Linux 3.16-rc2
[linux/fpc-iii.git] / fs / ceph / xattr.c
blobc9c2b887381ec2504ee48c673cf113e396f3b04e
1 #include <linux/ceph/ceph_debug.h>
3 #include "super.h"
4 #include "mds_client.h"
6 #include <linux/ceph/decode.h>
8 #include <linux/xattr.h>
9 #include <linux/posix_acl_xattr.h>
10 #include <linux/slab.h>
12 #define XATTR_CEPH_PREFIX "ceph."
13 #define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
15 static int __remove_xattr(struct ceph_inode_info *ci,
16 struct ceph_inode_xattr *xattr);
19 * List of handlers for synthetic system.* attributes. Other
20 * attributes are handled directly.
22 const struct xattr_handler *ceph_xattr_handlers[] = {
23 #ifdef CONFIG_CEPH_FS_POSIX_ACL
24 &posix_acl_access_xattr_handler,
25 &posix_acl_default_xattr_handler,
26 #endif
27 NULL,
30 static bool ceph_is_valid_xattr(const char *name)
32 return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
33 !strncmp(name, XATTR_SECURITY_PREFIX,
34 XATTR_SECURITY_PREFIX_LEN) ||
35 !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN) ||
36 !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
37 !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
41 * These define virtual xattrs exposing the recursive directory
42 * statistics and layout metadata.
44 struct ceph_vxattr {
45 char *name;
46 size_t name_size; /* strlen(name) + 1 (for '\0') */
47 size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
48 size_t size);
49 bool readonly, hidden;
50 bool (*exists_cb)(struct ceph_inode_info *ci);
53 /* layouts */
55 static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
57 size_t s;
58 char *p = (char *)&ci->i_layout;
60 for (s = 0; s < sizeof(ci->i_layout); s++, p++)
61 if (*p)
62 return true;
63 return false;
66 static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
67 size_t size)
69 int ret;
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 s64 pool = ceph_file_layout_pg_pool(ci->i_layout);
73 const char *pool_name;
74 char buf[128];
76 dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
77 down_read(&osdc->map_sem);
78 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
79 if (pool_name) {
80 size_t len = strlen(pool_name);
81 ret = snprintf(buf, sizeof(buf),
82 "stripe_unit=%lld stripe_count=%lld object_size=%lld pool=",
83 (unsigned long long)ceph_file_layout_su(ci->i_layout),
84 (unsigned long long)ceph_file_layout_stripe_count(ci->i_layout),
85 (unsigned long long)ceph_file_layout_object_size(ci->i_layout));
86 if (!size) {
87 ret += len;
88 } else if (ret + len > size) {
89 ret = -ERANGE;
90 } else {
91 memcpy(val, buf, ret);
92 memcpy(val + ret, pool_name, len);
93 ret += len;
95 } else {
96 ret = snprintf(buf, sizeof(buf),
97 "stripe_unit=%lld stripe_count=%lld object_size=%lld pool=%lld",
98 (unsigned long long)ceph_file_layout_su(ci->i_layout),
99 (unsigned long long)ceph_file_layout_stripe_count(ci->i_layout),
100 (unsigned long long)ceph_file_layout_object_size(ci->i_layout),
101 (unsigned long long)pool);
102 if (size) {
103 if (ret <= size)
104 memcpy(val, buf, ret);
105 else
106 ret = -ERANGE;
109 up_read(&osdc->map_sem);
110 return ret;
113 static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
114 char *val, size_t size)
116 return snprintf(val, size, "%lld",
117 (unsigned long long)ceph_file_layout_su(ci->i_layout));
120 static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
121 char *val, size_t size)
123 return snprintf(val, size, "%lld",
124 (unsigned long long)ceph_file_layout_stripe_count(ci->i_layout));
127 static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
128 char *val, size_t size)
130 return snprintf(val, size, "%lld",
131 (unsigned long long)ceph_file_layout_object_size(ci->i_layout));
134 static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
135 char *val, size_t size)
137 int ret;
138 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
139 struct ceph_osd_client *osdc = &fsc->client->osdc;
140 s64 pool = ceph_file_layout_pg_pool(ci->i_layout);
141 const char *pool_name;
143 down_read(&osdc->map_sem);
144 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
145 if (pool_name)
146 ret = snprintf(val, size, "%s", pool_name);
147 else
148 ret = snprintf(val, size, "%lld", (unsigned long long)pool);
149 up_read(&osdc->map_sem);
150 return ret;
153 /* directories */
155 static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
156 size_t size)
158 return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
161 static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
162 size_t size)
164 return snprintf(val, size, "%lld", ci->i_files);
167 static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
168 size_t size)
170 return snprintf(val, size, "%lld", ci->i_subdirs);
173 static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
174 size_t size)
176 return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
179 static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
180 size_t size)
182 return snprintf(val, size, "%lld", ci->i_rfiles);
185 static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
186 size_t size)
188 return snprintf(val, size, "%lld", ci->i_rsubdirs);
191 static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
192 size_t size)
194 return snprintf(val, size, "%lld", ci->i_rbytes);
197 static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
198 size_t size)
200 return snprintf(val, size, "%ld.09%ld", (long)ci->i_rctime.tv_sec,
201 (long)ci->i_rctime.tv_nsec);
205 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
206 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
207 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
209 #define XATTR_NAME_CEPH(_type, _name) \
211 .name = CEPH_XATTR_NAME(_type, _name), \
212 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
213 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
214 .readonly = true, \
215 .hidden = false, \
216 .exists_cb = NULL, \
218 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
220 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
221 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
222 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
223 .readonly = false, \
224 .hidden = true, \
225 .exists_cb = ceph_vxattrcb_layout_exists, \
228 static struct ceph_vxattr ceph_dir_vxattrs[] = {
230 .name = "ceph.dir.layout",
231 .name_size = sizeof("ceph.dir.layout"),
232 .getxattr_cb = ceph_vxattrcb_layout,
233 .readonly = false,
234 .hidden = true,
235 .exists_cb = ceph_vxattrcb_layout_exists,
237 XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
238 XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
239 XATTR_LAYOUT_FIELD(dir, layout, object_size),
240 XATTR_LAYOUT_FIELD(dir, layout, pool),
241 XATTR_NAME_CEPH(dir, entries),
242 XATTR_NAME_CEPH(dir, files),
243 XATTR_NAME_CEPH(dir, subdirs),
244 XATTR_NAME_CEPH(dir, rentries),
245 XATTR_NAME_CEPH(dir, rfiles),
246 XATTR_NAME_CEPH(dir, rsubdirs),
247 XATTR_NAME_CEPH(dir, rbytes),
248 XATTR_NAME_CEPH(dir, rctime),
249 { .name = NULL, 0 } /* Required table terminator */
251 static size_t ceph_dir_vxattrs_name_size; /* total size of all names */
253 /* files */
255 static struct ceph_vxattr ceph_file_vxattrs[] = {
257 .name = "ceph.file.layout",
258 .name_size = sizeof("ceph.file.layout"),
259 .getxattr_cb = ceph_vxattrcb_layout,
260 .readonly = false,
261 .hidden = true,
262 .exists_cb = ceph_vxattrcb_layout_exists,
264 XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
265 XATTR_LAYOUT_FIELD(file, layout, stripe_count),
266 XATTR_LAYOUT_FIELD(file, layout, object_size),
267 XATTR_LAYOUT_FIELD(file, layout, pool),
268 { .name = NULL, 0 } /* Required table terminator */
270 static size_t ceph_file_vxattrs_name_size; /* total size of all names */
272 static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
274 if (S_ISDIR(inode->i_mode))
275 return ceph_dir_vxattrs;
276 else if (S_ISREG(inode->i_mode))
277 return ceph_file_vxattrs;
278 return NULL;
281 static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
283 if (vxattrs == ceph_dir_vxattrs)
284 return ceph_dir_vxattrs_name_size;
285 if (vxattrs == ceph_file_vxattrs)
286 return ceph_file_vxattrs_name_size;
287 BUG();
289 return 0;
293 * Compute the aggregate size (including terminating '\0') of all
294 * virtual extended attribute names in the given vxattr table.
296 static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
298 struct ceph_vxattr *vxattr;
299 size_t size = 0;
301 for (vxattr = vxattrs; vxattr->name; vxattr++)
302 if (!vxattr->hidden)
303 size += vxattr->name_size;
305 return size;
308 /* Routines called at initialization and exit time */
310 void __init ceph_xattr_init(void)
312 ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
313 ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
316 void ceph_xattr_exit(void)
318 ceph_dir_vxattrs_name_size = 0;
319 ceph_file_vxattrs_name_size = 0;
322 static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
323 const char *name)
325 struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
327 if (vxattr) {
328 while (vxattr->name) {
329 if (!strcmp(vxattr->name, name))
330 return vxattr;
331 vxattr++;
335 return NULL;
338 static int __set_xattr(struct ceph_inode_info *ci,
339 const char *name, int name_len,
340 const char *val, int val_len,
341 int flags, int update_xattr,
342 struct ceph_inode_xattr **newxattr)
344 struct rb_node **p;
345 struct rb_node *parent = NULL;
346 struct ceph_inode_xattr *xattr = NULL;
347 int c;
348 int new = 0;
350 p = &ci->i_xattrs.index.rb_node;
351 while (*p) {
352 parent = *p;
353 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
354 c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
355 if (c < 0)
356 p = &(*p)->rb_left;
357 else if (c > 0)
358 p = &(*p)->rb_right;
359 else {
360 if (name_len == xattr->name_len)
361 break;
362 else if (name_len < xattr->name_len)
363 p = &(*p)->rb_left;
364 else
365 p = &(*p)->rb_right;
367 xattr = NULL;
370 if (update_xattr) {
371 int err = 0;
372 if (xattr && (flags & XATTR_CREATE))
373 err = -EEXIST;
374 else if (!xattr && (flags & XATTR_REPLACE))
375 err = -ENODATA;
376 if (err) {
377 kfree(name);
378 kfree(val);
379 return err;
381 if (update_xattr < 0) {
382 if (xattr)
383 __remove_xattr(ci, xattr);
384 kfree(name);
385 return 0;
389 if (!xattr) {
390 new = 1;
391 xattr = *newxattr;
392 xattr->name = name;
393 xattr->name_len = name_len;
394 xattr->should_free_name = update_xattr;
396 ci->i_xattrs.count++;
397 dout("__set_xattr count=%d\n", ci->i_xattrs.count);
398 } else {
399 kfree(*newxattr);
400 *newxattr = NULL;
401 if (xattr->should_free_val)
402 kfree((void *)xattr->val);
404 if (update_xattr) {
405 kfree((void *)name);
406 name = xattr->name;
408 ci->i_xattrs.names_size -= xattr->name_len;
409 ci->i_xattrs.vals_size -= xattr->val_len;
411 ci->i_xattrs.names_size += name_len;
412 ci->i_xattrs.vals_size += val_len;
413 if (val)
414 xattr->val = val;
415 else
416 xattr->val = "";
418 xattr->val_len = val_len;
419 xattr->dirty = update_xattr;
420 xattr->should_free_val = (val && update_xattr);
422 if (new) {
423 rb_link_node(&xattr->node, parent, p);
424 rb_insert_color(&xattr->node, &ci->i_xattrs.index);
425 dout("__set_xattr_val p=%p\n", p);
428 dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
429 ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);
431 return 0;
434 static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
435 const char *name)
437 struct rb_node **p;
438 struct rb_node *parent = NULL;
439 struct ceph_inode_xattr *xattr = NULL;
440 int name_len = strlen(name);
441 int c;
443 p = &ci->i_xattrs.index.rb_node;
444 while (*p) {
445 parent = *p;
446 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
447 c = strncmp(name, xattr->name, xattr->name_len);
448 if (c == 0 && name_len > xattr->name_len)
449 c = 1;
450 if (c < 0)
451 p = &(*p)->rb_left;
452 else if (c > 0)
453 p = &(*p)->rb_right;
454 else {
455 dout("__get_xattr %s: found %.*s\n", name,
456 xattr->val_len, xattr->val);
457 return xattr;
461 dout("__get_xattr %s: not found\n", name);
463 return NULL;
466 static void __free_xattr(struct ceph_inode_xattr *xattr)
468 BUG_ON(!xattr);
470 if (xattr->should_free_name)
471 kfree((void *)xattr->name);
472 if (xattr->should_free_val)
473 kfree((void *)xattr->val);
475 kfree(xattr);
478 static int __remove_xattr(struct ceph_inode_info *ci,
479 struct ceph_inode_xattr *xattr)
481 if (!xattr)
482 return -ENODATA;
484 rb_erase(&xattr->node, &ci->i_xattrs.index);
486 if (xattr->should_free_name)
487 kfree((void *)xattr->name);
488 if (xattr->should_free_val)
489 kfree((void *)xattr->val);
491 ci->i_xattrs.names_size -= xattr->name_len;
492 ci->i_xattrs.vals_size -= xattr->val_len;
493 ci->i_xattrs.count--;
494 kfree(xattr);
496 return 0;
499 static int __remove_xattr_by_name(struct ceph_inode_info *ci,
500 const char *name)
502 struct rb_node **p;
503 struct ceph_inode_xattr *xattr;
504 int err;
506 p = &ci->i_xattrs.index.rb_node;
507 xattr = __get_xattr(ci, name);
508 err = __remove_xattr(ci, xattr);
509 return err;
512 static char *__copy_xattr_names(struct ceph_inode_info *ci,
513 char *dest)
515 struct rb_node *p;
516 struct ceph_inode_xattr *xattr = NULL;
518 p = rb_first(&ci->i_xattrs.index);
519 dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
521 while (p) {
522 xattr = rb_entry(p, struct ceph_inode_xattr, node);
523 memcpy(dest, xattr->name, xattr->name_len);
524 dest[xattr->name_len] = '\0';
526 dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
527 xattr->name_len, ci->i_xattrs.names_size);
529 dest += xattr->name_len + 1;
530 p = rb_next(p);
533 return dest;
536 void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
538 struct rb_node *p, *tmp;
539 struct ceph_inode_xattr *xattr = NULL;
541 p = rb_first(&ci->i_xattrs.index);
543 dout("__ceph_destroy_xattrs p=%p\n", p);
545 while (p) {
546 xattr = rb_entry(p, struct ceph_inode_xattr, node);
547 tmp = p;
548 p = rb_next(tmp);
549 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
550 xattr->name_len, xattr->name);
551 rb_erase(tmp, &ci->i_xattrs.index);
553 __free_xattr(xattr);
556 ci->i_xattrs.names_size = 0;
557 ci->i_xattrs.vals_size = 0;
558 ci->i_xattrs.index_version = 0;
559 ci->i_xattrs.count = 0;
560 ci->i_xattrs.index = RB_ROOT;
563 static int __build_xattrs(struct inode *inode)
564 __releases(ci->i_ceph_lock)
565 __acquires(ci->i_ceph_lock)
567 u32 namelen;
568 u32 numattr = 0;
569 void *p, *end;
570 u32 len;
571 const char *name, *val;
572 struct ceph_inode_info *ci = ceph_inode(inode);
573 int xattr_version;
574 struct ceph_inode_xattr **xattrs = NULL;
575 int err = 0;
576 int i;
578 dout("__build_xattrs() len=%d\n",
579 ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
581 if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
582 return 0; /* already built */
584 __ceph_destroy_xattrs(ci);
586 start:
587 /* updated internal xattr rb tree */
588 if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
589 p = ci->i_xattrs.blob->vec.iov_base;
590 end = p + ci->i_xattrs.blob->vec.iov_len;
591 ceph_decode_32_safe(&p, end, numattr, bad);
592 xattr_version = ci->i_xattrs.version;
593 spin_unlock(&ci->i_ceph_lock);
595 xattrs = kcalloc(numattr, sizeof(struct ceph_xattr *),
596 GFP_NOFS);
597 err = -ENOMEM;
598 if (!xattrs)
599 goto bad_lock;
600 memset(xattrs, 0, numattr*sizeof(struct ceph_xattr *));
601 for (i = 0; i < numattr; i++) {
602 xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
603 GFP_NOFS);
604 if (!xattrs[i])
605 goto bad_lock;
608 spin_lock(&ci->i_ceph_lock);
609 if (ci->i_xattrs.version != xattr_version) {
610 /* lost a race, retry */
611 for (i = 0; i < numattr; i++)
612 kfree(xattrs[i]);
613 kfree(xattrs);
614 xattrs = NULL;
615 goto start;
617 err = -EIO;
618 while (numattr--) {
619 ceph_decode_32_safe(&p, end, len, bad);
620 namelen = len;
621 name = p;
622 p += len;
623 ceph_decode_32_safe(&p, end, len, bad);
624 val = p;
625 p += len;
627 err = __set_xattr(ci, name, namelen, val, len,
628 0, 0, &xattrs[numattr]);
630 if (err < 0)
631 goto bad;
633 kfree(xattrs);
635 ci->i_xattrs.index_version = ci->i_xattrs.version;
636 ci->i_xattrs.dirty = false;
638 return err;
639 bad_lock:
640 spin_lock(&ci->i_ceph_lock);
641 bad:
642 if (xattrs) {
643 for (i = 0; i < numattr; i++)
644 kfree(xattrs[i]);
645 kfree(xattrs);
647 ci->i_xattrs.names_size = 0;
648 return err;
651 static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
652 int val_size)
655 * 4 bytes for the length, and additional 4 bytes per each xattr name,
656 * 4 bytes per each value
658 int size = 4 + ci->i_xattrs.count*(4 + 4) +
659 ci->i_xattrs.names_size +
660 ci->i_xattrs.vals_size;
661 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
662 ci->i_xattrs.count, ci->i_xattrs.names_size,
663 ci->i_xattrs.vals_size);
665 if (name_size)
666 size += 4 + 4 + name_size + val_size;
668 return size;
672 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
673 * and swap into place.
675 void __ceph_build_xattrs_blob(struct ceph_inode_info *ci)
677 struct rb_node *p;
678 struct ceph_inode_xattr *xattr = NULL;
679 void *dest;
681 dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
682 if (ci->i_xattrs.dirty) {
683 int need = __get_required_blob_size(ci, 0, 0);
685 BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
687 p = rb_first(&ci->i_xattrs.index);
688 dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
690 ceph_encode_32(&dest, ci->i_xattrs.count);
691 while (p) {
692 xattr = rb_entry(p, struct ceph_inode_xattr, node);
694 ceph_encode_32(&dest, xattr->name_len);
695 memcpy(dest, xattr->name, xattr->name_len);
696 dest += xattr->name_len;
697 ceph_encode_32(&dest, xattr->val_len);
698 memcpy(dest, xattr->val, xattr->val_len);
699 dest += xattr->val_len;
701 p = rb_next(p);
704 /* adjust buffer len; it may be larger than we need */
705 ci->i_xattrs.prealloc_blob->vec.iov_len =
706 dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
708 if (ci->i_xattrs.blob)
709 ceph_buffer_put(ci->i_xattrs.blob);
710 ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
711 ci->i_xattrs.prealloc_blob = NULL;
712 ci->i_xattrs.dirty = false;
713 ci->i_xattrs.version++;
717 ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
718 size_t size)
720 struct ceph_inode_info *ci = ceph_inode(inode);
721 int err;
722 struct ceph_inode_xattr *xattr;
723 struct ceph_vxattr *vxattr = NULL;
725 if (!ceph_is_valid_xattr(name))
726 return -ENODATA;
728 /* let's see if a virtual xattr was requested */
729 vxattr = ceph_match_vxattr(inode, name);
730 if (vxattr && !(vxattr->exists_cb && !vxattr->exists_cb(ci))) {
731 err = vxattr->getxattr_cb(ci, value, size);
732 return err;
735 spin_lock(&ci->i_ceph_lock);
736 dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
737 ci->i_xattrs.version, ci->i_xattrs.index_version);
739 if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) &&
740 (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
741 goto get_xattr;
742 } else {
743 spin_unlock(&ci->i_ceph_lock);
744 /* get xattrs from mds (if we don't already have them) */
745 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
746 if (err)
747 return err;
750 spin_lock(&ci->i_ceph_lock);
752 err = __build_xattrs(inode);
753 if (err < 0)
754 goto out;
756 get_xattr:
757 err = -ENODATA; /* == ENOATTR */
758 xattr = __get_xattr(ci, name);
759 if (!xattr)
760 goto out;
762 err = -ERANGE;
763 if (size && size < xattr->val_len)
764 goto out;
766 err = xattr->val_len;
767 if (size == 0)
768 goto out;
770 memcpy(value, xattr->val, xattr->val_len);
772 out:
773 spin_unlock(&ci->i_ceph_lock);
774 return err;
777 ssize_t ceph_getxattr(struct dentry *dentry, const char *name, void *value,
778 size_t size)
780 if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
781 return generic_getxattr(dentry, name, value, size);
783 return __ceph_getxattr(dentry->d_inode, name, value, size);
786 ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
788 struct inode *inode = dentry->d_inode;
789 struct ceph_inode_info *ci = ceph_inode(inode);
790 struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
791 u32 vir_namelen = 0;
792 u32 namelen;
793 int err;
794 u32 len;
795 int i;
797 spin_lock(&ci->i_ceph_lock);
798 dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
799 ci->i_xattrs.version, ci->i_xattrs.index_version);
801 if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) &&
802 (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
803 goto list_xattr;
804 } else {
805 spin_unlock(&ci->i_ceph_lock);
806 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
807 if (err)
808 return err;
811 spin_lock(&ci->i_ceph_lock);
813 err = __build_xattrs(inode);
814 if (err < 0)
815 goto out;
817 list_xattr:
819 * Start with virtual dir xattr names (if any) (including
820 * terminating '\0' characters for each).
822 vir_namelen = ceph_vxattrs_name_size(vxattrs);
824 /* adding 1 byte per each variable due to the null termination */
825 namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
826 err = -ERANGE;
827 if (size && vir_namelen + namelen > size)
828 goto out;
830 err = namelen + vir_namelen;
831 if (size == 0)
832 goto out;
834 names = __copy_xattr_names(ci, names);
836 /* virtual xattr names, too */
837 err = namelen;
838 if (vxattrs) {
839 for (i = 0; vxattrs[i].name; i++) {
840 if (!vxattrs[i].hidden &&
841 !(vxattrs[i].exists_cb &&
842 !vxattrs[i].exists_cb(ci))) {
843 len = sprintf(names, "%s", vxattrs[i].name);
844 names += len + 1;
845 err += len + 1;
850 out:
851 spin_unlock(&ci->i_ceph_lock);
852 return err;
855 static int ceph_sync_setxattr(struct dentry *dentry, const char *name,
856 const char *value, size_t size, int flags)
858 struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
859 struct inode *inode = dentry->d_inode;
860 struct ceph_inode_info *ci = ceph_inode(inode);
861 struct ceph_mds_request *req;
862 struct ceph_mds_client *mdsc = fsc->mdsc;
863 int err;
864 int i, nr_pages;
865 struct page **pages = NULL;
866 void *kaddr;
868 /* copy value into some pages */
869 nr_pages = calc_pages_for(0, size);
870 if (nr_pages) {
871 pages = kmalloc(sizeof(pages[0])*nr_pages, GFP_NOFS);
872 if (!pages)
873 return -ENOMEM;
874 err = -ENOMEM;
875 for (i = 0; i < nr_pages; i++) {
876 pages[i] = __page_cache_alloc(GFP_NOFS);
877 if (!pages[i]) {
878 nr_pages = i;
879 goto out;
881 kaddr = kmap(pages[i]);
882 memcpy(kaddr, value + i*PAGE_CACHE_SIZE,
883 min(PAGE_CACHE_SIZE, size-i*PAGE_CACHE_SIZE));
887 dout("setxattr value=%.*s\n", (int)size, value);
889 if (!value)
890 flags |= CEPH_XATTR_REMOVE;
892 /* do request */
893 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETXATTR,
894 USE_AUTH_MDS);
895 if (IS_ERR(req)) {
896 err = PTR_ERR(req);
897 goto out;
899 req->r_inode = inode;
900 ihold(inode);
901 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
902 req->r_num_caps = 1;
903 req->r_args.setxattr.flags = cpu_to_le32(flags);
904 req->r_path2 = kstrdup(name, GFP_NOFS);
906 req->r_pages = pages;
907 req->r_num_pages = nr_pages;
908 req->r_data_len = size;
910 dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
911 err = ceph_mdsc_do_request(mdsc, NULL, req);
912 ceph_mdsc_put_request(req);
913 dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
915 out:
916 if (pages) {
917 for (i = 0; i < nr_pages; i++)
918 __free_page(pages[i]);
919 kfree(pages);
921 return err;
924 int __ceph_setxattr(struct dentry *dentry, const char *name,
925 const void *value, size_t size, int flags)
927 struct inode *inode = dentry->d_inode;
928 struct ceph_vxattr *vxattr;
929 struct ceph_inode_info *ci = ceph_inode(inode);
930 int issued;
931 int err;
932 int dirty = 0;
933 int name_len = strlen(name);
934 int val_len = size;
935 char *newname = NULL;
936 char *newval = NULL;
937 struct ceph_inode_xattr *xattr = NULL;
938 int required_blob_size;
940 if (!ceph_is_valid_xattr(name))
941 return -EOPNOTSUPP;
943 vxattr = ceph_match_vxattr(inode, name);
944 if (vxattr && vxattr->readonly)
945 return -EOPNOTSUPP;
947 /* pass any unhandled ceph.* xattrs through to the MDS */
948 if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
949 goto do_sync_unlocked;
951 /* preallocate memory for xattr name, value, index node */
952 err = -ENOMEM;
953 newname = kmemdup(name, name_len + 1, GFP_NOFS);
954 if (!newname)
955 goto out;
957 if (val_len) {
958 newval = kmemdup(value, val_len, GFP_NOFS);
959 if (!newval)
960 goto out;
963 xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
964 if (!xattr)
965 goto out;
967 spin_lock(&ci->i_ceph_lock);
968 retry:
969 issued = __ceph_caps_issued(ci, NULL);
970 dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
971 if (!(issued & CEPH_CAP_XATTR_EXCL))
972 goto do_sync;
973 __build_xattrs(inode);
975 required_blob_size = __get_required_blob_size(ci, name_len, val_len);
977 if (!ci->i_xattrs.prealloc_blob ||
978 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
979 struct ceph_buffer *blob;
981 spin_unlock(&ci->i_ceph_lock);
982 dout(" preaallocating new blob size=%d\n", required_blob_size);
983 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
984 if (!blob)
985 goto out;
986 spin_lock(&ci->i_ceph_lock);
987 if (ci->i_xattrs.prealloc_blob)
988 ceph_buffer_put(ci->i_xattrs.prealloc_blob);
989 ci->i_xattrs.prealloc_blob = blob;
990 goto retry;
993 err = __set_xattr(ci, newname, name_len, newval, val_len,
994 flags, value ? 1 : -1, &xattr);
996 if (!err) {
997 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
998 ci->i_xattrs.dirty = true;
999 inode->i_ctime = CURRENT_TIME;
1002 spin_unlock(&ci->i_ceph_lock);
1003 if (dirty)
1004 __mark_inode_dirty(inode, dirty);
1005 return err;
1007 do_sync:
1008 spin_unlock(&ci->i_ceph_lock);
1009 do_sync_unlocked:
1010 err = ceph_sync_setxattr(dentry, name, value, size, flags);
1011 out:
1012 kfree(newname);
1013 kfree(newval);
1014 kfree(xattr);
1015 return err;
1018 int ceph_setxattr(struct dentry *dentry, const char *name,
1019 const void *value, size_t size, int flags)
1021 if (ceph_snap(dentry->d_inode) != CEPH_NOSNAP)
1022 return -EROFS;
1024 if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
1025 return generic_setxattr(dentry, name, value, size, flags);
1027 return __ceph_setxattr(dentry, name, value, size, flags);
1030 static int ceph_send_removexattr(struct dentry *dentry, const char *name)
1032 struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
1033 struct ceph_mds_client *mdsc = fsc->mdsc;
1034 struct inode *inode = dentry->d_inode;
1035 struct ceph_mds_request *req;
1036 int err;
1038 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_RMXATTR,
1039 USE_AUTH_MDS);
1040 if (IS_ERR(req))
1041 return PTR_ERR(req);
1042 req->r_inode = inode;
1043 ihold(inode);
1044 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
1045 req->r_num_caps = 1;
1046 req->r_path2 = kstrdup(name, GFP_NOFS);
1048 err = ceph_mdsc_do_request(mdsc, NULL, req);
1049 ceph_mdsc_put_request(req);
1050 return err;
1053 int __ceph_removexattr(struct dentry *dentry, const char *name)
1055 struct inode *inode = dentry->d_inode;
1056 struct ceph_vxattr *vxattr;
1057 struct ceph_inode_info *ci = ceph_inode(inode);
1058 int issued;
1059 int err;
1060 int required_blob_size;
1061 int dirty;
1063 if (!ceph_is_valid_xattr(name))
1064 return -EOPNOTSUPP;
1066 vxattr = ceph_match_vxattr(inode, name);
1067 if (vxattr && vxattr->readonly)
1068 return -EOPNOTSUPP;
1070 /* pass any unhandled ceph.* xattrs through to the MDS */
1071 if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
1072 goto do_sync_unlocked;
1074 err = -ENOMEM;
1075 spin_lock(&ci->i_ceph_lock);
1076 retry:
1077 issued = __ceph_caps_issued(ci, NULL);
1078 dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued));
1080 if (!(issued & CEPH_CAP_XATTR_EXCL))
1081 goto do_sync;
1082 __build_xattrs(inode);
1084 required_blob_size = __get_required_blob_size(ci, 0, 0);
1086 if (!ci->i_xattrs.prealloc_blob ||
1087 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1088 struct ceph_buffer *blob;
1090 spin_unlock(&ci->i_ceph_lock);
1091 dout(" preaallocating new blob size=%d\n", required_blob_size);
1092 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1093 if (!blob)
1094 goto out;
1095 spin_lock(&ci->i_ceph_lock);
1096 if (ci->i_xattrs.prealloc_blob)
1097 ceph_buffer_put(ci->i_xattrs.prealloc_blob);
1098 ci->i_xattrs.prealloc_blob = blob;
1099 goto retry;
1102 err = __remove_xattr_by_name(ceph_inode(inode), name);
1104 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
1105 ci->i_xattrs.dirty = true;
1106 inode->i_ctime = CURRENT_TIME;
1107 spin_unlock(&ci->i_ceph_lock);
1108 if (dirty)
1109 __mark_inode_dirty(inode, dirty);
1110 return err;
1111 do_sync:
1112 spin_unlock(&ci->i_ceph_lock);
1113 do_sync_unlocked:
1114 err = ceph_send_removexattr(dentry, name);
1115 out:
1116 return err;
1119 int ceph_removexattr(struct dentry *dentry, const char *name)
1121 if (ceph_snap(dentry->d_inode) != CEPH_NOSNAP)
1122 return -EROFS;
1124 if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
1125 return generic_removexattr(dentry, name);
1127 return __ceph_removexattr(dentry, name);