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 int ceph_fmt_xattr(char *val
, size_t size
, const char *fmt
, ...)
123 char buf
[96]; /* NB: reevaluate size if new vxattrs are added */
126 ret
= vsnprintf(buf
, size
? sizeof(buf
) : 0, fmt
, args
);
130 if (size
&& ret
+ 1 > sizeof(buf
)) {
131 WARN_ONCE(true, "Returned length too big (%d)", ret
);
136 memcpy(val
, buf
, ret
);
140 static ssize_t
ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info
*ci
,
141 char *val
, size_t size
)
143 return ceph_fmt_xattr(val
, size
, "%u", ci
->i_layout
.stripe_unit
);
146 static ssize_t
ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info
*ci
,
147 char *val
, size_t size
)
149 return ceph_fmt_xattr(val
, size
, "%u", ci
->i_layout
.stripe_count
);
152 static ssize_t
ceph_vxattrcb_layout_object_size(struct ceph_inode_info
*ci
,
153 char *val
, size_t size
)
155 return ceph_fmt_xattr(val
, size
, "%u", ci
->i_layout
.object_size
);
158 static ssize_t
ceph_vxattrcb_layout_pool(struct ceph_inode_info
*ci
,
159 char *val
, size_t size
)
162 struct ceph_fs_client
*fsc
= ceph_sb_to_client(ci
->vfs_inode
.i_sb
);
163 struct ceph_osd_client
*osdc
= &fsc
->client
->osdc
;
164 s64 pool
= ci
->i_layout
.pool_id
;
165 const char *pool_name
;
167 down_read(&osdc
->lock
);
168 pool_name
= ceph_pg_pool_name_by_id(osdc
->osdmap
, pool
);
170 ret
= strlen(pool_name
);
172 memcpy(val
, pool_name
, ret
);
174 ret
= ceph_fmt_xattr(val
, size
, "%lld", pool
);
176 up_read(&osdc
->lock
);
180 static ssize_t
ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info
*ci
,
181 char *val
, size_t size
)
184 struct ceph_string
*ns
= ceph_try_get_string(ci
->i_layout
.pool_ns
);
189 memcpy(val
, ns
->str
, ret
);
197 static ssize_t
ceph_vxattrcb_dir_entries(struct ceph_inode_info
*ci
, char *val
,
200 return ceph_fmt_xattr(val
, size
, "%lld", ci
->i_files
+ ci
->i_subdirs
);
203 static ssize_t
ceph_vxattrcb_dir_files(struct ceph_inode_info
*ci
, char *val
,
206 return ceph_fmt_xattr(val
, size
, "%lld", ci
->i_files
);
209 static ssize_t
ceph_vxattrcb_dir_subdirs(struct ceph_inode_info
*ci
, char *val
,
212 return ceph_fmt_xattr(val
, size
, "%lld", ci
->i_subdirs
);
215 static ssize_t
ceph_vxattrcb_dir_rentries(struct ceph_inode_info
*ci
, char *val
,
218 return ceph_fmt_xattr(val
, size
, "%lld",
219 ci
->i_rfiles
+ ci
->i_rsubdirs
);
222 static ssize_t
ceph_vxattrcb_dir_rfiles(struct ceph_inode_info
*ci
, char *val
,
225 return ceph_fmt_xattr(val
, size
, "%lld", ci
->i_rfiles
);
228 static ssize_t
ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info
*ci
, char *val
,
231 return ceph_fmt_xattr(val
, size
, "%lld", ci
->i_rsubdirs
);
234 static ssize_t
ceph_vxattrcb_dir_rbytes(struct ceph_inode_info
*ci
, char *val
,
237 return ceph_fmt_xattr(val
, size
, "%lld", ci
->i_rbytes
);
240 static ssize_t
ceph_vxattrcb_dir_rctime(struct ceph_inode_info
*ci
, char *val
,
243 return ceph_fmt_xattr(val
, size
, "%lld.%09ld", ci
->i_rctime
.tv_sec
,
244 ci
->i_rctime
.tv_nsec
);
248 static bool ceph_vxattrcb_dir_pin_exists(struct ceph_inode_info
*ci
)
250 return ci
->i_dir_pin
!= -ENODATA
;
253 static ssize_t
ceph_vxattrcb_dir_pin(struct ceph_inode_info
*ci
, char *val
,
256 return ceph_fmt_xattr(val
, size
, "%d", (int)ci
->i_dir_pin
);
260 static bool ceph_vxattrcb_quota_exists(struct ceph_inode_info
*ci
)
263 spin_lock(&ci
->i_ceph_lock
);
264 if ((ci
->i_max_files
|| ci
->i_max_bytes
) &&
265 ci
->i_vino
.snap
== CEPH_NOSNAP
&&
267 ci
->i_snap_realm
->ino
== ci
->i_vino
.ino
)
269 spin_unlock(&ci
->i_ceph_lock
);
273 static ssize_t
ceph_vxattrcb_quota(struct ceph_inode_info
*ci
, char *val
,
276 return ceph_fmt_xattr(val
, size
, "max_bytes=%llu max_files=%llu",
277 ci
->i_max_bytes
, ci
->i_max_files
);
280 static ssize_t
ceph_vxattrcb_quota_max_bytes(struct ceph_inode_info
*ci
,
281 char *val
, size_t size
)
283 return ceph_fmt_xattr(val
, size
, "%llu", ci
->i_max_bytes
);
286 static ssize_t
ceph_vxattrcb_quota_max_files(struct ceph_inode_info
*ci
,
287 char *val
, size_t size
)
289 return ceph_fmt_xattr(val
, size
, "%llu", ci
->i_max_files
);
293 static bool ceph_vxattrcb_snap_btime_exists(struct ceph_inode_info
*ci
)
295 return (ci
->i_snap_btime
.tv_sec
!= 0 || ci
->i_snap_btime
.tv_nsec
!= 0);
298 static ssize_t
ceph_vxattrcb_snap_btime(struct ceph_inode_info
*ci
, char *val
,
301 return ceph_fmt_xattr(val
, size
, "%lld.%09ld", ci
->i_snap_btime
.tv_sec
,
302 ci
->i_snap_btime
.tv_nsec
);
305 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
306 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
307 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
309 #define XATTR_NAME_CEPH(_type, _name, _flags) \
311 .name = CEPH_XATTR_NAME(_type, _name), \
312 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
313 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
315 .flags = (VXATTR_FLAG_READONLY | _flags), \
317 #define XATTR_RSTAT_FIELD(_type, _name) \
318 XATTR_NAME_CEPH(_type, _name, VXATTR_FLAG_RSTAT)
319 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
321 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
322 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
323 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
324 .exists_cb = ceph_vxattrcb_layout_exists, \
325 .flags = VXATTR_FLAG_HIDDEN, \
327 #define XATTR_QUOTA_FIELD(_type, _name) \
329 .name = CEPH_XATTR_NAME(_type, _name), \
330 .name_size = sizeof(CEPH_XATTR_NAME(_type, _name)), \
331 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
332 .exists_cb = ceph_vxattrcb_quota_exists, \
333 .flags = VXATTR_FLAG_HIDDEN, \
336 static struct ceph_vxattr ceph_dir_vxattrs
[] = {
338 .name
= "ceph.dir.layout",
339 .name_size
= sizeof("ceph.dir.layout"),
340 .getxattr_cb
= ceph_vxattrcb_layout
,
341 .exists_cb
= ceph_vxattrcb_layout_exists
,
342 .flags
= VXATTR_FLAG_HIDDEN
,
344 XATTR_LAYOUT_FIELD(dir
, layout
, stripe_unit
),
345 XATTR_LAYOUT_FIELD(dir
, layout
, stripe_count
),
346 XATTR_LAYOUT_FIELD(dir
, layout
, object_size
),
347 XATTR_LAYOUT_FIELD(dir
, layout
, pool
),
348 XATTR_LAYOUT_FIELD(dir
, layout
, pool_namespace
),
349 XATTR_NAME_CEPH(dir
, entries
, 0),
350 XATTR_NAME_CEPH(dir
, files
, 0),
351 XATTR_NAME_CEPH(dir
, subdirs
, 0),
352 XATTR_RSTAT_FIELD(dir
, rentries
),
353 XATTR_RSTAT_FIELD(dir
, rfiles
),
354 XATTR_RSTAT_FIELD(dir
, rsubdirs
),
355 XATTR_RSTAT_FIELD(dir
, rbytes
),
356 XATTR_RSTAT_FIELD(dir
, rctime
),
358 .name
= "ceph.dir.pin",
359 .name_size
= sizeof("ceph.dir.pin"),
360 .getxattr_cb
= ceph_vxattrcb_dir_pin
,
361 .exists_cb
= ceph_vxattrcb_dir_pin_exists
,
362 .flags
= VXATTR_FLAG_HIDDEN
,
365 .name
= "ceph.quota",
366 .name_size
= sizeof("ceph.quota"),
367 .getxattr_cb
= ceph_vxattrcb_quota
,
368 .exists_cb
= ceph_vxattrcb_quota_exists
,
369 .flags
= VXATTR_FLAG_HIDDEN
,
371 XATTR_QUOTA_FIELD(quota
, max_bytes
),
372 XATTR_QUOTA_FIELD(quota
, max_files
),
374 .name
= "ceph.snap.btime",
375 .name_size
= sizeof("ceph.snap.btime"),
376 .getxattr_cb
= ceph_vxattrcb_snap_btime
,
377 .exists_cb
= ceph_vxattrcb_snap_btime_exists
,
378 .flags
= VXATTR_FLAG_READONLY
,
380 { .name
= NULL
, 0 } /* Required table terminator */
385 static struct ceph_vxattr ceph_file_vxattrs
[] = {
387 .name
= "ceph.file.layout",
388 .name_size
= sizeof("ceph.file.layout"),
389 .getxattr_cb
= ceph_vxattrcb_layout
,
390 .exists_cb
= ceph_vxattrcb_layout_exists
,
391 .flags
= VXATTR_FLAG_HIDDEN
,
393 XATTR_LAYOUT_FIELD(file
, layout
, stripe_unit
),
394 XATTR_LAYOUT_FIELD(file
, layout
, stripe_count
),
395 XATTR_LAYOUT_FIELD(file
, layout
, object_size
),
396 XATTR_LAYOUT_FIELD(file
, layout
, pool
),
397 XATTR_LAYOUT_FIELD(file
, layout
, pool_namespace
),
399 .name
= "ceph.snap.btime",
400 .name_size
= sizeof("ceph.snap.btime"),
401 .getxattr_cb
= ceph_vxattrcb_snap_btime
,
402 .exists_cb
= ceph_vxattrcb_snap_btime_exists
,
403 .flags
= VXATTR_FLAG_READONLY
,
405 { .name
= NULL
, 0 } /* Required table terminator */
408 static struct ceph_vxattr
*ceph_inode_vxattrs(struct inode
*inode
)
410 if (S_ISDIR(inode
->i_mode
))
411 return ceph_dir_vxattrs
;
412 else if (S_ISREG(inode
->i_mode
))
413 return ceph_file_vxattrs
;
417 static struct ceph_vxattr
*ceph_match_vxattr(struct inode
*inode
,
420 struct ceph_vxattr
*vxattr
= ceph_inode_vxattrs(inode
);
423 while (vxattr
->name
) {
424 if (!strcmp(vxattr
->name
, name
))
433 static int __set_xattr(struct ceph_inode_info
*ci
,
434 const char *name
, int name_len
,
435 const char *val
, int val_len
,
436 int flags
, int update_xattr
,
437 struct ceph_inode_xattr
**newxattr
)
440 struct rb_node
*parent
= NULL
;
441 struct ceph_inode_xattr
*xattr
= NULL
;
445 p
= &ci
->i_xattrs
.index
.rb_node
;
448 xattr
= rb_entry(parent
, struct ceph_inode_xattr
, node
);
449 c
= strncmp(name
, xattr
->name
, min(name_len
, xattr
->name_len
));
455 if (name_len
== xattr
->name_len
)
457 else if (name_len
< xattr
->name_len
)
468 if (xattr
&& (flags
& XATTR_CREATE
))
470 else if (!xattr
&& (flags
& XATTR_REPLACE
))
478 if (update_xattr
< 0) {
480 __remove_xattr(ci
, xattr
);
491 xattr
->name_len
= name_len
;
492 xattr
->should_free_name
= update_xattr
;
494 ci
->i_xattrs
.count
++;
495 dout("__set_xattr count=%d\n", ci
->i_xattrs
.count
);
499 if (xattr
->should_free_val
)
500 kfree((void *)xattr
->val
);
506 ci
->i_xattrs
.names_size
-= xattr
->name_len
;
507 ci
->i_xattrs
.vals_size
-= xattr
->val_len
;
509 ci
->i_xattrs
.names_size
+= name_len
;
510 ci
->i_xattrs
.vals_size
+= val_len
;
516 xattr
->val_len
= val_len
;
517 xattr
->dirty
= update_xattr
;
518 xattr
->should_free_val
= (val
&& update_xattr
);
521 rb_link_node(&xattr
->node
, parent
, p
);
522 rb_insert_color(&xattr
->node
, &ci
->i_xattrs
.index
);
523 dout("__set_xattr_val p=%p\n", p
);
526 dout("__set_xattr_val added %llx.%llx xattr %p %.*s=%.*s\n",
527 ceph_vinop(&ci
->vfs_inode
), xattr
, name_len
, name
, val_len
, val
);
532 static struct ceph_inode_xattr
*__get_xattr(struct ceph_inode_info
*ci
,
536 struct rb_node
*parent
= NULL
;
537 struct ceph_inode_xattr
*xattr
= NULL
;
538 int name_len
= strlen(name
);
541 p
= &ci
->i_xattrs
.index
.rb_node
;
544 xattr
= rb_entry(parent
, struct ceph_inode_xattr
, node
);
545 c
= strncmp(name
, xattr
->name
, xattr
->name_len
);
546 if (c
== 0 && name_len
> xattr
->name_len
)
553 dout("__get_xattr %s: found %.*s\n", name
,
554 xattr
->val_len
, xattr
->val
);
559 dout("__get_xattr %s: not found\n", name
);
564 static void __free_xattr(struct ceph_inode_xattr
*xattr
)
568 if (xattr
->should_free_name
)
569 kfree((void *)xattr
->name
);
570 if (xattr
->should_free_val
)
571 kfree((void *)xattr
->val
);
576 static int __remove_xattr(struct ceph_inode_info
*ci
,
577 struct ceph_inode_xattr
*xattr
)
582 rb_erase(&xattr
->node
, &ci
->i_xattrs
.index
);
584 if (xattr
->should_free_name
)
585 kfree((void *)xattr
->name
);
586 if (xattr
->should_free_val
)
587 kfree((void *)xattr
->val
);
589 ci
->i_xattrs
.names_size
-= xattr
->name_len
;
590 ci
->i_xattrs
.vals_size
-= xattr
->val_len
;
591 ci
->i_xattrs
.count
--;
597 static char *__copy_xattr_names(struct ceph_inode_info
*ci
,
601 struct ceph_inode_xattr
*xattr
= NULL
;
603 p
= rb_first(&ci
->i_xattrs
.index
);
604 dout("__copy_xattr_names count=%d\n", ci
->i_xattrs
.count
);
607 xattr
= rb_entry(p
, struct ceph_inode_xattr
, node
);
608 memcpy(dest
, xattr
->name
, xattr
->name_len
);
609 dest
[xattr
->name_len
] = '\0';
611 dout("dest=%s %p (%s) (%d/%d)\n", dest
, xattr
, xattr
->name
,
612 xattr
->name_len
, ci
->i_xattrs
.names_size
);
614 dest
+= xattr
->name_len
+ 1;
621 void __ceph_destroy_xattrs(struct ceph_inode_info
*ci
)
623 struct rb_node
*p
, *tmp
;
624 struct ceph_inode_xattr
*xattr
= NULL
;
626 p
= rb_first(&ci
->i_xattrs
.index
);
628 dout("__ceph_destroy_xattrs p=%p\n", p
);
631 xattr
= rb_entry(p
, struct ceph_inode_xattr
, node
);
634 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p
,
635 xattr
->name_len
, xattr
->name
);
636 rb_erase(tmp
, &ci
->i_xattrs
.index
);
641 ci
->i_xattrs
.names_size
= 0;
642 ci
->i_xattrs
.vals_size
= 0;
643 ci
->i_xattrs
.index_version
= 0;
644 ci
->i_xattrs
.count
= 0;
645 ci
->i_xattrs
.index
= RB_ROOT
;
648 static int __build_xattrs(struct inode
*inode
)
649 __releases(ci
->i_ceph_lock
)
650 __acquires(ci
->i_ceph_lock
)
656 const char *name
, *val
;
657 struct ceph_inode_info
*ci
= ceph_inode(inode
);
659 struct ceph_inode_xattr
**xattrs
= NULL
;
663 dout("__build_xattrs() len=%d\n",
664 ci
->i_xattrs
.blob
? (int)ci
->i_xattrs
.blob
->vec
.iov_len
: 0);
666 if (ci
->i_xattrs
.index_version
>= ci
->i_xattrs
.version
)
667 return 0; /* already built */
669 __ceph_destroy_xattrs(ci
);
672 /* updated internal xattr rb tree */
673 if (ci
->i_xattrs
.blob
&& ci
->i_xattrs
.blob
->vec
.iov_len
> 4) {
674 p
= ci
->i_xattrs
.blob
->vec
.iov_base
;
675 end
= p
+ ci
->i_xattrs
.blob
->vec
.iov_len
;
676 ceph_decode_32_safe(&p
, end
, numattr
, bad
);
677 xattr_version
= ci
->i_xattrs
.version
;
678 spin_unlock(&ci
->i_ceph_lock
);
680 xattrs
= kcalloc(numattr
, sizeof(struct ceph_inode_xattr
*),
686 for (i
= 0; i
< numattr
; i
++) {
687 xattrs
[i
] = kmalloc(sizeof(struct ceph_inode_xattr
),
693 spin_lock(&ci
->i_ceph_lock
);
694 if (ci
->i_xattrs
.version
!= xattr_version
) {
695 /* lost a race, retry */
696 for (i
= 0; i
< numattr
; i
++)
704 ceph_decode_32_safe(&p
, end
, len
, bad
);
708 ceph_decode_32_safe(&p
, end
, len
, bad
);
712 err
= __set_xattr(ci
, name
, namelen
, val
, len
,
713 0, 0, &xattrs
[numattr
]);
720 ci
->i_xattrs
.index_version
= ci
->i_xattrs
.version
;
721 ci
->i_xattrs
.dirty
= false;
725 spin_lock(&ci
->i_ceph_lock
);
728 for (i
= 0; i
< numattr
; i
++)
732 ci
->i_xattrs
.names_size
= 0;
736 static int __get_required_blob_size(struct ceph_inode_info
*ci
, int name_size
,
740 * 4 bytes for the length, and additional 4 bytes per each xattr name,
741 * 4 bytes per each value
743 int size
= 4 + ci
->i_xattrs
.count
*(4 + 4) +
744 ci
->i_xattrs
.names_size
+
745 ci
->i_xattrs
.vals_size
;
746 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
747 ci
->i_xattrs
.count
, ci
->i_xattrs
.names_size
,
748 ci
->i_xattrs
.vals_size
);
751 size
+= 4 + 4 + name_size
+ val_size
;
757 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
758 * and swap into place. It returns the old i_xattrs.blob (or NULL) so
759 * that it can be freed by the caller as the i_ceph_lock is likely to be
762 struct ceph_buffer
*__ceph_build_xattrs_blob(struct ceph_inode_info
*ci
)
765 struct ceph_inode_xattr
*xattr
= NULL
;
766 struct ceph_buffer
*old_blob
= NULL
;
769 dout("__build_xattrs_blob %p\n", &ci
->vfs_inode
);
770 if (ci
->i_xattrs
.dirty
) {
771 int need
= __get_required_blob_size(ci
, 0, 0);
773 BUG_ON(need
> ci
->i_xattrs
.prealloc_blob
->alloc_len
);
775 p
= rb_first(&ci
->i_xattrs
.index
);
776 dest
= ci
->i_xattrs
.prealloc_blob
->vec
.iov_base
;
778 ceph_encode_32(&dest
, ci
->i_xattrs
.count
);
780 xattr
= rb_entry(p
, struct ceph_inode_xattr
, node
);
782 ceph_encode_32(&dest
, xattr
->name_len
);
783 memcpy(dest
, xattr
->name
, xattr
->name_len
);
784 dest
+= xattr
->name_len
;
785 ceph_encode_32(&dest
, xattr
->val_len
);
786 memcpy(dest
, xattr
->val
, xattr
->val_len
);
787 dest
+= xattr
->val_len
;
792 /* adjust buffer len; it may be larger than we need */
793 ci
->i_xattrs
.prealloc_blob
->vec
.iov_len
=
794 dest
- ci
->i_xattrs
.prealloc_blob
->vec
.iov_base
;
796 if (ci
->i_xattrs
.blob
)
797 old_blob
= ci
->i_xattrs
.blob
;
798 ci
->i_xattrs
.blob
= ci
->i_xattrs
.prealloc_blob
;
799 ci
->i_xattrs
.prealloc_blob
= NULL
;
800 ci
->i_xattrs
.dirty
= false;
801 ci
->i_xattrs
.version
++;
807 static inline int __get_request_mask(struct inode
*in
) {
808 struct ceph_mds_request
*req
= current
->journal_info
;
810 if (req
&& req
->r_target_inode
== in
) {
811 if (req
->r_op
== CEPH_MDS_OP_LOOKUP
||
812 req
->r_op
== CEPH_MDS_OP_LOOKUPINO
||
813 req
->r_op
== CEPH_MDS_OP_LOOKUPPARENT
||
814 req
->r_op
== CEPH_MDS_OP_GETATTR
) {
815 mask
= le32_to_cpu(req
->r_args
.getattr
.mask
);
816 } else if (req
->r_op
== CEPH_MDS_OP_OPEN
||
817 req
->r_op
== CEPH_MDS_OP_CREATE
) {
818 mask
= le32_to_cpu(req
->r_args
.open
.mask
);
824 ssize_t
__ceph_getxattr(struct inode
*inode
, const char *name
, void *value
,
827 struct ceph_inode_info
*ci
= ceph_inode(inode
);
828 struct ceph_inode_xattr
*xattr
;
829 struct ceph_vxattr
*vxattr
= NULL
;
833 /* let's see if a virtual xattr was requested */
834 vxattr
= ceph_match_vxattr(inode
, name
);
837 if (vxattr
->flags
& VXATTR_FLAG_RSTAT
)
838 mask
|= CEPH_STAT_RSTAT
;
839 err
= ceph_do_getattr(inode
, mask
, true);
843 if (!(vxattr
->exists_cb
&& !vxattr
->exists_cb(ci
))) {
844 err
= vxattr
->getxattr_cb(ci
, value
, size
);
845 if (size
&& size
< err
)
851 req_mask
= __get_request_mask(inode
);
853 spin_lock(&ci
->i_ceph_lock
);
854 dout("getxattr %p name '%s' ver=%lld index_ver=%lld\n", inode
, name
,
855 ci
->i_xattrs
.version
, ci
->i_xattrs
.index_version
);
857 if (ci
->i_xattrs
.version
== 0 ||
858 !((req_mask
& CEPH_CAP_XATTR_SHARED
) ||
859 __ceph_caps_issued_mask(ci
, CEPH_CAP_XATTR_SHARED
, 1))) {
860 spin_unlock(&ci
->i_ceph_lock
);
862 /* security module gets xattr while filling trace */
863 if (current
->journal_info
) {
864 pr_warn_ratelimited("sync getxattr %p "
865 "during filling trace\n", inode
);
869 /* get xattrs from mds (if we don't already have them) */
870 err
= ceph_do_getattr(inode
, CEPH_STAT_CAP_XATTR
, true);
873 spin_lock(&ci
->i_ceph_lock
);
876 err
= __build_xattrs(inode
);
880 err
= -ENODATA
; /* == ENOATTR */
881 xattr
= __get_xattr(ci
, name
);
886 if (size
&& size
< xattr
->val_len
)
889 err
= xattr
->val_len
;
893 memcpy(value
, xattr
->val
, xattr
->val_len
);
895 if (current
->journal_info
&&
896 !strncmp(name
, XATTR_SECURITY_PREFIX
, XATTR_SECURITY_PREFIX_LEN
) &&
897 security_ismaclabel(name
+ XATTR_SECURITY_PREFIX_LEN
))
898 ci
->i_ceph_flags
|= CEPH_I_SEC_INITED
;
900 spin_unlock(&ci
->i_ceph_lock
);
904 ssize_t
ceph_listxattr(struct dentry
*dentry
, char *names
, size_t size
)
906 struct inode
*inode
= d_inode(dentry
);
907 struct ceph_inode_info
*ci
= ceph_inode(inode
);
908 bool len_only
= (size
== 0);
912 spin_lock(&ci
->i_ceph_lock
);
913 dout("listxattr %p ver=%lld index_ver=%lld\n", inode
,
914 ci
->i_xattrs
.version
, ci
->i_xattrs
.index_version
);
916 if (ci
->i_xattrs
.version
== 0 ||
917 !__ceph_caps_issued_mask(ci
, CEPH_CAP_XATTR_SHARED
, 1)) {
918 spin_unlock(&ci
->i_ceph_lock
);
919 err
= ceph_do_getattr(inode
, CEPH_STAT_CAP_XATTR
, true);
922 spin_lock(&ci
->i_ceph_lock
);
925 err
= __build_xattrs(inode
);
929 /* add 1 byte for each xattr due to the null termination */
930 namelen
= ci
->i_xattrs
.names_size
+ ci
->i_xattrs
.count
;
932 if (namelen
> size
) {
936 names
= __copy_xattr_names(ci
, names
);
941 spin_unlock(&ci
->i_ceph_lock
);
945 static int ceph_sync_setxattr(struct inode
*inode
, const char *name
,
946 const char *value
, size_t size
, int flags
)
948 struct ceph_fs_client
*fsc
= ceph_sb_to_client(inode
->i_sb
);
949 struct ceph_inode_info
*ci
= ceph_inode(inode
);
950 struct ceph_mds_request
*req
;
951 struct ceph_mds_client
*mdsc
= fsc
->mdsc
;
952 struct ceph_pagelist
*pagelist
= NULL
;
953 int op
= CEPH_MDS_OP_SETXATTR
;
957 /* copy value into pagelist */
958 pagelist
= ceph_pagelist_alloc(GFP_NOFS
);
962 err
= ceph_pagelist_append(pagelist
, value
, size
);
966 if (flags
& CEPH_XATTR_REPLACE
)
967 op
= CEPH_MDS_OP_RMXATTR
;
969 flags
|= CEPH_XATTR_REMOVE
;
972 dout("setxattr value=%.*s\n", (int)size
, value
);
975 req
= ceph_mdsc_create_request(mdsc
, op
, USE_AUTH_MDS
);
981 req
->r_path2
= kstrdup(name
, GFP_NOFS
);
983 ceph_mdsc_put_request(req
);
988 if (op
== CEPH_MDS_OP_SETXATTR
) {
989 req
->r_args
.setxattr
.flags
= cpu_to_le32(flags
);
990 req
->r_pagelist
= pagelist
;
994 req
->r_inode
= inode
;
997 req
->r_inode_drop
= CEPH_CAP_XATTR_SHARED
;
999 dout("xattr.ver (before): %lld\n", ci
->i_xattrs
.version
);
1000 err
= ceph_mdsc_do_request(mdsc
, NULL
, req
);
1001 ceph_mdsc_put_request(req
);
1002 dout("xattr.ver (after): %lld\n", ci
->i_xattrs
.version
);
1006 ceph_pagelist_release(pagelist
);
1010 int __ceph_setxattr(struct inode
*inode
, const char *name
,
1011 const void *value
, size_t size
, int flags
)
1013 struct ceph_vxattr
*vxattr
;
1014 struct ceph_inode_info
*ci
= ceph_inode(inode
);
1015 struct ceph_mds_client
*mdsc
= ceph_sb_to_client(inode
->i_sb
)->mdsc
;
1016 struct ceph_cap_flush
*prealloc_cf
= NULL
;
1017 struct ceph_buffer
*old_blob
= NULL
;
1021 int name_len
= strlen(name
);
1023 char *newname
= NULL
;
1024 char *newval
= NULL
;
1025 struct ceph_inode_xattr
*xattr
= NULL
;
1026 int required_blob_size
;
1027 bool check_realm
= false;
1028 bool lock_snap_rwsem
= false;
1030 if (ceph_snap(inode
) != CEPH_NOSNAP
)
1033 vxattr
= ceph_match_vxattr(inode
, name
);
1035 if (vxattr
->flags
& VXATTR_FLAG_READONLY
)
1037 if (value
&& !strncmp(vxattr
->name
, "ceph.quota", 10))
1041 /* pass any unhandled ceph.* xattrs through to the MDS */
1042 if (!strncmp(name
, XATTR_CEPH_PREFIX
, XATTR_CEPH_PREFIX_LEN
))
1043 goto do_sync_unlocked
;
1045 /* preallocate memory for xattr name, value, index node */
1047 newname
= kmemdup(name
, name_len
+ 1, GFP_NOFS
);
1052 newval
= kmemdup(value
, val_len
, GFP_NOFS
);
1057 xattr
= kmalloc(sizeof(struct ceph_inode_xattr
), GFP_NOFS
);
1061 prealloc_cf
= ceph_alloc_cap_flush();
1065 spin_lock(&ci
->i_ceph_lock
);
1067 issued
= __ceph_caps_issued(ci
, NULL
);
1068 if (ci
->i_xattrs
.version
== 0 || !(issued
& CEPH_CAP_XATTR_EXCL
))
1071 if (!lock_snap_rwsem
&& !ci
->i_head_snapc
) {
1072 lock_snap_rwsem
= true;
1073 if (!down_read_trylock(&mdsc
->snap_rwsem
)) {
1074 spin_unlock(&ci
->i_ceph_lock
);
1075 down_read(&mdsc
->snap_rwsem
);
1076 spin_lock(&ci
->i_ceph_lock
);
1081 dout("setxattr %p name '%s' issued %s\n", inode
, name
,
1082 ceph_cap_string(issued
));
1083 __build_xattrs(inode
);
1085 required_blob_size
= __get_required_blob_size(ci
, name_len
, val_len
);
1087 if (!ci
->i_xattrs
.prealloc_blob
||
1088 required_blob_size
> ci
->i_xattrs
.prealloc_blob
->alloc_len
) {
1089 struct ceph_buffer
*blob
;
1091 spin_unlock(&ci
->i_ceph_lock
);
1092 ceph_buffer_put(old_blob
); /* Shouldn't be required */
1093 dout(" pre-allocating new blob size=%d\n", required_blob_size
);
1094 blob
= ceph_buffer_new(required_blob_size
, GFP_NOFS
);
1096 goto do_sync_unlocked
;
1097 spin_lock(&ci
->i_ceph_lock
);
1098 /* prealloc_blob can't be released while holding i_ceph_lock */
1099 if (ci
->i_xattrs
.prealloc_blob
)
1100 old_blob
= ci
->i_xattrs
.prealloc_blob
;
1101 ci
->i_xattrs
.prealloc_blob
= blob
;
1105 err
= __set_xattr(ci
, newname
, name_len
, newval
, val_len
,
1106 flags
, value
? 1 : -1, &xattr
);
1109 dirty
= __ceph_mark_dirty_caps(ci
, CEPH_CAP_XATTR_EXCL
,
1111 ci
->i_xattrs
.dirty
= true;
1112 inode
->i_ctime
= current_time(inode
);
1115 spin_unlock(&ci
->i_ceph_lock
);
1116 ceph_buffer_put(old_blob
);
1117 if (lock_snap_rwsem
)
1118 up_read(&mdsc
->snap_rwsem
);
1120 __mark_inode_dirty(inode
, dirty
);
1121 ceph_free_cap_flush(prealloc_cf
);
1125 spin_unlock(&ci
->i_ceph_lock
);
1127 if (lock_snap_rwsem
)
1128 up_read(&mdsc
->snap_rwsem
);
1130 /* security module set xattr while filling trace */
1131 if (current
->journal_info
) {
1132 pr_warn_ratelimited("sync setxattr %p "
1133 "during filling trace\n", inode
);
1136 err
= ceph_sync_setxattr(inode
, name
, value
, size
, flags
);
1137 if (err
>= 0 && check_realm
) {
1138 /* check if snaprealm was created for quota inode */
1139 spin_lock(&ci
->i_ceph_lock
);
1140 if ((ci
->i_max_files
|| ci
->i_max_bytes
) &&
1141 !(ci
->i_snap_realm
&&
1142 ci
->i_snap_realm
->ino
== ci
->i_vino
.ino
))
1144 spin_unlock(&ci
->i_ceph_lock
);
1148 ceph_free_cap_flush(prealloc_cf
);
1155 static int ceph_get_xattr_handler(const struct xattr_handler
*handler
,
1156 struct dentry
*dentry
, struct inode
*inode
,
1157 const char *name
, void *value
, size_t size
)
1159 if (!ceph_is_valid_xattr(name
))
1161 return __ceph_getxattr(inode
, name
, value
, size
);
1164 static int ceph_set_xattr_handler(const struct xattr_handler
*handler
,
1165 struct dentry
*unused
, struct inode
*inode
,
1166 const char *name
, const void *value
,
1167 size_t size
, int flags
)
1169 if (!ceph_is_valid_xattr(name
))
1171 return __ceph_setxattr(inode
, name
, value
, size
, flags
);
1174 static const struct xattr_handler ceph_other_xattr_handler
= {
1175 .prefix
= "", /* match any name => handlers called with full name */
1176 .get
= ceph_get_xattr_handler
,
1177 .set
= ceph_set_xattr_handler
,
1180 #ifdef CONFIG_SECURITY
1181 bool ceph_security_xattr_wanted(struct inode
*in
)
1183 return in
->i_security
!= NULL
;
1186 bool ceph_security_xattr_deadlock(struct inode
*in
)
1188 struct ceph_inode_info
*ci
;
1190 if (!in
->i_security
)
1192 ci
= ceph_inode(in
);
1193 spin_lock(&ci
->i_ceph_lock
);
1194 ret
= !(ci
->i_ceph_flags
& CEPH_I_SEC_INITED
) &&
1195 !(ci
->i_xattrs
.version
> 0 &&
1196 __ceph_caps_issued_mask(ci
, CEPH_CAP_XATTR_SHARED
, 0));
1197 spin_unlock(&ci
->i_ceph_lock
);
1201 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1202 int ceph_security_init_secctx(struct dentry
*dentry
, umode_t mode
,
1203 struct ceph_acl_sec_ctx
*as_ctx
)
1205 struct ceph_pagelist
*pagelist
= as_ctx
->pagelist
;
1210 err
= security_dentry_init_security(dentry
, mode
, &dentry
->d_name
,
1212 &as_ctx
->sec_ctxlen
);
1214 WARN_ON_ONCE(err
!= -EOPNOTSUPP
);
1215 err
= 0; /* do nothing */
1221 pagelist
= ceph_pagelist_alloc(GFP_KERNEL
);
1224 err
= ceph_pagelist_reserve(pagelist
, PAGE_SIZE
);
1227 ceph_pagelist_encode_32(pagelist
, 1);
1231 * FIXME: Make security_dentry_init_security() generic. Currently
1232 * It only supports single security module and only selinux has
1233 * dentry_init_security hook.
1235 name
= XATTR_NAME_SELINUX
;
1236 name_len
= strlen(name
);
1237 err
= ceph_pagelist_reserve(pagelist
,
1238 4 * 2 + name_len
+ as_ctx
->sec_ctxlen
);
1242 if (as_ctx
->pagelist
) {
1243 /* update count of KV pairs */
1244 BUG_ON(pagelist
->length
<= sizeof(__le32
));
1245 if (list_is_singular(&pagelist
->head
)) {
1246 le32_add_cpu((__le32
*)pagelist
->mapped_tail
, 1);
1248 struct page
*page
= list_first_entry(&pagelist
->head
,
1250 void *addr
= kmap_atomic(page
);
1251 le32_add_cpu((__le32
*)addr
, 1);
1252 kunmap_atomic(addr
);
1255 as_ctx
->pagelist
= pagelist
;
1258 ceph_pagelist_encode_32(pagelist
, name_len
);
1259 ceph_pagelist_append(pagelist
, name
, name_len
);
1261 ceph_pagelist_encode_32(pagelist
, as_ctx
->sec_ctxlen
);
1262 ceph_pagelist_append(pagelist
, as_ctx
->sec_ctx
, as_ctx
->sec_ctxlen
);
1266 if (pagelist
&& !as_ctx
->pagelist
)
1267 ceph_pagelist_release(pagelist
);
1270 #endif /* CONFIG_CEPH_FS_SECURITY_LABEL */
1271 #endif /* CONFIG_SECURITY */
1273 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx
*as_ctx
)
1275 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1276 posix_acl_release(as_ctx
->acl
);
1277 posix_acl_release(as_ctx
->default_acl
);
1279 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1280 security_release_secctx(as_ctx
->sec_ctx
, as_ctx
->sec_ctxlen
);
1282 if (as_ctx
->pagelist
)
1283 ceph_pagelist_release(as_ctx
->pagelist
);
1287 * List of handlers for synthetic system.* attributes. Other
1288 * attributes are handled directly.
1290 const struct xattr_handler
*ceph_xattr_handlers
[] = {
1291 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1292 &posix_acl_access_xattr_handler
,
1293 &posix_acl_default_xattr_handler
,
1295 &ceph_other_xattr_handler
,