1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3 #include <linux/ceph/pagelist.h>
6 #include "mds_client.h"
8 #include <linux/ceph/decode.h>
10 #include <linux/xattr.h>
11 #include <linux/posix_acl_xattr.h>
12 #include <linux/slab.h>
14 #define XATTR_CEPH_PREFIX "ceph."
15 #define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
17 static int __remove_xattr(struct ceph_inode_info *ci,
18 struct ceph_inode_xattr *xattr);
20 static const struct xattr_handler ceph_other_xattr_handler;
23 * List of handlers for synthetic system.* attributes. Other
24 * attributes are handled directly.
26 const struct xattr_handler *ceph_xattr_handlers[] = {
27 #ifdef CONFIG_CEPH_FS_POSIX_ACL
28 &posix_acl_access_xattr_handler,
29 &posix_acl_default_xattr_handler,
31 &ceph_other_xattr_handler,
35 static bool ceph_is_valid_xattr(const char *name)
37 return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
38 !strncmp(name, XATTR_SECURITY_PREFIX,
39 XATTR_SECURITY_PREFIX_LEN) ||
40 !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
41 !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
45 * These define virtual xattrs exposing the recursive directory
46 * statistics and layout metadata.
50 size_t name_size; /* strlen(name) + 1 (for '\0') */
51 size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
53 bool (*exists_cb)(struct ceph_inode_info *ci);
57 #define VXATTR_FLAG_READONLY (1<<0)
58 #define VXATTR_FLAG_HIDDEN (1<<1)
59 #define VXATTR_FLAG_RSTAT (1<<2)
63 static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
65 struct ceph_file_layout *fl = &ci->i_layout;
66 return (fl->stripe_unit > 0 || fl->stripe_count > 0 ||
67 fl->object_size > 0 || fl->pool_id >= 0 ||
68 rcu_dereference_raw(fl->pool_ns) != NULL);
71 static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
74 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
75 struct ceph_osd_client *osdc = &fsc->client->osdc;
76 struct ceph_string *pool_ns;
77 s64 pool = ci->i_layout.pool_id;
78 const char *pool_name;
79 const char *ns_field = " pool_namespace=";
81 size_t len, total_len = 0;
84 pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
86 dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
87 down_read(&osdc->lock);
88 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
90 len = snprintf(buf, sizeof(buf),
91 "stripe_unit=%u stripe_count=%u object_size=%u pool=",
92 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
93 ci->i_layout.object_size);
94 total_len = len + strlen(pool_name);
96 len = snprintf(buf, sizeof(buf),
97 "stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
98 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
99 ci->i_layout.object_size, (unsigned long long)pool);
104 total_len += strlen(ns_field) + pool_ns->len;
107 if (size >= total_len) {
108 memcpy(val, buf, len);
111 len = strlen(pool_name);
112 memcpy(val + ret, pool_name, len);
116 len = strlen(ns_field);
117 memcpy(val + ret, ns_field, len);
119 memcpy(val + ret, pool_ns->str, pool_ns->len);
123 up_read(&osdc->lock);
124 ceph_put_string(pool_ns);
128 static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
129 char *val, size_t size)
131 return snprintf(val, size, "%u", ci->i_layout.stripe_unit);
134 static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
135 char *val, size_t size)
137 return snprintf(val, size, "%u", ci->i_layout.stripe_count);
140 static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
141 char *val, size_t size)
143 return snprintf(val, size, "%u", ci->i_layout.object_size);
146 static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
147 char *val, size_t size)
150 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
151 struct ceph_osd_client *osdc = &fsc->client->osdc;
152 s64 pool = ci->i_layout.pool_id;
153 const char *pool_name;
155 down_read(&osdc->lock);
156 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
158 ret = snprintf(val, size, "%s", pool_name);
160 ret = snprintf(val, size, "%lld", (unsigned long long)pool);
161 up_read(&osdc->lock);
165 static size_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info *ci,
166 char *val, size_t size)
169 struct ceph_string *ns = ceph_try_get_string(ci->i_layout.pool_ns);
171 ret = snprintf(val, size, "%.*s", (int)ns->len, ns->str);
179 static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
182 return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
185 static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
188 return snprintf(val, size, "%lld", ci->i_files);
191 static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
194 return snprintf(val, size, "%lld", ci->i_subdirs);
197 static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
200 return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
203 static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
206 return snprintf(val, size, "%lld", ci->i_rfiles);
209 static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
212 return snprintf(val, size, "%lld", ci->i_rsubdirs);
215 static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
218 return snprintf(val, size, "%lld", ci->i_rbytes);
221 static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
224 return snprintf(val, size, "%lld.%09ld", ci->i_rctime.tv_sec,
225 ci->i_rctime.tv_nsec);
230 static bool ceph_vxattrcb_quota_exists(struct ceph_inode_info *ci)
233 spin_lock(&ci->i_ceph_lock);
234 if ((ci->i_max_files || ci->i_max_bytes) &&
235 ci->i_vino.snap == CEPH_NOSNAP &&
237 ci->i_snap_realm->ino == ci->i_vino.ino)
239 spin_unlock(&ci->i_ceph_lock);
243 static size_t ceph_vxattrcb_quota(struct ceph_inode_info *ci, char *val,
246 return snprintf(val, size, "max_bytes=%llu max_files=%llu",
247 ci->i_max_bytes, ci->i_max_files);
250 static size_t ceph_vxattrcb_quota_max_bytes(struct ceph_inode_info *ci,
251 char *val, size_t size)
253 return snprintf(val, size, "%llu", ci->i_max_bytes);
256 static size_t ceph_vxattrcb_quota_max_files(struct ceph_inode_info *ci,
257 char *val, size_t size)
259 return snprintf(val, size, "%llu", ci->i_max_files);
262 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
263 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
264 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
266 #define XATTR_NAME_CEPH(_type, _name, _flags) \
268 .name = CEPH_XATTR_NAME(_type, _name), \
269 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
270 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
272 .flags = (VXATTR_FLAG_READONLY | _flags), \
274 #define XATTR_RSTAT_FIELD(_type, _name) \
275 XATTR_NAME_CEPH(_type, _name, VXATTR_FLAG_RSTAT)
276 #define XATTR_RSTAT_FIELD_UPDATABLE(_type, _name) \
278 .name = CEPH_XATTR_NAME(_type, _name), \
279 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
280 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
282 .flags = VXATTR_FLAG_RSTAT, \
284 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
286 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
287 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
288 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
289 .exists_cb = ceph_vxattrcb_layout_exists, \
290 .flags = VXATTR_FLAG_HIDDEN, \
292 #define XATTR_QUOTA_FIELD(_type, _name) \
294 .name = CEPH_XATTR_NAME(_type, _name), \
295 .name_size = sizeof(CEPH_XATTR_NAME(_type, _name)), \
296 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
297 .exists_cb = ceph_vxattrcb_quota_exists, \
298 .flags = VXATTR_FLAG_HIDDEN, \
301 static struct ceph_vxattr ceph_dir_vxattrs[] = {
303 .name = "ceph.dir.layout",
304 .name_size = sizeof("ceph.dir.layout"),
305 .getxattr_cb = ceph_vxattrcb_layout,
306 .exists_cb = ceph_vxattrcb_layout_exists,
307 .flags = VXATTR_FLAG_HIDDEN,
309 XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
310 XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
311 XATTR_LAYOUT_FIELD(dir, layout, object_size),
312 XATTR_LAYOUT_FIELD(dir, layout, pool),
313 XATTR_LAYOUT_FIELD(dir, layout, pool_namespace),
314 XATTR_NAME_CEPH(dir, entries, 0),
315 XATTR_NAME_CEPH(dir, files, 0),
316 XATTR_NAME_CEPH(dir, subdirs, 0),
317 XATTR_RSTAT_FIELD(dir, rentries),
318 XATTR_RSTAT_FIELD(dir, rfiles),
319 XATTR_RSTAT_FIELD(dir, rsubdirs),
320 XATTR_RSTAT_FIELD(dir, rbytes),
321 XATTR_RSTAT_FIELD_UPDATABLE(dir, rctime),
323 .name = "ceph.quota",
324 .name_size = sizeof("ceph.quota"),
325 .getxattr_cb = ceph_vxattrcb_quota,
326 .exists_cb = ceph_vxattrcb_quota_exists,
327 .flags = VXATTR_FLAG_HIDDEN,
329 XATTR_QUOTA_FIELD(quota, max_bytes),
330 XATTR_QUOTA_FIELD(quota, max_files),
331 { .name = NULL, 0 } /* Required table terminator */
333 static size_t ceph_dir_vxattrs_name_size; /* total size of all names */
337 static struct ceph_vxattr ceph_file_vxattrs[] = {
339 .name = "ceph.file.layout",
340 .name_size = sizeof("ceph.file.layout"),
341 .getxattr_cb = ceph_vxattrcb_layout,
342 .exists_cb = ceph_vxattrcb_layout_exists,
343 .flags = VXATTR_FLAG_HIDDEN,
345 XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
346 XATTR_LAYOUT_FIELD(file, layout, stripe_count),
347 XATTR_LAYOUT_FIELD(file, layout, object_size),
348 XATTR_LAYOUT_FIELD(file, layout, pool),
349 XATTR_LAYOUT_FIELD(file, layout, pool_namespace),
350 { .name = NULL, 0 } /* Required table terminator */
352 static size_t ceph_file_vxattrs_name_size; /* total size of all names */
354 static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
356 if (S_ISDIR(inode->i_mode))
357 return ceph_dir_vxattrs;
358 else if (S_ISREG(inode->i_mode))
359 return ceph_file_vxattrs;
363 static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
365 if (vxattrs == ceph_dir_vxattrs)
366 return ceph_dir_vxattrs_name_size;
367 if (vxattrs == ceph_file_vxattrs)
368 return ceph_file_vxattrs_name_size;
374 * Compute the aggregate size (including terminating '\0') of all
375 * virtual extended attribute names in the given vxattr table.
377 static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
379 struct ceph_vxattr *vxattr;
382 for (vxattr = vxattrs; vxattr->name; vxattr++) {
383 if (!(vxattr->flags & VXATTR_FLAG_HIDDEN))
384 size += vxattr->name_size;
390 /* Routines called at initialization and exit time */
392 void __init ceph_xattr_init(void)
394 ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
395 ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
398 void ceph_xattr_exit(void)
400 ceph_dir_vxattrs_name_size = 0;
401 ceph_file_vxattrs_name_size = 0;
404 static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
407 struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
410 while (vxattr->name) {
411 if (!strcmp(vxattr->name, name))
420 static int __set_xattr(struct ceph_inode_info *ci,
421 const char *name, int name_len,
422 const char *val, int val_len,
423 int flags, int update_xattr,
424 struct ceph_inode_xattr **newxattr)
427 struct rb_node *parent = NULL;
428 struct ceph_inode_xattr *xattr = NULL;
432 p = &ci->i_xattrs.index.rb_node;
435 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
436 c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
442 if (name_len == xattr->name_len)
444 else if (name_len < xattr->name_len)
455 if (xattr && (flags & XATTR_CREATE))
457 else if (!xattr && (flags & XATTR_REPLACE))
465 if (update_xattr < 0) {
467 __remove_xattr(ci, xattr);
478 xattr->name_len = name_len;
479 xattr->should_free_name = update_xattr;
481 ci->i_xattrs.count++;
482 dout("__set_xattr count=%d\n", ci->i_xattrs.count);
486 if (xattr->should_free_val)
487 kfree((void *)xattr->val);
493 ci->i_xattrs.names_size -= xattr->name_len;
494 ci->i_xattrs.vals_size -= xattr->val_len;
496 ci->i_xattrs.names_size += name_len;
497 ci->i_xattrs.vals_size += val_len;
503 xattr->val_len = val_len;
504 xattr->dirty = update_xattr;
505 xattr->should_free_val = (val && update_xattr);
508 rb_link_node(&xattr->node, parent, p);
509 rb_insert_color(&xattr->node, &ci->i_xattrs.index);
510 dout("__set_xattr_val p=%p\n", p);
513 dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
514 ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);
519 static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
523 struct rb_node *parent = NULL;
524 struct ceph_inode_xattr *xattr = NULL;
525 int name_len = strlen(name);
528 p = &ci->i_xattrs.index.rb_node;
531 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
532 c = strncmp(name, xattr->name, xattr->name_len);
533 if (c == 0 && name_len > xattr->name_len)
540 dout("__get_xattr %s: found %.*s\n", name,
541 xattr->val_len, xattr->val);
546 dout("__get_xattr %s: not found\n", name);
551 static void __free_xattr(struct ceph_inode_xattr *xattr)
555 if (xattr->should_free_name)
556 kfree((void *)xattr->name);
557 if (xattr->should_free_val)
558 kfree((void *)xattr->val);
563 static int __remove_xattr(struct ceph_inode_info *ci,
564 struct ceph_inode_xattr *xattr)
569 rb_erase(&xattr->node, &ci->i_xattrs.index);
571 if (xattr->should_free_name)
572 kfree((void *)xattr->name);
573 if (xattr->should_free_val)
574 kfree((void *)xattr->val);
576 ci->i_xattrs.names_size -= xattr->name_len;
577 ci->i_xattrs.vals_size -= xattr->val_len;
578 ci->i_xattrs.count--;
584 static char *__copy_xattr_names(struct ceph_inode_info *ci,
588 struct ceph_inode_xattr *xattr = NULL;
590 p = rb_first(&ci->i_xattrs.index);
591 dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
594 xattr = rb_entry(p, struct ceph_inode_xattr, node);
595 memcpy(dest, xattr->name, xattr->name_len);
596 dest[xattr->name_len] = '\0';
598 dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
599 xattr->name_len, ci->i_xattrs.names_size);
601 dest += xattr->name_len + 1;
608 void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
610 struct rb_node *p, *tmp;
611 struct ceph_inode_xattr *xattr = NULL;
613 p = rb_first(&ci->i_xattrs.index);
615 dout("__ceph_destroy_xattrs p=%p\n", p);
618 xattr = rb_entry(p, struct ceph_inode_xattr, node);
621 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
622 xattr->name_len, xattr->name);
623 rb_erase(tmp, &ci->i_xattrs.index);
628 ci->i_xattrs.names_size = 0;
629 ci->i_xattrs.vals_size = 0;
630 ci->i_xattrs.index_version = 0;
631 ci->i_xattrs.count = 0;
632 ci->i_xattrs.index = RB_ROOT;
635 static int __build_xattrs(struct inode *inode)
636 __releases(ci->i_ceph_lock)
637 __acquires(ci->i_ceph_lock)
643 const char *name, *val;
644 struct ceph_inode_info *ci = ceph_inode(inode);
646 struct ceph_inode_xattr **xattrs = NULL;
650 dout("__build_xattrs() len=%d\n",
651 ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
653 if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
654 return 0; /* already built */
656 __ceph_destroy_xattrs(ci);
659 /* updated internal xattr rb tree */
660 if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
661 p = ci->i_xattrs.blob->vec.iov_base;
662 end = p + ci->i_xattrs.blob->vec.iov_len;
663 ceph_decode_32_safe(&p, end, numattr, bad);
664 xattr_version = ci->i_xattrs.version;
665 spin_unlock(&ci->i_ceph_lock);
667 xattrs = kcalloc(numattr, sizeof(struct ceph_inode_xattr *),
673 for (i = 0; i < numattr; i++) {
674 xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
680 spin_lock(&ci->i_ceph_lock);
681 if (ci->i_xattrs.version != xattr_version) {
682 /* lost a race, retry */
683 for (i = 0; i < numattr; i++)
691 ceph_decode_32_safe(&p, end, len, bad);
695 ceph_decode_32_safe(&p, end, len, bad);
699 err = __set_xattr(ci, name, namelen, val, len,
700 0, 0, &xattrs[numattr]);
707 ci->i_xattrs.index_version = ci->i_xattrs.version;
708 ci->i_xattrs.dirty = false;
712 spin_lock(&ci->i_ceph_lock);
715 for (i = 0; i < numattr; i++)
719 ci->i_xattrs.names_size = 0;
723 static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
727 * 4 bytes for the length, and additional 4 bytes per each xattr name,
728 * 4 bytes per each value
730 int size = 4 + ci->i_xattrs.count*(4 + 4) +
731 ci->i_xattrs.names_size +
732 ci->i_xattrs.vals_size;
733 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
734 ci->i_xattrs.count, ci->i_xattrs.names_size,
735 ci->i_xattrs.vals_size);
738 size += 4 + 4 + name_size + val_size;
744 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
745 * and swap into place. It returns the old i_xattrs.blob (or NULL) so
746 * that it can be freed by the caller as the i_ceph_lock is likely to be
749 struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci)
752 struct ceph_inode_xattr *xattr = NULL;
753 struct ceph_buffer *old_blob = NULL;
756 dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
757 if (ci->i_xattrs.dirty) {
758 int need = __get_required_blob_size(ci, 0, 0);
760 BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
762 p = rb_first(&ci->i_xattrs.index);
763 dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
765 ceph_encode_32(&dest, ci->i_xattrs.count);
767 xattr = rb_entry(p, struct ceph_inode_xattr, node);
769 ceph_encode_32(&dest, xattr->name_len);
770 memcpy(dest, xattr->name, xattr->name_len);
771 dest += xattr->name_len;
772 ceph_encode_32(&dest, xattr->val_len);
773 memcpy(dest, xattr->val, xattr->val_len);
774 dest += xattr->val_len;
779 /* adjust buffer len; it may be larger than we need */
780 ci->i_xattrs.prealloc_blob->vec.iov_len =
781 dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
783 if (ci->i_xattrs.blob)
784 old_blob = ci->i_xattrs.blob;
785 ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
786 ci->i_xattrs.prealloc_blob = NULL;
787 ci->i_xattrs.dirty = false;
788 ci->i_xattrs.version++;
794 static inline int __get_request_mask(struct inode *in) {
795 struct ceph_mds_request *req = current->journal_info;
797 if (req && req->r_target_inode == in) {
798 if (req->r_op == CEPH_MDS_OP_LOOKUP ||
799 req->r_op == CEPH_MDS_OP_LOOKUPINO ||
800 req->r_op == CEPH_MDS_OP_LOOKUPPARENT ||
801 req->r_op == CEPH_MDS_OP_GETATTR) {
802 mask = le32_to_cpu(req->r_args.getattr.mask);
803 } else if (req->r_op == CEPH_MDS_OP_OPEN ||
804 req->r_op == CEPH_MDS_OP_CREATE) {
805 mask = le32_to_cpu(req->r_args.open.mask);
811 ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
814 struct ceph_inode_info *ci = ceph_inode(inode);
815 struct ceph_inode_xattr *xattr;
816 struct ceph_vxattr *vxattr = NULL;
820 /* let's see if a virtual xattr was requested */
821 vxattr = ceph_match_vxattr(inode, name);
824 if (vxattr->flags & VXATTR_FLAG_RSTAT)
825 mask |= CEPH_STAT_RSTAT;
826 err = ceph_do_getattr(inode, mask, true);
830 if (!(vxattr->exists_cb && !vxattr->exists_cb(ci))) {
831 err = vxattr->getxattr_cb(ci, value, size);
832 if (size && size < err)
838 req_mask = __get_request_mask(inode);
840 spin_lock(&ci->i_ceph_lock);
841 dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
842 ci->i_xattrs.version, ci->i_xattrs.index_version);
844 if (ci->i_xattrs.version == 0 ||
845 !((req_mask & CEPH_CAP_XATTR_SHARED) ||
846 __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1))) {
847 spin_unlock(&ci->i_ceph_lock);
849 /* security module gets xattr while filling trace */
850 if (current->journal_info) {
851 pr_warn_ratelimited("sync getxattr %p "
852 "during filling trace\n", inode);
856 /* get xattrs from mds (if we don't already have them) */
857 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
860 spin_lock(&ci->i_ceph_lock);
863 err = __build_xattrs(inode);
867 err = -ENODATA; /* == ENOATTR */
868 xattr = __get_xattr(ci, name);
873 if (size && size < xattr->val_len)
876 err = xattr->val_len;
880 memcpy(value, xattr->val, xattr->val_len);
882 if (current->journal_info &&
883 !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN))
884 ci->i_ceph_flags |= CEPH_I_SEC_INITED;
886 spin_unlock(&ci->i_ceph_lock);
890 ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
892 struct inode *inode = d_inode(dentry);
893 struct ceph_inode_info *ci = ceph_inode(inode);
894 struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
901 spin_lock(&ci->i_ceph_lock);
902 dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
903 ci->i_xattrs.version, ci->i_xattrs.index_version);
905 if (ci->i_xattrs.version == 0 ||
906 !__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1)) {
907 spin_unlock(&ci->i_ceph_lock);
908 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
911 spin_lock(&ci->i_ceph_lock);
914 err = __build_xattrs(inode);
918 * Start with virtual dir xattr names (if any) (including
919 * terminating '\0' characters for each).
921 vir_namelen = ceph_vxattrs_name_size(vxattrs);
923 /* adding 1 byte per each variable due to the null termination */
924 namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
926 if (size && vir_namelen + namelen > size)
929 err = namelen + vir_namelen;
933 names = __copy_xattr_names(ci, names);
935 /* virtual xattr names, too */
938 for (i = 0; vxattrs[i].name; i++) {
939 if (!(vxattrs[i].flags & VXATTR_FLAG_HIDDEN) &&
940 !(vxattrs[i].exists_cb &&
941 !vxattrs[i].exists_cb(ci))) {
942 len = sprintf(names, "%s", vxattrs[i].name);
950 spin_unlock(&ci->i_ceph_lock);
954 static int ceph_sync_setxattr(struct inode *inode, const char *name,
955 const char *value, size_t size, int flags)
957 struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
958 struct ceph_inode_info *ci = ceph_inode(inode);
959 struct ceph_mds_request *req;
960 struct ceph_mds_client *mdsc = fsc->mdsc;
961 struct ceph_pagelist *pagelist = NULL;
962 int op = CEPH_MDS_OP_SETXATTR;
966 /* copy value into pagelist */
967 pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
971 ceph_pagelist_init(pagelist);
972 err = ceph_pagelist_append(pagelist, value, size);
976 if (flags & CEPH_XATTR_REPLACE)
977 op = CEPH_MDS_OP_RMXATTR;
979 flags |= CEPH_XATTR_REMOVE;
982 dout("setxattr value=%.*s\n", (int)size, value);
985 req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
991 req->r_path2 = kstrdup(name, GFP_NOFS);
993 ceph_mdsc_put_request(req);
998 if (op == CEPH_MDS_OP_SETXATTR) {
999 req->r_args.setxattr.flags = cpu_to_le32(flags);
1000 req->r_pagelist = pagelist;
1004 req->r_inode = inode;
1006 req->r_num_caps = 1;
1007 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
1009 dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
1010 err = ceph_mdsc_do_request(mdsc, NULL, req);
1011 ceph_mdsc_put_request(req);
1012 dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
1016 ceph_pagelist_release(pagelist);
1020 int __ceph_setxattr(struct inode *inode, const char *name,
1021 const void *value, size_t size, int flags)
1023 struct ceph_vxattr *vxattr;
1024 struct ceph_inode_info *ci = ceph_inode(inode);
1025 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
1026 struct ceph_cap_flush *prealloc_cf = NULL;
1027 struct ceph_buffer *old_blob = NULL;
1031 int name_len = strlen(name);
1033 char *newname = NULL;
1034 char *newval = NULL;
1035 struct ceph_inode_xattr *xattr = NULL;
1036 int required_blob_size;
1037 bool check_realm = false;
1038 bool lock_snap_rwsem = false;
1040 if (ceph_snap(inode) != CEPH_NOSNAP)
1043 vxattr = ceph_match_vxattr(inode, name);
1045 if (vxattr->flags & VXATTR_FLAG_READONLY)
1047 if (value && !strncmp(vxattr->name, "ceph.quota", 10))
1051 /* pass any unhandled ceph.* xattrs through to the MDS */
1052 if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
1053 goto do_sync_unlocked;
1055 /* preallocate memory for xattr name, value, index node */
1057 newname = kmemdup(name, name_len + 1, GFP_NOFS);
1062 newval = kmemdup(value, val_len, GFP_NOFS);
1067 xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
1071 prealloc_cf = ceph_alloc_cap_flush();
1075 spin_lock(&ci->i_ceph_lock);
1077 issued = __ceph_caps_issued(ci, NULL);
1078 if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
1081 if (!lock_snap_rwsem && !ci->i_head_snapc) {
1082 lock_snap_rwsem = true;
1083 if (!down_read_trylock(&mdsc->snap_rwsem)) {
1084 spin_unlock(&ci->i_ceph_lock);
1085 down_read(&mdsc->snap_rwsem);
1086 spin_lock(&ci->i_ceph_lock);
1091 dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
1092 __build_xattrs(inode);
1094 required_blob_size = __get_required_blob_size(ci, name_len, val_len);
1096 if (!ci->i_xattrs.prealloc_blob ||
1097 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1098 struct ceph_buffer *blob;
1100 spin_unlock(&ci->i_ceph_lock);
1101 ceph_buffer_put(old_blob); /* Shouldn't be required */
1102 dout(" pre-allocating new blob size=%d\n", required_blob_size);
1103 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1105 goto do_sync_unlocked;
1106 spin_lock(&ci->i_ceph_lock);
1107 /* prealloc_blob can't be released while holding i_ceph_lock */
1108 if (ci->i_xattrs.prealloc_blob)
1109 old_blob = ci->i_xattrs.prealloc_blob;
1110 ci->i_xattrs.prealloc_blob = blob;
1114 err = __set_xattr(ci, newname, name_len, newval, val_len,
1115 flags, value ? 1 : -1, &xattr);
1118 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL,
1120 ci->i_xattrs.dirty = true;
1121 inode->i_ctime = current_time(inode);
1124 spin_unlock(&ci->i_ceph_lock);
1125 ceph_buffer_put(old_blob);
1126 if (lock_snap_rwsem)
1127 up_read(&mdsc->snap_rwsem);
1129 __mark_inode_dirty(inode, dirty);
1130 ceph_free_cap_flush(prealloc_cf);
1134 spin_unlock(&ci->i_ceph_lock);
1136 if (lock_snap_rwsem)
1137 up_read(&mdsc->snap_rwsem);
1139 /* security module set xattr while filling trace */
1140 if (current->journal_info) {
1141 pr_warn_ratelimited("sync setxattr %p "
1142 "during filling trace\n", inode);
1145 err = ceph_sync_setxattr(inode, name, value, size, flags);
1146 if (err >= 0 && check_realm) {
1147 /* check if snaprealm was created for quota inode */
1148 spin_lock(&ci->i_ceph_lock);
1149 if ((ci->i_max_files || ci->i_max_bytes) &&
1150 !(ci->i_snap_realm &&
1151 ci->i_snap_realm->ino == ci->i_vino.ino))
1153 spin_unlock(&ci->i_ceph_lock);
1157 ceph_free_cap_flush(prealloc_cf);
1164 static int ceph_get_xattr_handler(const struct xattr_handler *handler,
1165 struct dentry *dentry, struct inode *inode,
1166 const char *name, void *value, size_t size)
1168 if (!ceph_is_valid_xattr(name))
1170 return __ceph_getxattr(inode, name, value, size);
1173 static int ceph_set_xattr_handler(const struct xattr_handler *handler,
1174 struct dentry *unused, struct inode *inode,
1175 const char *name, const void *value,
1176 size_t size, int flags)
1178 if (!ceph_is_valid_xattr(name))
1180 return __ceph_setxattr(inode, name, value, size, flags);
1183 static const struct xattr_handler ceph_other_xattr_handler = {
1184 .prefix = "", /* match any name => handlers called with full name */
1185 .get = ceph_get_xattr_handler,
1186 .set = ceph_set_xattr_handler,
1189 #ifdef CONFIG_SECURITY
1190 bool ceph_security_xattr_wanted(struct inode *in)
1192 return in->i_security != NULL;
1195 bool ceph_security_xattr_deadlock(struct inode *in)
1197 struct ceph_inode_info *ci;
1199 if (!in->i_security)
1201 ci = ceph_inode(in);
1202 spin_lock(&ci->i_ceph_lock);
1203 ret = !(ci->i_ceph_flags & CEPH_I_SEC_INITED) &&
1204 !(ci->i_xattrs.version > 0 &&
1205 __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 0));
1206 spin_unlock(&ci->i_ceph_lock);