1 #include <linux/ceph/ceph_debug.h>
2 #include <linux/ceph/pagelist.h>
5 #include "mds_client.h"
7 #include <linux/ceph/decode.h>
9 #include <linux/xattr.h>
10 #include <linux/posix_acl_xattr.h>
11 #include <linux/slab.h>
13 #define XATTR_CEPH_PREFIX "ceph."
14 #define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
16 static int __remove_xattr(struct ceph_inode_info *ci,
17 struct ceph_inode_xattr *xattr);
19 static const struct xattr_handler ceph_other_xattr_handler;
22 * List of handlers for synthetic system.* attributes. Other
23 * attributes are handled directly.
25 const struct xattr_handler *ceph_xattr_handlers[] = {
26 #ifdef CONFIG_CEPH_FS_POSIX_ACL
27 &posix_acl_access_xattr_handler,
28 &posix_acl_default_xattr_handler,
30 &ceph_other_xattr_handler,
34 static bool ceph_is_valid_xattr(const char *name)
36 return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
37 !strncmp(name, XATTR_SECURITY_PREFIX,
38 XATTR_SECURITY_PREFIX_LEN) ||
39 !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
40 !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
44 * These define virtual xattrs exposing the recursive directory
45 * statistics and layout metadata.
49 size_t name_size; /* strlen(name) + 1 (for '\0') */
50 size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
52 bool readonly, hidden;
53 bool (*exists_cb)(struct ceph_inode_info *ci);
58 static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
60 struct ceph_file_layout *fl = &ci->i_layout;
61 return (fl->stripe_unit > 0 || fl->stripe_count > 0 ||
62 fl->object_size > 0 || fl->pool_id >= 0 ||
63 rcu_dereference_raw(fl->pool_ns) != NULL);
66 static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
69 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
70 struct ceph_osd_client *osdc = &fsc->client->osdc;
71 struct ceph_string *pool_ns;
72 s64 pool = ci->i_layout.pool_id;
73 const char *pool_name;
74 const char *ns_field = " pool_namespace=";
76 size_t len, total_len = 0;
79 pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
81 dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
82 down_read(&osdc->lock);
83 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
85 len = snprintf(buf, sizeof(buf),
86 "stripe_unit=%u stripe_count=%u object_size=%u pool=",
87 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
88 ci->i_layout.object_size);
89 total_len = len + strlen(pool_name);
91 len = snprintf(buf, sizeof(buf),
92 "stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
93 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
94 ci->i_layout.object_size, (unsigned long long)pool);
99 total_len += strlen(ns_field) + pool_ns->len;
103 } else if (total_len > size) {
106 memcpy(val, buf, len);
109 len = strlen(pool_name);
110 memcpy(val + ret, pool_name, len);
114 len = strlen(ns_field);
115 memcpy(val + ret, ns_field, len);
117 memcpy(val + ret, pool_ns->str, pool_ns->len);
121 up_read(&osdc->lock);
122 ceph_put_string(pool_ns);
126 static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
127 char *val, size_t size)
129 return snprintf(val, size, "%u", ci->i_layout.stripe_unit);
132 static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
133 char *val, size_t size)
135 return snprintf(val, size, "%u", ci->i_layout.stripe_count);
138 static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
139 char *val, size_t size)
141 return snprintf(val, size, "%u", ci->i_layout.object_size);
144 static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
145 char *val, size_t size)
148 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
149 struct ceph_osd_client *osdc = &fsc->client->osdc;
150 s64 pool = ci->i_layout.pool_id;
151 const char *pool_name;
153 down_read(&osdc->lock);
154 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
156 ret = snprintf(val, size, "%s", pool_name);
158 ret = snprintf(val, size, "%lld", (unsigned long long)pool);
159 up_read(&osdc->lock);
163 static size_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info *ci,
164 char *val, size_t size)
167 struct ceph_string *ns = ceph_try_get_string(ci->i_layout.pool_ns);
169 ret = snprintf(val, size, "%.*s", (int)ns->len, ns->str);
177 static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
180 return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
183 static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
186 return snprintf(val, size, "%lld", ci->i_files);
189 static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
192 return snprintf(val, size, "%lld", ci->i_subdirs);
195 static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
198 return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
201 static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
204 return snprintf(val, size, "%lld", ci->i_rfiles);
207 static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
210 return snprintf(val, size, "%lld", ci->i_rsubdirs);
213 static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
216 return snprintf(val, size, "%lld", ci->i_rbytes);
219 static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
222 return snprintf(val, size, "%ld.09%ld", (long)ci->i_rctime.tv_sec,
223 (long)ci->i_rctime.tv_nsec);
227 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
228 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
229 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
231 #define XATTR_NAME_CEPH(_type, _name) \
233 .name = CEPH_XATTR_NAME(_type, _name), \
234 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
235 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
240 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
242 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
243 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
244 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
247 .exists_cb = ceph_vxattrcb_layout_exists, \
250 static struct ceph_vxattr ceph_dir_vxattrs[] = {
252 .name = "ceph.dir.layout",
253 .name_size = sizeof("ceph.dir.layout"),
254 .getxattr_cb = ceph_vxattrcb_layout,
257 .exists_cb = ceph_vxattrcb_layout_exists,
259 XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
260 XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
261 XATTR_LAYOUT_FIELD(dir, layout, object_size),
262 XATTR_LAYOUT_FIELD(dir, layout, pool),
263 XATTR_LAYOUT_FIELD(dir, layout, pool_namespace),
264 XATTR_NAME_CEPH(dir, entries),
265 XATTR_NAME_CEPH(dir, files),
266 XATTR_NAME_CEPH(dir, subdirs),
267 XATTR_NAME_CEPH(dir, rentries),
268 XATTR_NAME_CEPH(dir, rfiles),
269 XATTR_NAME_CEPH(dir, rsubdirs),
270 XATTR_NAME_CEPH(dir, rbytes),
271 XATTR_NAME_CEPH(dir, rctime),
272 { .name = NULL, 0 } /* Required table terminator */
274 static size_t ceph_dir_vxattrs_name_size; /* total size of all names */
278 static struct ceph_vxattr ceph_file_vxattrs[] = {
280 .name = "ceph.file.layout",
281 .name_size = sizeof("ceph.file.layout"),
282 .getxattr_cb = ceph_vxattrcb_layout,
285 .exists_cb = ceph_vxattrcb_layout_exists,
287 XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
288 XATTR_LAYOUT_FIELD(file, layout, stripe_count),
289 XATTR_LAYOUT_FIELD(file, layout, object_size),
290 XATTR_LAYOUT_FIELD(file, layout, pool),
291 XATTR_LAYOUT_FIELD(file, layout, pool_namespace),
292 { .name = NULL, 0 } /* Required table terminator */
294 static size_t ceph_file_vxattrs_name_size; /* total size of all names */
296 static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
298 if (S_ISDIR(inode->i_mode))
299 return ceph_dir_vxattrs;
300 else if (S_ISREG(inode->i_mode))
301 return ceph_file_vxattrs;
305 static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
307 if (vxattrs == ceph_dir_vxattrs)
308 return ceph_dir_vxattrs_name_size;
309 if (vxattrs == ceph_file_vxattrs)
310 return ceph_file_vxattrs_name_size;
316 * Compute the aggregate size (including terminating '\0') of all
317 * virtual extended attribute names in the given vxattr table.
319 static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
321 struct ceph_vxattr *vxattr;
324 for (vxattr = vxattrs; vxattr->name; vxattr++)
326 size += vxattr->name_size;
331 /* Routines called at initialization and exit time */
333 void __init ceph_xattr_init(void)
335 ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
336 ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
339 void ceph_xattr_exit(void)
341 ceph_dir_vxattrs_name_size = 0;
342 ceph_file_vxattrs_name_size = 0;
345 static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
348 struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
351 while (vxattr->name) {
352 if (!strcmp(vxattr->name, name))
361 static int __set_xattr(struct ceph_inode_info *ci,
362 const char *name, int name_len,
363 const char *val, int val_len,
364 int flags, int update_xattr,
365 struct ceph_inode_xattr **newxattr)
368 struct rb_node *parent = NULL;
369 struct ceph_inode_xattr *xattr = NULL;
373 p = &ci->i_xattrs.index.rb_node;
376 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
377 c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
383 if (name_len == xattr->name_len)
385 else if (name_len < xattr->name_len)
396 if (xattr && (flags & XATTR_CREATE))
398 else if (!xattr && (flags & XATTR_REPLACE))
406 if (update_xattr < 0) {
408 __remove_xattr(ci, xattr);
419 xattr->name_len = name_len;
420 xattr->should_free_name = update_xattr;
422 ci->i_xattrs.count++;
423 dout("__set_xattr count=%d\n", ci->i_xattrs.count);
427 if (xattr->should_free_val)
428 kfree((void *)xattr->val);
434 ci->i_xattrs.names_size -= xattr->name_len;
435 ci->i_xattrs.vals_size -= xattr->val_len;
437 ci->i_xattrs.names_size += name_len;
438 ci->i_xattrs.vals_size += val_len;
444 xattr->val_len = val_len;
445 xattr->dirty = update_xattr;
446 xattr->should_free_val = (val && update_xattr);
449 rb_link_node(&xattr->node, parent, p);
450 rb_insert_color(&xattr->node, &ci->i_xattrs.index);
451 dout("__set_xattr_val p=%p\n", p);
454 dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
455 ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);
460 static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
464 struct rb_node *parent = NULL;
465 struct ceph_inode_xattr *xattr = NULL;
466 int name_len = strlen(name);
469 p = &ci->i_xattrs.index.rb_node;
472 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
473 c = strncmp(name, xattr->name, xattr->name_len);
474 if (c == 0 && name_len > xattr->name_len)
481 dout("__get_xattr %s: found %.*s\n", name,
482 xattr->val_len, xattr->val);
487 dout("__get_xattr %s: not found\n", name);
492 static void __free_xattr(struct ceph_inode_xattr *xattr)
496 if (xattr->should_free_name)
497 kfree((void *)xattr->name);
498 if (xattr->should_free_val)
499 kfree((void *)xattr->val);
504 static int __remove_xattr(struct ceph_inode_info *ci,
505 struct ceph_inode_xattr *xattr)
510 rb_erase(&xattr->node, &ci->i_xattrs.index);
512 if (xattr->should_free_name)
513 kfree((void *)xattr->name);
514 if (xattr->should_free_val)
515 kfree((void *)xattr->val);
517 ci->i_xattrs.names_size -= xattr->name_len;
518 ci->i_xattrs.vals_size -= xattr->val_len;
519 ci->i_xattrs.count--;
525 static char *__copy_xattr_names(struct ceph_inode_info *ci,
529 struct ceph_inode_xattr *xattr = NULL;
531 p = rb_first(&ci->i_xattrs.index);
532 dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
535 xattr = rb_entry(p, struct ceph_inode_xattr, node);
536 memcpy(dest, xattr->name, xattr->name_len);
537 dest[xattr->name_len] = '\0';
539 dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
540 xattr->name_len, ci->i_xattrs.names_size);
542 dest += xattr->name_len + 1;
549 void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
551 struct rb_node *p, *tmp;
552 struct ceph_inode_xattr *xattr = NULL;
554 p = rb_first(&ci->i_xattrs.index);
556 dout("__ceph_destroy_xattrs p=%p\n", p);
559 xattr = rb_entry(p, struct ceph_inode_xattr, node);
562 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
563 xattr->name_len, xattr->name);
564 rb_erase(tmp, &ci->i_xattrs.index);
569 ci->i_xattrs.names_size = 0;
570 ci->i_xattrs.vals_size = 0;
571 ci->i_xattrs.index_version = 0;
572 ci->i_xattrs.count = 0;
573 ci->i_xattrs.index = RB_ROOT;
576 static int __build_xattrs(struct inode *inode)
577 __releases(ci->i_ceph_lock)
578 __acquires(ci->i_ceph_lock)
584 const char *name, *val;
585 struct ceph_inode_info *ci = ceph_inode(inode);
587 struct ceph_inode_xattr **xattrs = NULL;
591 dout("__build_xattrs() len=%d\n",
592 ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
594 if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
595 return 0; /* already built */
597 __ceph_destroy_xattrs(ci);
600 /* updated internal xattr rb tree */
601 if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
602 p = ci->i_xattrs.blob->vec.iov_base;
603 end = p + ci->i_xattrs.blob->vec.iov_len;
604 ceph_decode_32_safe(&p, end, numattr, bad);
605 xattr_version = ci->i_xattrs.version;
606 spin_unlock(&ci->i_ceph_lock);
608 xattrs = kcalloc(numattr, sizeof(struct ceph_inode_xattr *),
614 for (i = 0; i < numattr; i++) {
615 xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
621 spin_lock(&ci->i_ceph_lock);
622 if (ci->i_xattrs.version != xattr_version) {
623 /* lost a race, retry */
624 for (i = 0; i < numattr; i++)
632 ceph_decode_32_safe(&p, end, len, bad);
636 ceph_decode_32_safe(&p, end, len, bad);
640 err = __set_xattr(ci, name, namelen, val, len,
641 0, 0, &xattrs[numattr]);
648 ci->i_xattrs.index_version = ci->i_xattrs.version;
649 ci->i_xattrs.dirty = false;
653 spin_lock(&ci->i_ceph_lock);
656 for (i = 0; i < numattr; i++)
660 ci->i_xattrs.names_size = 0;
664 static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
668 * 4 bytes for the length, and additional 4 bytes per each xattr name,
669 * 4 bytes per each value
671 int size = 4 + ci->i_xattrs.count*(4 + 4) +
672 ci->i_xattrs.names_size +
673 ci->i_xattrs.vals_size;
674 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
675 ci->i_xattrs.count, ci->i_xattrs.names_size,
676 ci->i_xattrs.vals_size);
679 size += 4 + 4 + name_size + val_size;
685 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
686 * and swap into place.
688 void __ceph_build_xattrs_blob(struct ceph_inode_info *ci)
691 struct ceph_inode_xattr *xattr = NULL;
694 dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
695 if (ci->i_xattrs.dirty) {
696 int need = __get_required_blob_size(ci, 0, 0);
698 BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
700 p = rb_first(&ci->i_xattrs.index);
701 dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
703 ceph_encode_32(&dest, ci->i_xattrs.count);
705 xattr = rb_entry(p, struct ceph_inode_xattr, node);
707 ceph_encode_32(&dest, xattr->name_len);
708 memcpy(dest, xattr->name, xattr->name_len);
709 dest += xattr->name_len;
710 ceph_encode_32(&dest, xattr->val_len);
711 memcpy(dest, xattr->val, xattr->val_len);
712 dest += xattr->val_len;
717 /* adjust buffer len; it may be larger than we need */
718 ci->i_xattrs.prealloc_blob->vec.iov_len =
719 dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
721 if (ci->i_xattrs.blob)
722 ceph_buffer_put(ci->i_xattrs.blob);
723 ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
724 ci->i_xattrs.prealloc_blob = NULL;
725 ci->i_xattrs.dirty = false;
726 ci->i_xattrs.version++;
730 static inline int __get_request_mask(struct inode *in) {
731 struct ceph_mds_request *req = current->journal_info;
733 if (req && req->r_target_inode == in) {
734 if (req->r_op == CEPH_MDS_OP_LOOKUP ||
735 req->r_op == CEPH_MDS_OP_LOOKUPINO ||
736 req->r_op == CEPH_MDS_OP_LOOKUPPARENT ||
737 req->r_op == CEPH_MDS_OP_GETATTR) {
738 mask = le32_to_cpu(req->r_args.getattr.mask);
739 } else if (req->r_op == CEPH_MDS_OP_OPEN ||
740 req->r_op == CEPH_MDS_OP_CREATE) {
741 mask = le32_to_cpu(req->r_args.open.mask);
747 ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
750 struct ceph_inode_info *ci = ceph_inode(inode);
751 struct ceph_inode_xattr *xattr;
752 struct ceph_vxattr *vxattr = NULL;
756 /* let's see if a virtual xattr was requested */
757 vxattr = ceph_match_vxattr(inode, name);
759 err = ceph_do_getattr(inode, 0, true);
763 if (!(vxattr->exists_cb && !vxattr->exists_cb(ci)))
764 err = vxattr->getxattr_cb(ci, value, size);
768 req_mask = __get_request_mask(inode);
770 spin_lock(&ci->i_ceph_lock);
771 dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
772 ci->i_xattrs.version, ci->i_xattrs.index_version);
774 if (ci->i_xattrs.version == 0 ||
775 !((req_mask & CEPH_CAP_XATTR_SHARED) ||
776 __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1))) {
777 spin_unlock(&ci->i_ceph_lock);
779 /* security module gets xattr while filling trace */
780 if (current->journal_info) {
781 pr_warn_ratelimited("sync getxattr %p "
782 "during filling trace\n", inode);
786 /* get xattrs from mds (if we don't already have them) */
787 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
790 spin_lock(&ci->i_ceph_lock);
793 err = __build_xattrs(inode);
797 err = -ENODATA; /* == ENOATTR */
798 xattr = __get_xattr(ci, name);
803 if (size && size < xattr->val_len)
806 err = xattr->val_len;
810 memcpy(value, xattr->val, xattr->val_len);
812 if (current->journal_info &&
813 !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN))
814 ci->i_ceph_flags |= CEPH_I_SEC_INITED;
816 spin_unlock(&ci->i_ceph_lock);
820 ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
822 struct inode *inode = d_inode(dentry);
823 struct ceph_inode_info *ci = ceph_inode(inode);
824 struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
831 spin_lock(&ci->i_ceph_lock);
832 dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
833 ci->i_xattrs.version, ci->i_xattrs.index_version);
835 if (ci->i_xattrs.version == 0 ||
836 !__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1)) {
837 spin_unlock(&ci->i_ceph_lock);
838 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
841 spin_lock(&ci->i_ceph_lock);
844 err = __build_xattrs(inode);
848 * Start with virtual dir xattr names (if any) (including
849 * terminating '\0' characters for each).
851 vir_namelen = ceph_vxattrs_name_size(vxattrs);
853 /* adding 1 byte per each variable due to the null termination */
854 namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
856 if (size && vir_namelen + namelen > size)
859 err = namelen + vir_namelen;
863 names = __copy_xattr_names(ci, names);
865 /* virtual xattr names, too */
868 for (i = 0; vxattrs[i].name; i++) {
869 if (!vxattrs[i].hidden &&
870 !(vxattrs[i].exists_cb &&
871 !vxattrs[i].exists_cb(ci))) {
872 len = sprintf(names, "%s", vxattrs[i].name);
880 spin_unlock(&ci->i_ceph_lock);
884 static int ceph_sync_setxattr(struct inode *inode, const char *name,
885 const char *value, size_t size, int flags)
887 struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
888 struct ceph_inode_info *ci = ceph_inode(inode);
889 struct ceph_mds_request *req;
890 struct ceph_mds_client *mdsc = fsc->mdsc;
891 struct ceph_pagelist *pagelist = NULL;
892 int op = CEPH_MDS_OP_SETXATTR;
896 /* copy value into pagelist */
897 pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
901 ceph_pagelist_init(pagelist);
902 err = ceph_pagelist_append(pagelist, value, size);
906 if (flags & CEPH_XATTR_REPLACE)
907 op = CEPH_MDS_OP_RMXATTR;
909 flags |= CEPH_XATTR_REMOVE;
912 dout("setxattr value=%.*s\n", (int)size, value);
915 req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
921 req->r_path2 = kstrdup(name, GFP_NOFS);
923 ceph_mdsc_put_request(req);
928 if (op == CEPH_MDS_OP_SETXATTR) {
929 req->r_args.setxattr.flags = cpu_to_le32(flags);
930 req->r_pagelist = pagelist;
934 req->r_inode = inode;
937 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
939 dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
940 err = ceph_mdsc_do_request(mdsc, NULL, req);
941 ceph_mdsc_put_request(req);
942 dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
946 ceph_pagelist_release(pagelist);
950 int __ceph_setxattr(struct inode *inode, const char *name,
951 const void *value, size_t size, int flags)
953 struct ceph_vxattr *vxattr;
954 struct ceph_inode_info *ci = ceph_inode(inode);
955 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
956 struct ceph_cap_flush *prealloc_cf = NULL;
960 int name_len = strlen(name);
962 char *newname = NULL;
964 struct ceph_inode_xattr *xattr = NULL;
965 int required_blob_size;
966 bool lock_snap_rwsem = false;
968 if (ceph_snap(inode) != CEPH_NOSNAP)
971 vxattr = ceph_match_vxattr(inode, name);
972 if (vxattr && vxattr->readonly)
975 /* pass any unhandled ceph.* xattrs through to the MDS */
976 if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
977 goto do_sync_unlocked;
979 /* preallocate memory for xattr name, value, index node */
981 newname = kmemdup(name, name_len + 1, GFP_NOFS);
986 newval = kmemdup(value, val_len, GFP_NOFS);
991 xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
995 prealloc_cf = ceph_alloc_cap_flush();
999 spin_lock(&ci->i_ceph_lock);
1001 issued = __ceph_caps_issued(ci, NULL);
1002 if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
1005 if (!lock_snap_rwsem && !ci->i_head_snapc) {
1006 lock_snap_rwsem = true;
1007 if (!down_read_trylock(&mdsc->snap_rwsem)) {
1008 spin_unlock(&ci->i_ceph_lock);
1009 down_read(&mdsc->snap_rwsem);
1010 spin_lock(&ci->i_ceph_lock);
1015 dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
1016 __build_xattrs(inode);
1018 required_blob_size = __get_required_blob_size(ci, name_len, val_len);
1020 if (!ci->i_xattrs.prealloc_blob ||
1021 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1022 struct ceph_buffer *blob;
1024 spin_unlock(&ci->i_ceph_lock);
1025 dout(" preaallocating new blob size=%d\n", required_blob_size);
1026 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1028 goto do_sync_unlocked;
1029 spin_lock(&ci->i_ceph_lock);
1030 if (ci->i_xattrs.prealloc_blob)
1031 ceph_buffer_put(ci->i_xattrs.prealloc_blob);
1032 ci->i_xattrs.prealloc_blob = blob;
1036 err = __set_xattr(ci, newname, name_len, newval, val_len,
1037 flags, value ? 1 : -1, &xattr);
1040 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL,
1042 ci->i_xattrs.dirty = true;
1043 inode->i_ctime = current_time(inode);
1046 spin_unlock(&ci->i_ceph_lock);
1047 if (lock_snap_rwsem)
1048 up_read(&mdsc->snap_rwsem);
1050 __mark_inode_dirty(inode, dirty);
1051 ceph_free_cap_flush(prealloc_cf);
1055 spin_unlock(&ci->i_ceph_lock);
1057 if (lock_snap_rwsem)
1058 up_read(&mdsc->snap_rwsem);
1060 /* security module set xattr while filling trace */
1061 if (current->journal_info) {
1062 pr_warn_ratelimited("sync setxattr %p "
1063 "during filling trace\n", inode);
1066 err = ceph_sync_setxattr(inode, name, value, size, flags);
1069 ceph_free_cap_flush(prealloc_cf);
1076 static int ceph_get_xattr_handler(const struct xattr_handler *handler,
1077 struct dentry *dentry, struct inode *inode,
1078 const char *name, void *value, size_t size)
1080 if (!ceph_is_valid_xattr(name))
1082 return __ceph_getxattr(inode, name, value, size);
1085 static int ceph_set_xattr_handler(const struct xattr_handler *handler,
1086 struct dentry *unused, struct inode *inode,
1087 const char *name, const void *value,
1088 size_t size, int flags)
1090 if (!ceph_is_valid_xattr(name))
1092 return __ceph_setxattr(inode, name, value, size, flags);
1095 static const struct xattr_handler ceph_other_xattr_handler = {
1096 .prefix = "", /* match any name => handlers called with full name */
1097 .get = ceph_get_xattr_handler,
1098 .set = ceph_set_xattr_handler,
1101 #ifdef CONFIG_SECURITY
1102 bool ceph_security_xattr_wanted(struct inode *in)
1104 return in->i_security != NULL;
1107 bool ceph_security_xattr_deadlock(struct inode *in)
1109 struct ceph_inode_info *ci;
1111 if (!in->i_security)
1113 ci = ceph_inode(in);
1114 spin_lock(&ci->i_ceph_lock);
1115 ret = !(ci->i_ceph_flags & CEPH_I_SEC_INITED) &&
1116 !(ci->i_xattrs.version > 0 &&
1117 __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 0));
1118 spin_unlock(&ci->i_ceph_lock);