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 readonly
, hidden
;
54 bool (*exists_cb
)(struct ceph_inode_info
*ci
);
59 static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info
*ci
)
61 struct ceph_file_layout
*fl
= &ci
->i_layout
;
62 return (fl
->stripe_unit
> 0 || fl
->stripe_count
> 0 ||
63 fl
->object_size
> 0 || fl
->pool_id
>= 0 ||
64 rcu_dereference_raw(fl
->pool_ns
) != NULL
);
67 static size_t ceph_vxattrcb_layout(struct ceph_inode_info
*ci
, char *val
,
70 struct ceph_fs_client
*fsc
= ceph_sb_to_client(ci
->vfs_inode
.i_sb
);
71 struct ceph_osd_client
*osdc
= &fsc
->client
->osdc
;
72 struct ceph_string
*pool_ns
;
73 s64 pool
= ci
->i_layout
.pool_id
;
74 const char *pool_name
;
75 const char *ns_field
= " pool_namespace=";
77 size_t len
, total_len
= 0;
80 pool_ns
= ceph_try_get_string(ci
->i_layout
.pool_ns
);
82 dout("ceph_vxattrcb_layout %p\n", &ci
->vfs_inode
);
83 down_read(&osdc
->lock
);
84 pool_name
= ceph_pg_pool_name_by_id(osdc
->osdmap
, pool
);
86 len
= snprintf(buf
, sizeof(buf
),
87 "stripe_unit=%u stripe_count=%u object_size=%u pool=",
88 ci
->i_layout
.stripe_unit
, ci
->i_layout
.stripe_count
,
89 ci
->i_layout
.object_size
);
90 total_len
= len
+ strlen(pool_name
);
92 len
= snprintf(buf
, sizeof(buf
),
93 "stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
94 ci
->i_layout
.stripe_unit
, ci
->i_layout
.stripe_count
,
95 ci
->i_layout
.object_size
, (unsigned long long)pool
);
100 total_len
+= strlen(ns_field
) + pool_ns
->len
;
104 } else if (total_len
> size
) {
107 memcpy(val
, buf
, len
);
110 len
= strlen(pool_name
);
111 memcpy(val
+ ret
, pool_name
, len
);
115 len
= strlen(ns_field
);
116 memcpy(val
+ ret
, ns_field
, len
);
118 memcpy(val
+ ret
, pool_ns
->str
, pool_ns
->len
);
122 up_read(&osdc
->lock
);
123 ceph_put_string(pool_ns
);
127 static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info
*ci
,
128 char *val
, size_t size
)
130 return snprintf(val
, size
, "%u", ci
->i_layout
.stripe_unit
);
133 static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info
*ci
,
134 char *val
, size_t size
)
136 return snprintf(val
, size
, "%u", ci
->i_layout
.stripe_count
);
139 static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info
*ci
,
140 char *val
, size_t size
)
142 return snprintf(val
, size
, "%u", ci
->i_layout
.object_size
);
145 static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info
*ci
,
146 char *val
, size_t size
)
149 struct ceph_fs_client
*fsc
= ceph_sb_to_client(ci
->vfs_inode
.i_sb
);
150 struct ceph_osd_client
*osdc
= &fsc
->client
->osdc
;
151 s64 pool
= ci
->i_layout
.pool_id
;
152 const char *pool_name
;
154 down_read(&osdc
->lock
);
155 pool_name
= ceph_pg_pool_name_by_id(osdc
->osdmap
, pool
);
157 ret
= snprintf(val
, size
, "%s", pool_name
);
159 ret
= snprintf(val
, size
, "%lld", (unsigned long long)pool
);
160 up_read(&osdc
->lock
);
164 static size_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info
*ci
,
165 char *val
, size_t size
)
168 struct ceph_string
*ns
= ceph_try_get_string(ci
->i_layout
.pool_ns
);
170 ret
= snprintf(val
, size
, "%.*s", (int)ns
->len
, ns
->str
);
178 static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info
*ci
, char *val
,
181 return snprintf(val
, size
, "%lld", ci
->i_files
+ ci
->i_subdirs
);
184 static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info
*ci
, char *val
,
187 return snprintf(val
, size
, "%lld", ci
->i_files
);
190 static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info
*ci
, char *val
,
193 return snprintf(val
, size
, "%lld", ci
->i_subdirs
);
196 static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info
*ci
, char *val
,
199 return snprintf(val
, size
, "%lld", ci
->i_rfiles
+ ci
->i_rsubdirs
);
202 static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info
*ci
, char *val
,
205 return snprintf(val
, size
, "%lld", ci
->i_rfiles
);
208 static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info
*ci
, char *val
,
211 return snprintf(val
, size
, "%lld", ci
->i_rsubdirs
);
214 static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info
*ci
, char *val
,
217 return snprintf(val
, size
, "%lld", ci
->i_rbytes
);
220 static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info
*ci
, char *val
,
223 return snprintf(val
, size
, "%ld.09%ld", (long)ci
->i_rctime
.tv_sec
,
224 (long)ci
->i_rctime
.tv_nsec
);
229 static bool ceph_vxattrcb_quota_exists(struct ceph_inode_info
*ci
)
231 return (ci
->i_max_files
|| ci
->i_max_bytes
);
234 static size_t ceph_vxattrcb_quota(struct ceph_inode_info
*ci
, char *val
,
237 return snprintf(val
, size
, "max_bytes=%llu max_files=%llu",
238 ci
->i_max_bytes
, ci
->i_max_files
);
241 static size_t ceph_vxattrcb_quota_max_bytes(struct ceph_inode_info
*ci
,
242 char *val
, size_t size
)
244 return snprintf(val
, size
, "%llu", ci
->i_max_bytes
);
247 static size_t ceph_vxattrcb_quota_max_files(struct ceph_inode_info
*ci
,
248 char *val
, size_t size
)
250 return snprintf(val
, size
, "%llu", ci
->i_max_files
);
253 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
254 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
255 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
257 #define XATTR_NAME_CEPH(_type, _name) \
259 .name = CEPH_XATTR_NAME(_type, _name), \
260 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
261 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
266 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
268 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
269 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
270 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
273 .exists_cb = ceph_vxattrcb_layout_exists, \
275 #define XATTR_QUOTA_FIELD(_type, _name) \
277 .name = CEPH_XATTR_NAME(_type, _name), \
278 .name_size = sizeof(CEPH_XATTR_NAME(_type, _name)), \
279 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
282 .exists_cb = ceph_vxattrcb_quota_exists, \
285 static struct ceph_vxattr ceph_dir_vxattrs
[] = {
287 .name
= "ceph.dir.layout",
288 .name_size
= sizeof("ceph.dir.layout"),
289 .getxattr_cb
= ceph_vxattrcb_layout
,
292 .exists_cb
= ceph_vxattrcb_layout_exists
,
294 XATTR_LAYOUT_FIELD(dir
, layout
, stripe_unit
),
295 XATTR_LAYOUT_FIELD(dir
, layout
, stripe_count
),
296 XATTR_LAYOUT_FIELD(dir
, layout
, object_size
),
297 XATTR_LAYOUT_FIELD(dir
, layout
, pool
),
298 XATTR_LAYOUT_FIELD(dir
, layout
, pool_namespace
),
299 XATTR_NAME_CEPH(dir
, entries
),
300 XATTR_NAME_CEPH(dir
, files
),
301 XATTR_NAME_CEPH(dir
, subdirs
),
302 XATTR_NAME_CEPH(dir
, rentries
),
303 XATTR_NAME_CEPH(dir
, rfiles
),
304 XATTR_NAME_CEPH(dir
, rsubdirs
),
305 XATTR_NAME_CEPH(dir
, rbytes
),
306 XATTR_NAME_CEPH(dir
, rctime
),
308 .name
= "ceph.quota",
309 .name_size
= sizeof("ceph.quota"),
310 .getxattr_cb
= ceph_vxattrcb_quota
,
313 .exists_cb
= ceph_vxattrcb_quota_exists
,
315 XATTR_QUOTA_FIELD(quota
, max_bytes
),
316 XATTR_QUOTA_FIELD(quota
, max_files
),
317 { .name
= NULL
, 0 } /* Required table terminator */
319 static size_t ceph_dir_vxattrs_name_size
; /* total size of all names */
323 static struct ceph_vxattr ceph_file_vxattrs
[] = {
325 .name
= "ceph.file.layout",
326 .name_size
= sizeof("ceph.file.layout"),
327 .getxattr_cb
= ceph_vxattrcb_layout
,
330 .exists_cb
= ceph_vxattrcb_layout_exists
,
332 XATTR_LAYOUT_FIELD(file
, layout
, stripe_unit
),
333 XATTR_LAYOUT_FIELD(file
, layout
, stripe_count
),
334 XATTR_LAYOUT_FIELD(file
, layout
, object_size
),
335 XATTR_LAYOUT_FIELD(file
, layout
, pool
),
336 XATTR_LAYOUT_FIELD(file
, layout
, pool_namespace
),
337 { .name
= NULL
, 0 } /* Required table terminator */
339 static size_t ceph_file_vxattrs_name_size
; /* total size of all names */
341 static struct ceph_vxattr
*ceph_inode_vxattrs(struct inode
*inode
)
343 if (S_ISDIR(inode
->i_mode
))
344 return ceph_dir_vxattrs
;
345 else if (S_ISREG(inode
->i_mode
))
346 return ceph_file_vxattrs
;
350 static size_t ceph_vxattrs_name_size(struct ceph_vxattr
*vxattrs
)
352 if (vxattrs
== ceph_dir_vxattrs
)
353 return ceph_dir_vxattrs_name_size
;
354 if (vxattrs
== ceph_file_vxattrs
)
355 return ceph_file_vxattrs_name_size
;
361 * Compute the aggregate size (including terminating '\0') of all
362 * virtual extended attribute names in the given vxattr table.
364 static size_t __init
vxattrs_name_size(struct ceph_vxattr
*vxattrs
)
366 struct ceph_vxattr
*vxattr
;
369 for (vxattr
= vxattrs
; vxattr
->name
; vxattr
++)
371 size
+= vxattr
->name_size
;
376 /* Routines called at initialization and exit time */
378 void __init
ceph_xattr_init(void)
380 ceph_dir_vxattrs_name_size
= vxattrs_name_size(ceph_dir_vxattrs
);
381 ceph_file_vxattrs_name_size
= vxattrs_name_size(ceph_file_vxattrs
);
384 void ceph_xattr_exit(void)
386 ceph_dir_vxattrs_name_size
= 0;
387 ceph_file_vxattrs_name_size
= 0;
390 static struct ceph_vxattr
*ceph_match_vxattr(struct inode
*inode
,
393 struct ceph_vxattr
*vxattr
= ceph_inode_vxattrs(inode
);
396 while (vxattr
->name
) {
397 if (!strcmp(vxattr
->name
, name
))
406 static int __set_xattr(struct ceph_inode_info
*ci
,
407 const char *name
, int name_len
,
408 const char *val
, int val_len
,
409 int flags
, int update_xattr
,
410 struct ceph_inode_xattr
**newxattr
)
413 struct rb_node
*parent
= NULL
;
414 struct ceph_inode_xattr
*xattr
= NULL
;
418 p
= &ci
->i_xattrs
.index
.rb_node
;
421 xattr
= rb_entry(parent
, struct ceph_inode_xattr
, node
);
422 c
= strncmp(name
, xattr
->name
, min(name_len
, xattr
->name_len
));
428 if (name_len
== xattr
->name_len
)
430 else if (name_len
< xattr
->name_len
)
441 if (xattr
&& (flags
& XATTR_CREATE
))
443 else if (!xattr
&& (flags
& XATTR_REPLACE
))
451 if (update_xattr
< 0) {
453 __remove_xattr(ci
, xattr
);
464 xattr
->name_len
= name_len
;
465 xattr
->should_free_name
= update_xattr
;
467 ci
->i_xattrs
.count
++;
468 dout("__set_xattr count=%d\n", ci
->i_xattrs
.count
);
472 if (xattr
->should_free_val
)
473 kfree((void *)xattr
->val
);
479 ci
->i_xattrs
.names_size
-= xattr
->name_len
;
480 ci
->i_xattrs
.vals_size
-= xattr
->val_len
;
482 ci
->i_xattrs
.names_size
+= name_len
;
483 ci
->i_xattrs
.vals_size
+= val_len
;
489 xattr
->val_len
= val_len
;
490 xattr
->dirty
= update_xattr
;
491 xattr
->should_free_val
= (val
&& update_xattr
);
494 rb_link_node(&xattr
->node
, parent
, p
);
495 rb_insert_color(&xattr
->node
, &ci
->i_xattrs
.index
);
496 dout("__set_xattr_val p=%p\n", p
);
499 dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
500 ceph_vinop(&ci
->vfs_inode
), xattr
, name
, val_len
, val
);
505 static struct ceph_inode_xattr
*__get_xattr(struct ceph_inode_info
*ci
,
509 struct rb_node
*parent
= NULL
;
510 struct ceph_inode_xattr
*xattr
= NULL
;
511 int name_len
= strlen(name
);
514 p
= &ci
->i_xattrs
.index
.rb_node
;
517 xattr
= rb_entry(parent
, struct ceph_inode_xattr
, node
);
518 c
= strncmp(name
, xattr
->name
, xattr
->name_len
);
519 if (c
== 0 && name_len
> xattr
->name_len
)
526 dout("__get_xattr %s: found %.*s\n", name
,
527 xattr
->val_len
, xattr
->val
);
532 dout("__get_xattr %s: not found\n", name
);
537 static void __free_xattr(struct ceph_inode_xattr
*xattr
)
541 if (xattr
->should_free_name
)
542 kfree((void *)xattr
->name
);
543 if (xattr
->should_free_val
)
544 kfree((void *)xattr
->val
);
549 static int __remove_xattr(struct ceph_inode_info
*ci
,
550 struct ceph_inode_xattr
*xattr
)
555 rb_erase(&xattr
->node
, &ci
->i_xattrs
.index
);
557 if (xattr
->should_free_name
)
558 kfree((void *)xattr
->name
);
559 if (xattr
->should_free_val
)
560 kfree((void *)xattr
->val
);
562 ci
->i_xattrs
.names_size
-= xattr
->name_len
;
563 ci
->i_xattrs
.vals_size
-= xattr
->val_len
;
564 ci
->i_xattrs
.count
--;
570 static char *__copy_xattr_names(struct ceph_inode_info
*ci
,
574 struct ceph_inode_xattr
*xattr
= NULL
;
576 p
= rb_first(&ci
->i_xattrs
.index
);
577 dout("__copy_xattr_names count=%d\n", ci
->i_xattrs
.count
);
580 xattr
= rb_entry(p
, struct ceph_inode_xattr
, node
);
581 memcpy(dest
, xattr
->name
, xattr
->name_len
);
582 dest
[xattr
->name_len
] = '\0';
584 dout("dest=%s %p (%s) (%d/%d)\n", dest
, xattr
, xattr
->name
,
585 xattr
->name_len
, ci
->i_xattrs
.names_size
);
587 dest
+= xattr
->name_len
+ 1;
594 void __ceph_destroy_xattrs(struct ceph_inode_info
*ci
)
596 struct rb_node
*p
, *tmp
;
597 struct ceph_inode_xattr
*xattr
= NULL
;
599 p
= rb_first(&ci
->i_xattrs
.index
);
601 dout("__ceph_destroy_xattrs p=%p\n", p
);
604 xattr
= rb_entry(p
, struct ceph_inode_xattr
, node
);
607 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p
,
608 xattr
->name_len
, xattr
->name
);
609 rb_erase(tmp
, &ci
->i_xattrs
.index
);
614 ci
->i_xattrs
.names_size
= 0;
615 ci
->i_xattrs
.vals_size
= 0;
616 ci
->i_xattrs
.index_version
= 0;
617 ci
->i_xattrs
.count
= 0;
618 ci
->i_xattrs
.index
= RB_ROOT
;
621 static int __build_xattrs(struct inode
*inode
)
622 __releases(ci
->i_ceph_lock
)
623 __acquires(ci
->i_ceph_lock
)
629 const char *name
, *val
;
630 struct ceph_inode_info
*ci
= ceph_inode(inode
);
632 struct ceph_inode_xattr
**xattrs
= NULL
;
636 dout("__build_xattrs() len=%d\n",
637 ci
->i_xattrs
.blob
? (int)ci
->i_xattrs
.blob
->vec
.iov_len
: 0);
639 if (ci
->i_xattrs
.index_version
>= ci
->i_xattrs
.version
)
640 return 0; /* already built */
642 __ceph_destroy_xattrs(ci
);
645 /* updated internal xattr rb tree */
646 if (ci
->i_xattrs
.blob
&& ci
->i_xattrs
.blob
->vec
.iov_len
> 4) {
647 p
= ci
->i_xattrs
.blob
->vec
.iov_base
;
648 end
= p
+ ci
->i_xattrs
.blob
->vec
.iov_len
;
649 ceph_decode_32_safe(&p
, end
, numattr
, bad
);
650 xattr_version
= ci
->i_xattrs
.version
;
651 spin_unlock(&ci
->i_ceph_lock
);
653 xattrs
= kcalloc(numattr
, sizeof(struct ceph_inode_xattr
*),
659 for (i
= 0; i
< numattr
; i
++) {
660 xattrs
[i
] = kmalloc(sizeof(struct ceph_inode_xattr
),
666 spin_lock(&ci
->i_ceph_lock
);
667 if (ci
->i_xattrs
.version
!= xattr_version
) {
668 /* lost a race, retry */
669 for (i
= 0; i
< numattr
; i
++)
677 ceph_decode_32_safe(&p
, end
, len
, bad
);
681 ceph_decode_32_safe(&p
, end
, len
, bad
);
685 err
= __set_xattr(ci
, name
, namelen
, val
, len
,
686 0, 0, &xattrs
[numattr
]);
693 ci
->i_xattrs
.index_version
= ci
->i_xattrs
.version
;
694 ci
->i_xattrs
.dirty
= false;
698 spin_lock(&ci
->i_ceph_lock
);
701 for (i
= 0; i
< numattr
; i
++)
705 ci
->i_xattrs
.names_size
= 0;
709 static int __get_required_blob_size(struct ceph_inode_info
*ci
, int name_size
,
713 * 4 bytes for the length, and additional 4 bytes per each xattr name,
714 * 4 bytes per each value
716 int size
= 4 + ci
->i_xattrs
.count
*(4 + 4) +
717 ci
->i_xattrs
.names_size
+
718 ci
->i_xattrs
.vals_size
;
719 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
720 ci
->i_xattrs
.count
, ci
->i_xattrs
.names_size
,
721 ci
->i_xattrs
.vals_size
);
724 size
+= 4 + 4 + name_size
+ val_size
;
730 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
731 * and swap into place.
733 void __ceph_build_xattrs_blob(struct ceph_inode_info
*ci
)
736 struct ceph_inode_xattr
*xattr
= NULL
;
739 dout("__build_xattrs_blob %p\n", &ci
->vfs_inode
);
740 if (ci
->i_xattrs
.dirty
) {
741 int need
= __get_required_blob_size(ci
, 0, 0);
743 BUG_ON(need
> ci
->i_xattrs
.prealloc_blob
->alloc_len
);
745 p
= rb_first(&ci
->i_xattrs
.index
);
746 dest
= ci
->i_xattrs
.prealloc_blob
->vec
.iov_base
;
748 ceph_encode_32(&dest
, ci
->i_xattrs
.count
);
750 xattr
= rb_entry(p
, struct ceph_inode_xattr
, node
);
752 ceph_encode_32(&dest
, xattr
->name_len
);
753 memcpy(dest
, xattr
->name
, xattr
->name_len
);
754 dest
+= xattr
->name_len
;
755 ceph_encode_32(&dest
, xattr
->val_len
);
756 memcpy(dest
, xattr
->val
, xattr
->val_len
);
757 dest
+= xattr
->val_len
;
762 /* adjust buffer len; it may be larger than we need */
763 ci
->i_xattrs
.prealloc_blob
->vec
.iov_len
=
764 dest
- ci
->i_xattrs
.prealloc_blob
->vec
.iov_base
;
766 if (ci
->i_xattrs
.blob
)
767 ceph_buffer_put(ci
->i_xattrs
.blob
);
768 ci
->i_xattrs
.blob
= ci
->i_xattrs
.prealloc_blob
;
769 ci
->i_xattrs
.prealloc_blob
= NULL
;
770 ci
->i_xattrs
.dirty
= false;
771 ci
->i_xattrs
.version
++;
775 static inline int __get_request_mask(struct inode
*in
) {
776 struct ceph_mds_request
*req
= current
->journal_info
;
778 if (req
&& req
->r_target_inode
== in
) {
779 if (req
->r_op
== CEPH_MDS_OP_LOOKUP
||
780 req
->r_op
== CEPH_MDS_OP_LOOKUPINO
||
781 req
->r_op
== CEPH_MDS_OP_LOOKUPPARENT
||
782 req
->r_op
== CEPH_MDS_OP_GETATTR
) {
783 mask
= le32_to_cpu(req
->r_args
.getattr
.mask
);
784 } else if (req
->r_op
== CEPH_MDS_OP_OPEN
||
785 req
->r_op
== CEPH_MDS_OP_CREATE
) {
786 mask
= le32_to_cpu(req
->r_args
.open
.mask
);
792 ssize_t
__ceph_getxattr(struct inode
*inode
, const char *name
, void *value
,
795 struct ceph_inode_info
*ci
= ceph_inode(inode
);
796 struct ceph_inode_xattr
*xattr
;
797 struct ceph_vxattr
*vxattr
= NULL
;
801 /* let's see if a virtual xattr was requested */
802 vxattr
= ceph_match_vxattr(inode
, name
);
804 err
= ceph_do_getattr(inode
, 0, true);
808 if (!(vxattr
->exists_cb
&& !vxattr
->exists_cb(ci
)))
809 err
= vxattr
->getxattr_cb(ci
, value
, size
);
813 req_mask
= __get_request_mask(inode
);
815 spin_lock(&ci
->i_ceph_lock
);
816 dout("getxattr %p ver=%lld index_ver=%lld\n", inode
,
817 ci
->i_xattrs
.version
, ci
->i_xattrs
.index_version
);
819 if (ci
->i_xattrs
.version
== 0 ||
820 !((req_mask
& CEPH_CAP_XATTR_SHARED
) ||
821 __ceph_caps_issued_mask(ci
, CEPH_CAP_XATTR_SHARED
, 1))) {
822 spin_unlock(&ci
->i_ceph_lock
);
824 /* security module gets xattr while filling trace */
825 if (current
->journal_info
) {
826 pr_warn_ratelimited("sync getxattr %p "
827 "during filling trace\n", inode
);
831 /* get xattrs from mds (if we don't already have them) */
832 err
= ceph_do_getattr(inode
, CEPH_STAT_CAP_XATTR
, true);
835 spin_lock(&ci
->i_ceph_lock
);
838 err
= __build_xattrs(inode
);
842 err
= -ENODATA
; /* == ENOATTR */
843 xattr
= __get_xattr(ci
, name
);
848 if (size
&& size
< xattr
->val_len
)
851 err
= xattr
->val_len
;
855 memcpy(value
, xattr
->val
, xattr
->val_len
);
857 if (current
->journal_info
&&
858 !strncmp(name
, XATTR_SECURITY_PREFIX
, XATTR_SECURITY_PREFIX_LEN
))
859 ci
->i_ceph_flags
|= CEPH_I_SEC_INITED
;
861 spin_unlock(&ci
->i_ceph_lock
);
865 ssize_t
ceph_listxattr(struct dentry
*dentry
, char *names
, size_t size
)
867 struct inode
*inode
= d_inode(dentry
);
868 struct ceph_inode_info
*ci
= ceph_inode(inode
);
869 struct ceph_vxattr
*vxattrs
= ceph_inode_vxattrs(inode
);
876 spin_lock(&ci
->i_ceph_lock
);
877 dout("listxattr %p ver=%lld index_ver=%lld\n", inode
,
878 ci
->i_xattrs
.version
, ci
->i_xattrs
.index_version
);
880 if (ci
->i_xattrs
.version
== 0 ||
881 !__ceph_caps_issued_mask(ci
, CEPH_CAP_XATTR_SHARED
, 1)) {
882 spin_unlock(&ci
->i_ceph_lock
);
883 err
= ceph_do_getattr(inode
, CEPH_STAT_CAP_XATTR
, true);
886 spin_lock(&ci
->i_ceph_lock
);
889 err
= __build_xattrs(inode
);
893 * Start with virtual dir xattr names (if any) (including
894 * terminating '\0' characters for each).
896 vir_namelen
= ceph_vxattrs_name_size(vxattrs
);
898 /* adding 1 byte per each variable due to the null termination */
899 namelen
= ci
->i_xattrs
.names_size
+ ci
->i_xattrs
.count
;
901 if (size
&& vir_namelen
+ namelen
> size
)
904 err
= namelen
+ vir_namelen
;
908 names
= __copy_xattr_names(ci
, names
);
910 /* virtual xattr names, too */
913 for (i
= 0; vxattrs
[i
].name
; i
++) {
914 if (!vxattrs
[i
].hidden
&&
915 !(vxattrs
[i
].exists_cb
&&
916 !vxattrs
[i
].exists_cb(ci
))) {
917 len
= sprintf(names
, "%s", vxattrs
[i
].name
);
925 spin_unlock(&ci
->i_ceph_lock
);
929 static int ceph_sync_setxattr(struct inode
*inode
, const char *name
,
930 const char *value
, size_t size
, int flags
)
932 struct ceph_fs_client
*fsc
= ceph_sb_to_client(inode
->i_sb
);
933 struct ceph_inode_info
*ci
= ceph_inode(inode
);
934 struct ceph_mds_request
*req
;
935 struct ceph_mds_client
*mdsc
= fsc
->mdsc
;
936 struct ceph_pagelist
*pagelist
= NULL
;
937 int op
= CEPH_MDS_OP_SETXATTR
;
941 /* copy value into pagelist */
942 pagelist
= kmalloc(sizeof(*pagelist
), GFP_NOFS
);
946 ceph_pagelist_init(pagelist
);
947 err
= ceph_pagelist_append(pagelist
, value
, size
);
951 if (flags
& CEPH_XATTR_REPLACE
)
952 op
= CEPH_MDS_OP_RMXATTR
;
954 flags
|= CEPH_XATTR_REMOVE
;
957 dout("setxattr value=%.*s\n", (int)size
, value
);
960 req
= ceph_mdsc_create_request(mdsc
, op
, USE_AUTH_MDS
);
966 req
->r_path2
= kstrdup(name
, GFP_NOFS
);
968 ceph_mdsc_put_request(req
);
973 if (op
== CEPH_MDS_OP_SETXATTR
) {
974 req
->r_args
.setxattr
.flags
= cpu_to_le32(flags
);
975 req
->r_pagelist
= pagelist
;
979 req
->r_inode
= inode
;
982 req
->r_inode_drop
= CEPH_CAP_XATTR_SHARED
;
984 dout("xattr.ver (before): %lld\n", ci
->i_xattrs
.version
);
985 err
= ceph_mdsc_do_request(mdsc
, NULL
, req
);
986 ceph_mdsc_put_request(req
);
987 dout("xattr.ver (after): %lld\n", ci
->i_xattrs
.version
);
991 ceph_pagelist_release(pagelist
);
995 int __ceph_setxattr(struct inode
*inode
, const char *name
,
996 const void *value
, size_t size
, int flags
)
998 struct ceph_vxattr
*vxattr
;
999 struct ceph_inode_info
*ci
= ceph_inode(inode
);
1000 struct ceph_mds_client
*mdsc
= ceph_sb_to_client(inode
->i_sb
)->mdsc
;
1001 struct ceph_cap_flush
*prealloc_cf
= NULL
;
1005 int name_len
= strlen(name
);
1007 char *newname
= NULL
;
1008 char *newval
= NULL
;
1009 struct ceph_inode_xattr
*xattr
= NULL
;
1010 int required_blob_size
;
1011 bool lock_snap_rwsem
= false;
1013 if (ceph_snap(inode
) != CEPH_NOSNAP
)
1016 vxattr
= ceph_match_vxattr(inode
, name
);
1017 if (vxattr
&& vxattr
->readonly
)
1020 /* pass any unhandled ceph.* xattrs through to the MDS */
1021 if (!strncmp(name
, XATTR_CEPH_PREFIX
, XATTR_CEPH_PREFIX_LEN
))
1022 goto do_sync_unlocked
;
1024 /* preallocate memory for xattr name, value, index node */
1026 newname
= kmemdup(name
, name_len
+ 1, GFP_NOFS
);
1031 newval
= kmemdup(value
, val_len
, GFP_NOFS
);
1036 xattr
= kmalloc(sizeof(struct ceph_inode_xattr
), GFP_NOFS
);
1040 prealloc_cf
= ceph_alloc_cap_flush();
1044 spin_lock(&ci
->i_ceph_lock
);
1046 issued
= __ceph_caps_issued(ci
, NULL
);
1047 if (ci
->i_xattrs
.version
== 0 || !(issued
& CEPH_CAP_XATTR_EXCL
))
1050 if (!lock_snap_rwsem
&& !ci
->i_head_snapc
) {
1051 lock_snap_rwsem
= true;
1052 if (!down_read_trylock(&mdsc
->snap_rwsem
)) {
1053 spin_unlock(&ci
->i_ceph_lock
);
1054 down_read(&mdsc
->snap_rwsem
);
1055 spin_lock(&ci
->i_ceph_lock
);
1060 dout("setxattr %p issued %s\n", inode
, ceph_cap_string(issued
));
1061 __build_xattrs(inode
);
1063 required_blob_size
= __get_required_blob_size(ci
, name_len
, val_len
);
1065 if (!ci
->i_xattrs
.prealloc_blob
||
1066 required_blob_size
> ci
->i_xattrs
.prealloc_blob
->alloc_len
) {
1067 struct ceph_buffer
*blob
;
1069 spin_unlock(&ci
->i_ceph_lock
);
1070 dout(" preaallocating new blob size=%d\n", required_blob_size
);
1071 blob
= ceph_buffer_new(required_blob_size
, GFP_NOFS
);
1073 goto do_sync_unlocked
;
1074 spin_lock(&ci
->i_ceph_lock
);
1075 if (ci
->i_xattrs
.prealloc_blob
)
1076 ceph_buffer_put(ci
->i_xattrs
.prealloc_blob
);
1077 ci
->i_xattrs
.prealloc_blob
= blob
;
1081 err
= __set_xattr(ci
, newname
, name_len
, newval
, val_len
,
1082 flags
, value
? 1 : -1, &xattr
);
1085 dirty
= __ceph_mark_dirty_caps(ci
, CEPH_CAP_XATTR_EXCL
,
1087 ci
->i_xattrs
.dirty
= true;
1088 inode
->i_ctime
= current_time(inode
);
1091 spin_unlock(&ci
->i_ceph_lock
);
1092 if (lock_snap_rwsem
)
1093 up_read(&mdsc
->snap_rwsem
);
1095 __mark_inode_dirty(inode
, dirty
);
1096 ceph_free_cap_flush(prealloc_cf
);
1100 spin_unlock(&ci
->i_ceph_lock
);
1102 if (lock_snap_rwsem
)
1103 up_read(&mdsc
->snap_rwsem
);
1105 /* security module set xattr while filling trace */
1106 if (current
->journal_info
) {
1107 pr_warn_ratelimited("sync setxattr %p "
1108 "during filling trace\n", inode
);
1111 err
= ceph_sync_setxattr(inode
, name
, value
, size
, flags
);
1114 ceph_free_cap_flush(prealloc_cf
);
1121 static int ceph_get_xattr_handler(const struct xattr_handler
*handler
,
1122 struct dentry
*dentry
, struct inode
*inode
,
1123 const char *name
, void *value
, size_t size
)
1125 if (!ceph_is_valid_xattr(name
))
1127 return __ceph_getxattr(inode
, name
, value
, size
);
1130 static int ceph_set_xattr_handler(const struct xattr_handler
*handler
,
1131 struct dentry
*unused
, struct inode
*inode
,
1132 const char *name
, const void *value
,
1133 size_t size
, int flags
)
1135 if (!ceph_is_valid_xattr(name
))
1137 return __ceph_setxattr(inode
, name
, value
, size
, flags
);
1140 static const struct xattr_handler ceph_other_xattr_handler
= {
1141 .prefix
= "", /* match any name => handlers called with full name */
1142 .get
= ceph_get_xattr_handler
,
1143 .set
= ceph_set_xattr_handler
,
1146 #ifdef CONFIG_SECURITY
1147 bool ceph_security_xattr_wanted(struct inode
*in
)
1149 return in
->i_security
!= NULL
;
1152 bool ceph_security_xattr_deadlock(struct inode
*in
)
1154 struct ceph_inode_info
*ci
;
1156 if (!in
->i_security
)
1158 ci
= ceph_inode(in
);
1159 spin_lock(&ci
->i_ceph_lock
);
1160 ret
= !(ci
->i_ceph_flags
& CEPH_I_SEC_INITED
) &&
1161 !(ci
->i_xattrs
.version
> 0 &&
1162 __ceph_caps_issued_mask(ci
, CEPH_CAP_XATTR_SHARED
, 0));
1163 spin_unlock(&ci
->i_ceph_lock
);