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/security.h>
12 #include <linux/posix_acl_xattr.h>
13 #include <linux/slab.h>
15 #define XATTR_CEPH_PREFIX "ceph."
16 #define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
18 static int __remove_xattr(struct ceph_inode_info *ci,
19 struct ceph_inode_xattr *xattr);
21 static bool ceph_is_valid_xattr(const char *name)
23 return !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
24 !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
25 !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
26 !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
30 * These define virtual xattrs exposing the recursive directory
31 * statistics and layout metadata.
35 size_t name_size; /* strlen(name) + 1 (for '\0') */
36 ssize_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
38 bool (*exists_cb)(struct ceph_inode_info *ci);
42 #define VXATTR_FLAG_READONLY (1<<0)
43 #define VXATTR_FLAG_HIDDEN (1<<1)
44 #define VXATTR_FLAG_RSTAT (1<<2)
48 static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
50 struct ceph_file_layout *fl = &ci->i_layout;
51 return (fl->stripe_unit > 0 || fl->stripe_count > 0 ||
52 fl->object_size > 0 || fl->pool_id >= 0 ||
53 rcu_dereference_raw(fl->pool_ns) != NULL);
56 static ssize_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
59 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
60 struct ceph_osd_client *osdc = &fsc->client->osdc;
61 struct ceph_string *pool_ns;
62 s64 pool = ci->i_layout.pool_id;
63 const char *pool_name;
64 const char *ns_field = " pool_namespace=";
66 size_t len, total_len = 0;
69 pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
71 dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
72 down_read(&osdc->lock);
73 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
75 len = snprintf(buf, sizeof(buf),
76 "stripe_unit=%u stripe_count=%u object_size=%u pool=",
77 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
78 ci->i_layout.object_size);
79 total_len = len + strlen(pool_name);
81 len = snprintf(buf, sizeof(buf),
82 "stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
83 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
84 ci->i_layout.object_size, pool);
89 total_len += strlen(ns_field) + pool_ns->len;
92 if (size >= total_len) {
93 memcpy(val, buf, len);
96 len = strlen(pool_name);
97 memcpy(val + ret, pool_name, len);
101 len = strlen(ns_field);
102 memcpy(val + ret, ns_field, len);
104 memcpy(val + ret, pool_ns->str, pool_ns->len);
108 up_read(&osdc->lock);
109 ceph_put_string(pool_ns);
114 * The convention with strings in xattrs is that they should not be NULL
115 * terminated, since we're returning the length with them. snprintf always
116 * NULL terminates however, so call it on a temporary buffer and then memcpy
117 * the result into place.
119 static __printf(3, 4)
120 int ceph_fmt_xattr(char *val, size_t size, const char *fmt, ...)
124 char buf[96]; /* NB: reevaluate size if new vxattrs are added */
127 ret = vsnprintf(buf, size ? sizeof(buf) : 0, fmt, args);
131 if (size && ret + 1 > sizeof(buf)) {
132 WARN_ONCE(true, "Returned length too big (%d)", ret);
137 memcpy(val, buf, ret);
141 static ssize_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
142 char *val, size_t size)
144 return ceph_fmt_xattr(val, size, "%u", ci->i_layout.stripe_unit);
147 static ssize_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
148 char *val, size_t size)
150 return ceph_fmt_xattr(val, size, "%u", ci->i_layout.stripe_count);
153 static ssize_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
154 char *val, size_t size)
156 return ceph_fmt_xattr(val, size, "%u", ci->i_layout.object_size);
159 static ssize_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
160 char *val, size_t size)
163 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
164 struct ceph_osd_client *osdc = &fsc->client->osdc;
165 s64 pool = ci->i_layout.pool_id;
166 const char *pool_name;
168 down_read(&osdc->lock);
169 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
171 ret = strlen(pool_name);
173 memcpy(val, pool_name, ret);
175 ret = ceph_fmt_xattr(val, size, "%lld", pool);
177 up_read(&osdc->lock);
181 static ssize_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info *ci,
182 char *val, size_t size)
185 struct ceph_string *ns = ceph_try_get_string(ci->i_layout.pool_ns);
190 memcpy(val, ns->str, ret);
198 static ssize_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
201 return ceph_fmt_xattr(val, size, "%lld", ci->i_files + ci->i_subdirs);
204 static ssize_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
207 return ceph_fmt_xattr(val, size, "%lld", ci->i_files);
210 static ssize_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
213 return ceph_fmt_xattr(val, size, "%lld", ci->i_subdirs);
216 static ssize_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
219 return ceph_fmt_xattr(val, size, "%lld",
220 ci->i_rfiles + ci->i_rsubdirs);
223 static ssize_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
226 return ceph_fmt_xattr(val, size, "%lld", ci->i_rfiles);
229 static ssize_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
232 return ceph_fmt_xattr(val, size, "%lld", ci->i_rsubdirs);
235 static ssize_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
238 return ceph_fmt_xattr(val, size, "%lld", ci->i_rbytes);
241 static ssize_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
244 return ceph_fmt_xattr(val, size, "%lld.%09ld", ci->i_rctime.tv_sec,
245 ci->i_rctime.tv_nsec);
249 static bool ceph_vxattrcb_dir_pin_exists(struct ceph_inode_info *ci)
251 return ci->i_dir_pin != -ENODATA;
254 static ssize_t ceph_vxattrcb_dir_pin(struct ceph_inode_info *ci, char *val,
257 return ceph_fmt_xattr(val, size, "%d", (int)ci->i_dir_pin);
261 static bool ceph_vxattrcb_quota_exists(struct ceph_inode_info *ci)
264 spin_lock(&ci->i_ceph_lock);
265 if ((ci->i_max_files || ci->i_max_bytes) &&
266 ci->i_vino.snap == CEPH_NOSNAP &&
268 ci->i_snap_realm->ino == ci->i_vino.ino)
270 spin_unlock(&ci->i_ceph_lock);
274 static ssize_t ceph_vxattrcb_quota(struct ceph_inode_info *ci, char *val,
277 return ceph_fmt_xattr(val, size, "max_bytes=%llu max_files=%llu",
278 ci->i_max_bytes, ci->i_max_files);
281 static ssize_t ceph_vxattrcb_quota_max_bytes(struct ceph_inode_info *ci,
282 char *val, size_t size)
284 return ceph_fmt_xattr(val, size, "%llu", ci->i_max_bytes);
287 static ssize_t ceph_vxattrcb_quota_max_files(struct ceph_inode_info *ci,
288 char *val, size_t size)
290 return ceph_fmt_xattr(val, size, "%llu", ci->i_max_files);
294 static bool ceph_vxattrcb_snap_btime_exists(struct ceph_inode_info *ci)
296 return (ci->i_snap_btime.tv_sec != 0 || ci->i_snap_btime.tv_nsec != 0);
299 static ssize_t ceph_vxattrcb_snap_btime(struct ceph_inode_info *ci, char *val,
302 return ceph_fmt_xattr(val, size, "%lld.%09ld", ci->i_snap_btime.tv_sec,
303 ci->i_snap_btime.tv_nsec);
306 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
307 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
308 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
310 #define XATTR_NAME_CEPH(_type, _name, _flags) \
312 .name = CEPH_XATTR_NAME(_type, _name), \
313 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
314 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
316 .flags = (VXATTR_FLAG_READONLY | _flags), \
318 #define XATTR_RSTAT_FIELD(_type, _name) \
319 XATTR_NAME_CEPH(_type, _name, VXATTR_FLAG_RSTAT)
320 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
322 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
323 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
324 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
325 .exists_cb = ceph_vxattrcb_layout_exists, \
326 .flags = VXATTR_FLAG_HIDDEN, \
328 #define XATTR_QUOTA_FIELD(_type, _name) \
330 .name = CEPH_XATTR_NAME(_type, _name), \
331 .name_size = sizeof(CEPH_XATTR_NAME(_type, _name)), \
332 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
333 .exists_cb = ceph_vxattrcb_quota_exists, \
334 .flags = VXATTR_FLAG_HIDDEN, \
337 static struct ceph_vxattr ceph_dir_vxattrs[] = {
339 .name = "ceph.dir.layout",
340 .name_size = sizeof("ceph.dir.layout"),
341 .getxattr_cb = ceph_vxattrcb_layout,
342 .exists_cb = ceph_vxattrcb_layout_exists,
343 .flags = VXATTR_FLAG_HIDDEN,
345 XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
346 XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
347 XATTR_LAYOUT_FIELD(dir, layout, object_size),
348 XATTR_LAYOUT_FIELD(dir, layout, pool),
349 XATTR_LAYOUT_FIELD(dir, layout, pool_namespace),
350 XATTR_NAME_CEPH(dir, entries, 0),
351 XATTR_NAME_CEPH(dir, files, 0),
352 XATTR_NAME_CEPH(dir, subdirs, 0),
353 XATTR_RSTAT_FIELD(dir, rentries),
354 XATTR_RSTAT_FIELD(dir, rfiles),
355 XATTR_RSTAT_FIELD(dir, rsubdirs),
356 XATTR_RSTAT_FIELD(dir, rbytes),
357 XATTR_RSTAT_FIELD(dir, rctime),
359 .name = "ceph.dir.pin",
360 .name_size = sizeof("ceph.dir.pin"),
361 .getxattr_cb = ceph_vxattrcb_dir_pin,
362 .exists_cb = ceph_vxattrcb_dir_pin_exists,
363 .flags = VXATTR_FLAG_HIDDEN,
366 .name = "ceph.quota",
367 .name_size = sizeof("ceph.quota"),
368 .getxattr_cb = ceph_vxattrcb_quota,
369 .exists_cb = ceph_vxattrcb_quota_exists,
370 .flags = VXATTR_FLAG_HIDDEN,
372 XATTR_QUOTA_FIELD(quota, max_bytes),
373 XATTR_QUOTA_FIELD(quota, max_files),
375 .name = "ceph.snap.btime",
376 .name_size = sizeof("ceph.snap.btime"),
377 .getxattr_cb = ceph_vxattrcb_snap_btime,
378 .exists_cb = ceph_vxattrcb_snap_btime_exists,
379 .flags = VXATTR_FLAG_READONLY,
381 { .name = NULL, 0 } /* Required table terminator */
386 static struct ceph_vxattr ceph_file_vxattrs[] = {
388 .name = "ceph.file.layout",
389 .name_size = sizeof("ceph.file.layout"),
390 .getxattr_cb = ceph_vxattrcb_layout,
391 .exists_cb = ceph_vxattrcb_layout_exists,
392 .flags = VXATTR_FLAG_HIDDEN,
394 XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
395 XATTR_LAYOUT_FIELD(file, layout, stripe_count),
396 XATTR_LAYOUT_FIELD(file, layout, object_size),
397 XATTR_LAYOUT_FIELD(file, layout, pool),
398 XATTR_LAYOUT_FIELD(file, layout, pool_namespace),
400 .name = "ceph.snap.btime",
401 .name_size = sizeof("ceph.snap.btime"),
402 .getxattr_cb = ceph_vxattrcb_snap_btime,
403 .exists_cb = ceph_vxattrcb_snap_btime_exists,
404 .flags = VXATTR_FLAG_READONLY,
406 { .name = NULL, 0 } /* Required table terminator */
409 static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
411 if (S_ISDIR(inode->i_mode))
412 return ceph_dir_vxattrs;
413 else if (S_ISREG(inode->i_mode))
414 return ceph_file_vxattrs;
418 static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
421 struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
424 while (vxattr->name) {
425 if (!strcmp(vxattr->name, name))
434 static int __set_xattr(struct ceph_inode_info *ci,
435 const char *name, int name_len,
436 const char *val, int val_len,
437 int flags, int update_xattr,
438 struct ceph_inode_xattr **newxattr)
441 struct rb_node *parent = NULL;
442 struct ceph_inode_xattr *xattr = NULL;
446 p = &ci->i_xattrs.index.rb_node;
449 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
450 c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
456 if (name_len == xattr->name_len)
458 else if (name_len < xattr->name_len)
469 if (xattr && (flags & XATTR_CREATE))
471 else if (!xattr && (flags & XATTR_REPLACE))
479 if (update_xattr < 0) {
481 __remove_xattr(ci, xattr);
492 xattr->name_len = name_len;
493 xattr->should_free_name = update_xattr;
495 ci->i_xattrs.count++;
496 dout("__set_xattr count=%d\n", ci->i_xattrs.count);
500 if (xattr->should_free_val)
507 ci->i_xattrs.names_size -= xattr->name_len;
508 ci->i_xattrs.vals_size -= xattr->val_len;
510 ci->i_xattrs.names_size += name_len;
511 ci->i_xattrs.vals_size += val_len;
517 xattr->val_len = val_len;
518 xattr->dirty = update_xattr;
519 xattr->should_free_val = (val && update_xattr);
522 rb_link_node(&xattr->node, parent, p);
523 rb_insert_color(&xattr->node, &ci->i_xattrs.index);
524 dout("__set_xattr_val p=%p\n", p);
527 dout("__set_xattr_val added %llx.%llx xattr %p %.*s=%.*s\n",
528 ceph_vinop(&ci->vfs_inode), xattr, name_len, name, val_len, val);
533 static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
537 struct rb_node *parent = NULL;
538 struct ceph_inode_xattr *xattr = NULL;
539 int name_len = strlen(name);
542 p = &ci->i_xattrs.index.rb_node;
545 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
546 c = strncmp(name, xattr->name, xattr->name_len);
547 if (c == 0 && name_len > xattr->name_len)
554 dout("__get_xattr %s: found %.*s\n", name,
555 xattr->val_len, xattr->val);
560 dout("__get_xattr %s: not found\n", name);
565 static void __free_xattr(struct ceph_inode_xattr *xattr)
569 if (xattr->should_free_name)
571 if (xattr->should_free_val)
577 static int __remove_xattr(struct ceph_inode_info *ci,
578 struct ceph_inode_xattr *xattr)
583 rb_erase(&xattr->node, &ci->i_xattrs.index);
585 if (xattr->should_free_name)
587 if (xattr->should_free_val)
590 ci->i_xattrs.names_size -= xattr->name_len;
591 ci->i_xattrs.vals_size -= xattr->val_len;
592 ci->i_xattrs.count--;
598 static char *__copy_xattr_names(struct ceph_inode_info *ci,
602 struct ceph_inode_xattr *xattr = NULL;
604 p = rb_first(&ci->i_xattrs.index);
605 dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
608 xattr = rb_entry(p, struct ceph_inode_xattr, node);
609 memcpy(dest, xattr->name, xattr->name_len);
610 dest[xattr->name_len] = '\0';
612 dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
613 xattr->name_len, ci->i_xattrs.names_size);
615 dest += xattr->name_len + 1;
622 void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
624 struct rb_node *p, *tmp;
625 struct ceph_inode_xattr *xattr = NULL;
627 p = rb_first(&ci->i_xattrs.index);
629 dout("__ceph_destroy_xattrs p=%p\n", p);
632 xattr = rb_entry(p, struct ceph_inode_xattr, node);
635 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
636 xattr->name_len, xattr->name);
637 rb_erase(tmp, &ci->i_xattrs.index);
642 ci->i_xattrs.names_size = 0;
643 ci->i_xattrs.vals_size = 0;
644 ci->i_xattrs.index_version = 0;
645 ci->i_xattrs.count = 0;
646 ci->i_xattrs.index = RB_ROOT;
649 static int __build_xattrs(struct inode *inode)
650 __releases(ci->i_ceph_lock)
651 __acquires(ci->i_ceph_lock)
657 const char *name, *val;
658 struct ceph_inode_info *ci = ceph_inode(inode);
660 struct ceph_inode_xattr **xattrs = NULL;
664 dout("__build_xattrs() len=%d\n",
665 ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
667 if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
668 return 0; /* already built */
670 __ceph_destroy_xattrs(ci);
673 /* updated internal xattr rb tree */
674 if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
675 p = ci->i_xattrs.blob->vec.iov_base;
676 end = p + ci->i_xattrs.blob->vec.iov_len;
677 ceph_decode_32_safe(&p, end, numattr, bad);
678 xattr_version = ci->i_xattrs.version;
679 spin_unlock(&ci->i_ceph_lock);
681 xattrs = kcalloc(numattr, sizeof(struct ceph_inode_xattr *),
687 for (i = 0; i < numattr; i++) {
688 xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
694 spin_lock(&ci->i_ceph_lock);
695 if (ci->i_xattrs.version != xattr_version) {
696 /* lost a race, retry */
697 for (i = 0; i < numattr; i++)
705 ceph_decode_32_safe(&p, end, len, bad);
709 ceph_decode_32_safe(&p, end, len, bad);
713 err = __set_xattr(ci, name, namelen, val, len,
714 0, 0, &xattrs[numattr]);
721 ci->i_xattrs.index_version = ci->i_xattrs.version;
722 ci->i_xattrs.dirty = false;
726 spin_lock(&ci->i_ceph_lock);
729 for (i = 0; i < numattr; i++)
733 ci->i_xattrs.names_size = 0;
737 static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
741 * 4 bytes for the length, and additional 4 bytes per each xattr name,
742 * 4 bytes per each value
744 int size = 4 + ci->i_xattrs.count*(4 + 4) +
745 ci->i_xattrs.names_size +
746 ci->i_xattrs.vals_size;
747 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
748 ci->i_xattrs.count, ci->i_xattrs.names_size,
749 ci->i_xattrs.vals_size);
752 size += 4 + 4 + name_size + val_size;
758 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
759 * and swap into place. It returns the old i_xattrs.blob (or NULL) so
760 * that it can be freed by the caller as the i_ceph_lock is likely to be
763 struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci)
766 struct ceph_inode_xattr *xattr = NULL;
767 struct ceph_buffer *old_blob = NULL;
770 dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
771 if (ci->i_xattrs.dirty) {
772 int need = __get_required_blob_size(ci, 0, 0);
774 BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
776 p = rb_first(&ci->i_xattrs.index);
777 dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
779 ceph_encode_32(&dest, ci->i_xattrs.count);
781 xattr = rb_entry(p, struct ceph_inode_xattr, node);
783 ceph_encode_32(&dest, xattr->name_len);
784 memcpy(dest, xattr->name, xattr->name_len);
785 dest += xattr->name_len;
786 ceph_encode_32(&dest, xattr->val_len);
787 memcpy(dest, xattr->val, xattr->val_len);
788 dest += xattr->val_len;
793 /* adjust buffer len; it may be larger than we need */
794 ci->i_xattrs.prealloc_blob->vec.iov_len =
795 dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
797 if (ci->i_xattrs.blob)
798 old_blob = ci->i_xattrs.blob;
799 ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
800 ci->i_xattrs.prealloc_blob = NULL;
801 ci->i_xattrs.dirty = false;
802 ci->i_xattrs.version++;
808 static inline int __get_request_mask(struct inode *in) {
809 struct ceph_mds_request *req = current->journal_info;
811 if (req && req->r_target_inode == in) {
812 if (req->r_op == CEPH_MDS_OP_LOOKUP ||
813 req->r_op == CEPH_MDS_OP_LOOKUPINO ||
814 req->r_op == CEPH_MDS_OP_LOOKUPPARENT ||
815 req->r_op == CEPH_MDS_OP_GETATTR) {
816 mask = le32_to_cpu(req->r_args.getattr.mask);
817 } else if (req->r_op == CEPH_MDS_OP_OPEN ||
818 req->r_op == CEPH_MDS_OP_CREATE) {
819 mask = le32_to_cpu(req->r_args.open.mask);
825 ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
828 struct ceph_inode_info *ci = ceph_inode(inode);
829 struct ceph_inode_xattr *xattr;
830 struct ceph_vxattr *vxattr = NULL;
834 /* let's see if a virtual xattr was requested */
835 vxattr = ceph_match_vxattr(inode, name);
838 if (vxattr->flags & VXATTR_FLAG_RSTAT)
839 mask |= CEPH_STAT_RSTAT;
840 err = ceph_do_getattr(inode, mask, true);
844 if (!(vxattr->exists_cb && !vxattr->exists_cb(ci))) {
845 err = vxattr->getxattr_cb(ci, value, size);
846 if (size && size < err)
852 req_mask = __get_request_mask(inode);
854 spin_lock(&ci->i_ceph_lock);
855 dout("getxattr %p name '%s' ver=%lld index_ver=%lld\n", inode, name,
856 ci->i_xattrs.version, ci->i_xattrs.index_version);
858 if (ci->i_xattrs.version == 0 ||
859 !((req_mask & CEPH_CAP_XATTR_SHARED) ||
860 __ceph_caps_issued_mask_metric(ci, CEPH_CAP_XATTR_SHARED, 1))) {
861 spin_unlock(&ci->i_ceph_lock);
863 /* security module gets xattr while filling trace */
864 if (current->journal_info) {
865 pr_warn_ratelimited("sync getxattr %p "
866 "during filling trace\n", inode);
870 /* get xattrs from mds (if we don't already have them) */
871 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
874 spin_lock(&ci->i_ceph_lock);
877 err = __build_xattrs(inode);
881 err = -ENODATA; /* == ENOATTR */
882 xattr = __get_xattr(ci, name);
887 if (size && size < xattr->val_len)
890 err = xattr->val_len;
894 memcpy(value, xattr->val, xattr->val_len);
896 if (current->journal_info &&
897 !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) &&
898 security_ismaclabel(name + XATTR_SECURITY_PREFIX_LEN))
899 ci->i_ceph_flags |= CEPH_I_SEC_INITED;
901 spin_unlock(&ci->i_ceph_lock);
905 ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
907 struct inode *inode = d_inode(dentry);
908 struct ceph_inode_info *ci = ceph_inode(inode);
909 bool len_only = (size == 0);
913 spin_lock(&ci->i_ceph_lock);
914 dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
915 ci->i_xattrs.version, ci->i_xattrs.index_version);
917 if (ci->i_xattrs.version == 0 ||
918 !__ceph_caps_issued_mask_metric(ci, CEPH_CAP_XATTR_SHARED, 1)) {
919 spin_unlock(&ci->i_ceph_lock);
920 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
923 spin_lock(&ci->i_ceph_lock);
926 err = __build_xattrs(inode);
930 /* add 1 byte for each xattr due to the null termination */
931 namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
933 if (namelen > size) {
937 names = __copy_xattr_names(ci, names);
942 spin_unlock(&ci->i_ceph_lock);
946 static int ceph_sync_setxattr(struct inode *inode, const char *name,
947 const char *value, size_t size, int flags)
949 struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
950 struct ceph_inode_info *ci = ceph_inode(inode);
951 struct ceph_mds_request *req;
952 struct ceph_mds_client *mdsc = fsc->mdsc;
953 struct ceph_pagelist *pagelist = NULL;
954 int op = CEPH_MDS_OP_SETXATTR;
958 /* copy value into pagelist */
959 pagelist = ceph_pagelist_alloc(GFP_NOFS);
963 err = ceph_pagelist_append(pagelist, value, size);
967 if (flags & CEPH_XATTR_REPLACE)
968 op = CEPH_MDS_OP_RMXATTR;
970 flags |= CEPH_XATTR_REMOVE;
973 dout("setxattr value=%.*s\n", (int)size, value);
976 req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
982 req->r_path2 = kstrdup(name, GFP_NOFS);
984 ceph_mdsc_put_request(req);
989 if (op == CEPH_MDS_OP_SETXATTR) {
990 req->r_args.setxattr.flags = cpu_to_le32(flags);
991 req->r_pagelist = pagelist;
995 req->r_inode = inode;
998 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
1000 dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
1001 err = ceph_mdsc_do_request(mdsc, NULL, req);
1002 ceph_mdsc_put_request(req);
1003 dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
1007 ceph_pagelist_release(pagelist);
1011 int __ceph_setxattr(struct inode *inode, const char *name,
1012 const void *value, size_t size, int flags)
1014 struct ceph_vxattr *vxattr;
1015 struct ceph_inode_info *ci = ceph_inode(inode);
1016 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
1017 struct ceph_cap_flush *prealloc_cf = NULL;
1018 struct ceph_buffer *old_blob = NULL;
1022 int name_len = strlen(name);
1024 char *newname = NULL;
1025 char *newval = NULL;
1026 struct ceph_inode_xattr *xattr = NULL;
1027 int required_blob_size;
1028 bool check_realm = false;
1029 bool lock_snap_rwsem = false;
1031 if (ceph_snap(inode) != CEPH_NOSNAP)
1034 vxattr = ceph_match_vxattr(inode, name);
1036 if (vxattr->flags & VXATTR_FLAG_READONLY)
1038 if (value && !strncmp(vxattr->name, "ceph.quota", 10))
1042 /* pass any unhandled ceph.* xattrs through to the MDS */
1043 if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
1044 goto do_sync_unlocked;
1046 /* preallocate memory for xattr name, value, index node */
1048 newname = kmemdup(name, name_len + 1, GFP_NOFS);
1053 newval = kmemdup(value, val_len, GFP_NOFS);
1058 xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
1062 prealloc_cf = ceph_alloc_cap_flush();
1066 spin_lock(&ci->i_ceph_lock);
1068 issued = __ceph_caps_issued(ci, NULL);
1069 if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
1072 if (!lock_snap_rwsem && !ci->i_head_snapc) {
1073 lock_snap_rwsem = true;
1074 if (!down_read_trylock(&mdsc->snap_rwsem)) {
1075 spin_unlock(&ci->i_ceph_lock);
1076 down_read(&mdsc->snap_rwsem);
1077 spin_lock(&ci->i_ceph_lock);
1082 dout("setxattr %p name '%s' issued %s\n", inode, name,
1083 ceph_cap_string(issued));
1084 __build_xattrs(inode);
1086 required_blob_size = __get_required_blob_size(ci, name_len, val_len);
1088 if (!ci->i_xattrs.prealloc_blob ||
1089 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1090 struct ceph_buffer *blob;
1092 spin_unlock(&ci->i_ceph_lock);
1093 ceph_buffer_put(old_blob); /* Shouldn't be required */
1094 dout(" pre-allocating new blob size=%d\n", required_blob_size);
1095 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1097 goto do_sync_unlocked;
1098 spin_lock(&ci->i_ceph_lock);
1099 /* prealloc_blob can't be released while holding i_ceph_lock */
1100 if (ci->i_xattrs.prealloc_blob)
1101 old_blob = ci->i_xattrs.prealloc_blob;
1102 ci->i_xattrs.prealloc_blob = blob;
1106 err = __set_xattr(ci, newname, name_len, newval, val_len,
1107 flags, value ? 1 : -1, &xattr);
1110 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL,
1112 ci->i_xattrs.dirty = true;
1113 inode->i_ctime = current_time(inode);
1116 spin_unlock(&ci->i_ceph_lock);
1117 ceph_buffer_put(old_blob);
1118 if (lock_snap_rwsem)
1119 up_read(&mdsc->snap_rwsem);
1121 __mark_inode_dirty(inode, dirty);
1122 ceph_free_cap_flush(prealloc_cf);
1126 spin_unlock(&ci->i_ceph_lock);
1128 if (lock_snap_rwsem)
1129 up_read(&mdsc->snap_rwsem);
1131 /* security module set xattr while filling trace */
1132 if (current->journal_info) {
1133 pr_warn_ratelimited("sync setxattr %p "
1134 "during filling trace\n", inode);
1137 err = ceph_sync_setxattr(inode, name, value, size, flags);
1138 if (err >= 0 && check_realm) {
1139 /* check if snaprealm was created for quota inode */
1140 spin_lock(&ci->i_ceph_lock);
1141 if ((ci->i_max_files || ci->i_max_bytes) &&
1142 !(ci->i_snap_realm &&
1143 ci->i_snap_realm->ino == ci->i_vino.ino))
1145 spin_unlock(&ci->i_ceph_lock);
1149 ceph_free_cap_flush(prealloc_cf);
1156 static int ceph_get_xattr_handler(const struct xattr_handler *handler,
1157 struct dentry *dentry, struct inode *inode,
1158 const char *name, void *value, size_t size)
1160 if (!ceph_is_valid_xattr(name))
1162 return __ceph_getxattr(inode, name, value, size);
1165 static int ceph_set_xattr_handler(const struct xattr_handler *handler,
1166 struct dentry *unused, struct inode *inode,
1167 const char *name, const void *value,
1168 size_t size, int flags)
1170 if (!ceph_is_valid_xattr(name))
1172 return __ceph_setxattr(inode, name, value, size, flags);
1175 static const struct xattr_handler ceph_other_xattr_handler = {
1176 .prefix = "", /* match any name => handlers called with full name */
1177 .get = ceph_get_xattr_handler,
1178 .set = ceph_set_xattr_handler,
1181 #ifdef CONFIG_SECURITY
1182 bool ceph_security_xattr_wanted(struct inode *in)
1184 return in->i_security != NULL;
1187 bool ceph_security_xattr_deadlock(struct inode *in)
1189 struct ceph_inode_info *ci;
1191 if (!in->i_security)
1193 ci = ceph_inode(in);
1194 spin_lock(&ci->i_ceph_lock);
1195 ret = !(ci->i_ceph_flags & CEPH_I_SEC_INITED) &&
1196 !(ci->i_xattrs.version > 0 &&
1197 __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 0));
1198 spin_unlock(&ci->i_ceph_lock);
1202 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1203 int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1204 struct ceph_acl_sec_ctx *as_ctx)
1206 struct ceph_pagelist *pagelist = as_ctx->pagelist;
1211 err = security_dentry_init_security(dentry, mode, &dentry->d_name,
1213 &as_ctx->sec_ctxlen);
1215 WARN_ON_ONCE(err != -EOPNOTSUPP);
1216 err = 0; /* do nothing */
1222 pagelist = ceph_pagelist_alloc(GFP_KERNEL);
1225 err = ceph_pagelist_reserve(pagelist, PAGE_SIZE);
1228 ceph_pagelist_encode_32(pagelist, 1);
1232 * FIXME: Make security_dentry_init_security() generic. Currently
1233 * It only supports single security module and only selinux has
1234 * dentry_init_security hook.
1236 name = XATTR_NAME_SELINUX;
1237 name_len = strlen(name);
1238 err = ceph_pagelist_reserve(pagelist,
1239 4 * 2 + name_len + as_ctx->sec_ctxlen);
1243 if (as_ctx->pagelist) {
1244 /* update count of KV pairs */
1245 BUG_ON(pagelist->length <= sizeof(__le32));
1246 if (list_is_singular(&pagelist->head)) {
1247 le32_add_cpu((__le32*)pagelist->mapped_tail, 1);
1249 struct page *page = list_first_entry(&pagelist->head,
1251 void *addr = kmap_atomic(page);
1252 le32_add_cpu((__le32*)addr, 1);
1253 kunmap_atomic(addr);
1256 as_ctx->pagelist = pagelist;
1259 ceph_pagelist_encode_32(pagelist, name_len);
1260 ceph_pagelist_append(pagelist, name, name_len);
1262 ceph_pagelist_encode_32(pagelist, as_ctx->sec_ctxlen);
1263 ceph_pagelist_append(pagelist, as_ctx->sec_ctx, as_ctx->sec_ctxlen);
1267 if (pagelist && !as_ctx->pagelist)
1268 ceph_pagelist_release(pagelist);
1271 #endif /* CONFIG_CEPH_FS_SECURITY_LABEL */
1272 #endif /* CONFIG_SECURITY */
1274 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx)
1276 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1277 posix_acl_release(as_ctx->acl);
1278 posix_acl_release(as_ctx->default_acl);
1280 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1281 security_release_secctx(as_ctx->sec_ctx, as_ctx->sec_ctxlen);
1283 if (as_ctx->pagelist)
1284 ceph_pagelist_release(as_ctx->pagelist);
1288 * List of handlers for synthetic system.* attributes. Other
1289 * attributes are handled directly.
1291 const struct xattr_handler *ceph_xattr_handlers[] = {
1292 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1293 &posix_acl_access_xattr_handler,
1294 &posix_acl_default_xattr_handler,
1296 &ceph_other_xattr_handler,