSETFACL &&
154 rce->rce_ops != RMT_RSETFACL))
157 if (rce->rce_ops == RMT_LSETFACL) {
158 struct eacl_entry *ee;
160 ee = et_search_del(&sbi->ll_et, current_pid(),
161 ll_inode2fid(inode), xattr_type);
163 if (valid & OBD_MD_FLXATTR) {
164 acl = lustre_acl_xattr_merge2ext(
165 (posix_acl_xattr_header *)value,
171 size = CFS_ACL_XATTR_SIZE(\
172 le32_to_cpu(acl->a_count), \
174 pv = (const char *)acl;
177 } else if (rce->rce_ops == RMT_RSETFACL) {
178 size = lustre_posix_acl_xattr_filter(
179 (posix_acl_xattr_header *)value,
181 if (unlikely(size < 0))
184 pv = (const char *)new_value;
188 valid |= rce_ops2valid(rce->rce_ops);
191 oc = ll_mdscapa_get(inode);
192 rc = md_setxattr(sbi->ll_md_exp, ll_inode2fid(inode), oc,
193 valid, name, pv, size, 0, flags,
194 ll_i2suppgid(inode), &req);
196 #ifdef CONFIG_FS_POSIX_ACL
197 if (new_value != NULL)
198 lustre_posix_acl_xattr_free(new_value, size);
200 lustre_ext_acl_xattr_free(acl);
203 if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
204 LCONSOLE_INFO("Disabling user_xattr feature because it is not supported on the server\n");
205 sbi->ll_flags &= ~LL_SBI_USER_XATTR;
210 ptlrpc_req_finished(req);
214 int ll_setxattr(struct dentry *dentry, const char *name,
215 const void *value, size_t size, int flags)
217 struct inode *inode = d_inode(dentry);
222 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), xattr %s\n",
223 inode->i_ino, inode->i_generation, inode, name);
225 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_SETXATTR, 1);
227 if ((strncmp(name, XATTR_TRUSTED_PREFIX,
228 sizeof(XATTR_TRUSTED_PREFIX) - 1) == 0 &&
229 strcmp(name + sizeof(XATTR_TRUSTED_PREFIX) - 1, "lov") == 0) ||
230 (strncmp(name, XATTR_LUSTRE_PREFIX,
231 sizeof(XATTR_LUSTRE_PREFIX) - 1) == 0 &&
232 strcmp(name + sizeof(XATTR_LUSTRE_PREFIX) - 1, "lov") == 0)) {
233 struct lov_user_md *lump = (struct lov_user_md *)value;
236 if (size != 0 && size < sizeof(struct lov_user_md))
239 /* Attributes that are saved via getxattr will always have
240 * the stripe_offset as 0. Instead, the MDS should be
241 * allowed to pick the starting OST index. b=17846 */
242 if (lump != NULL && lump->lmm_stripe_offset == 0)
243 lump->lmm_stripe_offset = -1;
245 if (lump != NULL && S_ISREG(inode->i_mode)) {
246 int flags = FMODE_WRITE;
247 int lum_size = (lump->lmm_magic == LOV_USER_MAGIC_V1) ?
248 sizeof(*lump) : sizeof(struct lov_user_md_v3);
250 rc = ll_lov_setstripe_ea_info(inode, dentry, flags, lump,
252 /* b10667: rc always be 0 here for now */
254 } else if (S_ISDIR(inode->i_mode)) {
255 rc = ll_dir_setstripe(inode, lump, 0);
260 } else if (strcmp(name, XATTR_NAME_LMA) == 0 ||
261 strcmp(name, XATTR_NAME_LINK) == 0)
264 return ll_setxattr_common(inode, name, value, size, flags,
268 int ll_removexattr(struct dentry *dentry, const char *name)
270 struct inode *inode = d_inode(dentry);
275 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), xattr %s\n",
276 inode->i_ino, inode->i_generation, inode, name);
278 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_REMOVEXATTR, 1);
279 return ll_setxattr_common(inode, name, NULL, 0, 0,
284 int ll_getxattr_common(struct inode *inode, const char *name,
285 void *buffer, size_t size, __u64 valid)
287 struct ll_sb_info *sbi = ll_i2sbi(inode);
288 struct ptlrpc_request *req = NULL;
289 struct mdt_body *body;
293 struct rmtacl_ctl_entry *rce = NULL;
294 struct ll_inode_info *lli = ll_i2info(inode);
296 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n",
297 inode->i_ino, inode->i_generation, inode);
299 /* listxattr have slightly different behavior from of ext3:
300 * without 'user_xattr' ext3 will list all xattr names but
301 * filtered out "^user..*"; we list them all for simplicity.
304 xattr_type = XATTR_OTHER_T;
308 xattr_type = get_xattr_type(name);
309 rc = xattr_type_filter(sbi, xattr_type);
313 /* b15587: ignore security.capability xattr for now */
314 if ((xattr_type == XATTR_SECURITY_T &&
315 strcmp(name, "security.capability") == 0))
318 /* LU-549: Disable security.selinux when selinux is disabled */
319 if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
320 strcmp(name, "security.selinux") == 0)
323 #ifdef CONFIG_FS_POSIX_ACL
324 if (sbi->ll_flags & LL_SBI_RMT_CLIENT &&
325 (xattr_type == XATTR_ACL_ACCESS_T ||
326 xattr_type == XATTR_ACL_DEFAULT_T)) {
327 rce = rct_search(&sbi->ll_rct, current_pid());
329 (rce->rce_ops != RMT_LSETFACL &&
330 rce->rce_ops != RMT_LGETFACL &&
331 rce->rce_ops != RMT_RSETFACL &&
332 rce->rce_ops != RMT_RGETFACL))
336 /* posix acl is under protection of LOOKUP lock. when calling to this,
337 * we just have path resolution to the target inode, so we have great
338 * chance that cached ACL is uptodate.
340 if (xattr_type == XATTR_ACL_ACCESS_T &&
341 !(sbi->ll_flags & LL_SBI_RMT_CLIENT)) {
343 struct posix_acl *acl;
345 spin_lock(&lli->lli_lock);
346 acl = posix_acl_dup(lli->lli_posix_acl);
347 spin_unlock(&lli->lli_lock);
352 rc = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
353 posix_acl_release(acl);
356 if (xattr_type == XATTR_ACL_DEFAULT_T && !S_ISDIR(inode->i_mode))
361 if (sbi->ll_xattr_cache_enabled && xattr_type != XATTR_ACL_ACCESS_T) {
362 rc = ll_xattr_cache_get(inode, name, buffer, size, valid);
364 goto getxattr_nocache;
368 /* Add "system.posix_acl_access" to the list */
369 if (lli->lli_posix_acl != NULL && valid & OBD_MD_FLXATTRLS) {
371 rc += sizeof(XATTR_NAME_ACL_ACCESS);
372 } else if (size - rc >= sizeof(XATTR_NAME_ACL_ACCESS)) {
373 memcpy(buffer + rc, XATTR_NAME_ACL_ACCESS,
374 sizeof(XATTR_NAME_ACL_ACCESS));
375 rc += sizeof(XATTR_NAME_ACL_ACCESS);
383 oc = ll_mdscapa_get(inode);
384 rc = md_getxattr(sbi->ll_md_exp, ll_inode2fid(inode), oc,
385 valid | (rce ? rce_ops2valid(rce->rce_ops) : 0),
386 name, NULL, 0, size, 0, &req);
392 body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
395 /* only detect the xattr size */
397 rc = body->eadatasize;
401 if (size < body->eadatasize) {
402 CERROR("server bug: replied size %u > %u\n",
403 body->eadatasize, (int)size);
408 if (body->eadatasize == 0) {
413 /* do not need swab xattr data */
414 xdata = req_capsule_server_sized_get(&req->rq_pill, &RMF_EADATA,
421 memcpy(buffer, xdata, body->eadatasize);
422 rc = body->eadatasize;
425 #ifdef CONFIG_FS_POSIX_ACL
426 if (rce && rce->rce_ops == RMT_LSETFACL) {
427 ext_acl_xattr_header *acl;
429 acl = lustre_posix_acl_xattr_2ext(
430 (posix_acl_xattr_header *)buffer, rc);
436 rc = ee_add(&sbi->ll_et, current_pid(), ll_inode2fid(inode),
438 if (unlikely(rc < 0)) {
439 lustre_ext_acl_xattr_free(acl);
446 if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
448 "%s: disabling user_xattr feature because it is not supported on the server: rc = %d\n",
449 ll_get_fsname(inode->i_sb, NULL, 0), rc);
450 sbi->ll_flags &= ~LL_SBI_USER_XATTR;
453 ptlrpc_req_finished(req);
457 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
458 void *buffer, size_t size)
460 struct inode *inode = d_inode(dentry);
465 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), xattr %s\n",
466 inode->i_ino, inode->i_generation, inode, name);
468 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_GETXATTR, 1);
470 if ((strncmp(name, XATTR_TRUSTED_PREFIX,
471 sizeof(XATTR_TRUSTED_PREFIX) - 1) == 0 &&
472 strcmp(name + sizeof(XATTR_TRUSTED_PREFIX) - 1, "lov") == 0) ||
473 (strncmp(name, XATTR_LUSTRE_PREFIX,
474 sizeof(XATTR_LUSTRE_PREFIX) - 1) == 0 &&
475 strcmp(name + sizeof(XATTR_LUSTRE_PREFIX) - 1, "lov") == 0)) {
476 struct lov_stripe_md *lsm;
477 struct lov_user_md *lump;
478 struct lov_mds_md *lmm = NULL;
479 struct ptlrpc_request *request = NULL;
480 int rc = 0, lmmsize = 0;
482 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
485 if (size == 0 && S_ISDIR(inode->i_mode)) {
486 /* XXX directory EA is fix for now, optimize to save
488 rc = sizeof(struct lov_user_md);
492 lsm = ccc_inode_lsm_get(inode);
494 if (S_ISDIR(inode->i_mode)) {
495 rc = ll_dir_getstripe(inode, &lmm,
501 /* LSM is present already after lookup/getattr call.
502 * we need to grab layout lock once it is implemented */
503 rc = obd_packmd(ll_i2dtexp(inode), &lmm, lsm);
506 ccc_inode_lsm_put(inode, lsm);
512 /* used to call ll_get_max_mdsize() forward to get
513 * the maximum buffer size, while some apps (such as
514 * rsync 3.0.x) care much about the exact xattr value
520 if (size < lmmsize) {
521 CERROR("server bug: replied size %d > %d for %pd (%s)\n",
522 lmmsize, (int)size, dentry, name);
527 lump = (struct lov_user_md *)buffer;
528 memcpy(lump, lmm, lmmsize);
529 /* do not return layout gen for getxattr otherwise it would
530 * confuse tar --xattr by recognizing layout gen as stripe
531 * offset when the file is restored. See LU-2809. */
532 lump->lmm_layout_gen = 0;
537 ptlrpc_req_finished(request);
539 obd_free_diskmd(ll_i2dtexp(inode), &lmm);
543 return ll_getxattr_common(inode, name, buffer, size, OBD_MD_FLXATTR);
546 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size)
548 struct inode *inode = d_inode(dentry);
550 struct lov_mds_md *lmm = NULL;
551 struct ptlrpc_request *request = NULL;
556 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n",
557 inode->i_ino, inode->i_generation, inode);
559 ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_LISTXATTR, 1);
561 rc = ll_getxattr_common(inode, NULL, buffer, size, OBD_MD_FLXATTRLS);
565 if (buffer != NULL) {
566 struct ll_sb_info *sbi = ll_i2sbi(inode);
567 char *xattr_name = buffer;
571 xlen = strnlen(xattr_name, rem - 1) + 1;
573 if (xattr_type_filter(sbi,
574 get_xattr_type(xattr_name)) == 0) {
575 /* skip OK xattr type
581 /* move up remaining xattrs in buffer
582 * removing the xattr that is not OK
584 memmove(xattr_name, xattr_name + xlen, rem);
588 if (S_ISREG(inode->i_mode)) {
589 if (!ll_i2info(inode)->lli_has_smd)
591 } else if (S_ISDIR(inode->i_mode)) {
592 rc2 = ll_dir_getstripe(inode, &lmm, &lmmsize, &request);
598 } else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) {
599 const int prefix_len = sizeof(XATTR_LUSTRE_PREFIX) - 1;
600 const size_t name_len = sizeof("lov") - 1;
601 const size_t total_len = prefix_len + name_len + 1;
603 if (((rc + total_len) > size) && (buffer != NULL)) {
604 ptlrpc_req_finished(request);
608 if (buffer != NULL) {
610 memcpy(buffer, XATTR_LUSTRE_PREFIX, prefix_len);
611 memcpy(buffer + prefix_len, "lov", name_len);
612 buffer[prefix_len + name_len] = '\0';
617 ptlrpc_req_finished(request);