slab: Common name for the per node structures
[linux-2.6-block.git] / fs / ceph / xattr.c
CommitLineData
3d14c5d2
YS
1#include <linux/ceph/ceph_debug.h>
2
355da1eb 3#include "super.h"
3d14c5d2
YS
4#include "mds_client.h"
5
6#include <linux/ceph/decode.h>
355da1eb
SW
7
8#include <linux/xattr.h>
5a0e3ad6 9#include <linux/slab.h>
355da1eb 10
22891907
AE
11#define XATTR_CEPH_PREFIX "ceph."
12#define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
13
355da1eb
SW
14static bool ceph_is_valid_xattr(const char *name)
15{
22891907 16 return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
1a756278 17 !strncmp(name, XATTR_SECURITY_PREFIX,
355da1eb
SW
18 XATTR_SECURITY_PREFIX_LEN) ||
19 !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
20 !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
21}
22
23/*
24 * These define virtual xattrs exposing the recursive directory
25 * statistics and layout metadata.
26 */
881a5fa2 27struct ceph_vxattr {
355da1eb 28 char *name;
3ce6cd12 29 size_t name_size; /* strlen(name) + 1 (for '\0') */
355da1eb
SW
30 size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
31 size_t size);
eb788084 32 bool readonly;
355da1eb
SW
33};
34
35/* directories */
36
aa4066ed 37static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
355da1eb
SW
38 size_t size)
39{
40 return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
41}
42
aa4066ed 43static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
355da1eb
SW
44 size_t size)
45{
46 return snprintf(val, size, "%lld", ci->i_files);
47}
48
aa4066ed 49static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
355da1eb
SW
50 size_t size)
51{
52 return snprintf(val, size, "%lld", ci->i_subdirs);
53}
54
aa4066ed 55static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
355da1eb
SW
56 size_t size)
57{
58 return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
59}
60
aa4066ed 61static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
355da1eb
SW
62 size_t size)
63{
64 return snprintf(val, size, "%lld", ci->i_rfiles);
65}
66
aa4066ed 67static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
355da1eb
SW
68 size_t size)
69{
70 return snprintf(val, size, "%lld", ci->i_rsubdirs);
71}
72
aa4066ed 73static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
355da1eb
SW
74 size_t size)
75{
76 return snprintf(val, size, "%lld", ci->i_rbytes);
77}
78
aa4066ed 79static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
355da1eb
SW
80 size_t size)
81{
3489b42a 82 return snprintf(val, size, "%ld.09%ld", (long)ci->i_rctime.tv_sec,
355da1eb
SW
83 (long)ci->i_rctime.tv_nsec);
84}
85
eb788084
AE
86#define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
87
88#define XATTR_NAME_CEPH(_type, _name) \
89 { \
90 .name = CEPH_XATTR_NAME(_type, _name), \
3ce6cd12 91 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
aa4066ed 92 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
eb788084
AE
93 .readonly = true, \
94 }
95
881a5fa2 96static struct ceph_vxattr ceph_dir_vxattrs[] = {
eb788084
AE
97 XATTR_NAME_CEPH(dir, entries),
98 XATTR_NAME_CEPH(dir, files),
99 XATTR_NAME_CEPH(dir, subdirs),
100 XATTR_NAME_CEPH(dir, rentries),
101 XATTR_NAME_CEPH(dir, rfiles),
102 XATTR_NAME_CEPH(dir, rsubdirs),
103 XATTR_NAME_CEPH(dir, rbytes),
104 XATTR_NAME_CEPH(dir, rctime),
105 { 0 } /* Required table terminator */
355da1eb 106};
3ce6cd12 107static size_t ceph_dir_vxattrs_name_size; /* total size of all names */
355da1eb
SW
108
109/* files */
110
aa4066ed 111static size_t ceph_vxattrcb_file_layout(struct ceph_inode_info *ci, char *val,
355da1eb
SW
112 size_t size)
113{
b195befd
SW
114 int ret;
115
116 ret = snprintf(val, size,
355da1eb
SW
117 "chunk_bytes=%lld\nstripe_count=%lld\nobject_size=%lld\n",
118 (unsigned long long)ceph_file_layout_su(ci->i_layout),
119 (unsigned long long)ceph_file_layout_stripe_count(ci->i_layout),
120 (unsigned long long)ceph_file_layout_object_size(ci->i_layout));
b195befd 121 return ret;
355da1eb
SW
122}
123
881a5fa2 124static struct ceph_vxattr ceph_file_vxattrs[] = {
eb788084 125 XATTR_NAME_CEPH(file, layout),
114fc474 126 /* The following extended attribute name is deprecated */
eb788084
AE
127 {
128 .name = XATTR_CEPH_PREFIX "layout",
3ce6cd12 129 .name_size = sizeof (XATTR_CEPH_PREFIX "layout"),
aa4066ed 130 .getxattr_cb = ceph_vxattrcb_file_layout,
eb788084
AE
131 .readonly = true,
132 },
133 { 0 } /* Required table terminator */
355da1eb 134};
3ce6cd12 135static size_t ceph_file_vxattrs_name_size; /* total size of all names */
355da1eb 136
881a5fa2 137static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
355da1eb
SW
138{
139 if (S_ISDIR(inode->i_mode))
140 return ceph_dir_vxattrs;
141 else if (S_ISREG(inode->i_mode))
142 return ceph_file_vxattrs;
143 return NULL;
144}
145
3ce6cd12
AE
146static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
147{
148 if (vxattrs == ceph_dir_vxattrs)
149 return ceph_dir_vxattrs_name_size;
150 if (vxattrs == ceph_file_vxattrs)
151 return ceph_file_vxattrs_name_size;
152 BUG();
153
154 return 0;
155}
156
157/*
158 * Compute the aggregate size (including terminating '\0') of all
159 * virtual extended attribute names in the given vxattr table.
160 */
161static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
162{
163 struct ceph_vxattr *vxattr;
164 size_t size = 0;
165
166 for (vxattr = vxattrs; vxattr->name; vxattr++)
167 size += vxattr->name_size;
168
169 return size;
170}
171
172/* Routines called at initialization and exit time */
173
174void __init ceph_xattr_init(void)
175{
176 ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
177 ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
178}
179
180void ceph_xattr_exit(void)
181{
182 ceph_dir_vxattrs_name_size = 0;
183 ceph_file_vxattrs_name_size = 0;
184}
185
881a5fa2 186static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
355da1eb
SW
187 const char *name)
188{
881a5fa2 189 struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
06476a69
AE
190
191 if (vxattr) {
192 while (vxattr->name) {
193 if (!strcmp(vxattr->name, name))
194 return vxattr;
195 vxattr++;
196 }
197 }
198
355da1eb
SW
199 return NULL;
200}
201
202static int __set_xattr(struct ceph_inode_info *ci,
203 const char *name, int name_len,
204 const char *val, int val_len,
205 int dirty,
206 int should_free_name, int should_free_val,
207 struct ceph_inode_xattr **newxattr)
208{
209 struct rb_node **p;
210 struct rb_node *parent = NULL;
211 struct ceph_inode_xattr *xattr = NULL;
212 int c;
213 int new = 0;
214
215 p = &ci->i_xattrs.index.rb_node;
216 while (*p) {
217 parent = *p;
218 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
219 c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
220 if (c < 0)
221 p = &(*p)->rb_left;
222 else if (c > 0)
223 p = &(*p)->rb_right;
224 else {
225 if (name_len == xattr->name_len)
226 break;
227 else if (name_len < xattr->name_len)
228 p = &(*p)->rb_left;
229 else
230 p = &(*p)->rb_right;
231 }
232 xattr = NULL;
233 }
234
235 if (!xattr) {
236 new = 1;
237 xattr = *newxattr;
238 xattr->name = name;
239 xattr->name_len = name_len;
240 xattr->should_free_name = should_free_name;
241
242 ci->i_xattrs.count++;
243 dout("__set_xattr count=%d\n", ci->i_xattrs.count);
244 } else {
245 kfree(*newxattr);
246 *newxattr = NULL;
247 if (xattr->should_free_val)
248 kfree((void *)xattr->val);
249
250 if (should_free_name) {
251 kfree((void *)name);
252 name = xattr->name;
253 }
254 ci->i_xattrs.names_size -= xattr->name_len;
255 ci->i_xattrs.vals_size -= xattr->val_len;
256 }
355da1eb
SW
257 ci->i_xattrs.names_size += name_len;
258 ci->i_xattrs.vals_size += val_len;
259 if (val)
260 xattr->val = val;
261 else
262 xattr->val = "";
263
264 xattr->val_len = val_len;
265 xattr->dirty = dirty;
266 xattr->should_free_val = (val && should_free_val);
267
268 if (new) {
269 rb_link_node(&xattr->node, parent, p);
270 rb_insert_color(&xattr->node, &ci->i_xattrs.index);
271 dout("__set_xattr_val p=%p\n", p);
272 }
273
274 dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
275 ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);
276
277 return 0;
278}
279
280static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
281 const char *name)
282{
283 struct rb_node **p;
284 struct rb_node *parent = NULL;
285 struct ceph_inode_xattr *xattr = NULL;
17db143f 286 int name_len = strlen(name);
355da1eb
SW
287 int c;
288
289 p = &ci->i_xattrs.index.rb_node;
290 while (*p) {
291 parent = *p;
292 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
293 c = strncmp(name, xattr->name, xattr->name_len);
17db143f
SW
294 if (c == 0 && name_len > xattr->name_len)
295 c = 1;
355da1eb
SW
296 if (c < 0)
297 p = &(*p)->rb_left;
298 else if (c > 0)
299 p = &(*p)->rb_right;
300 else {
301 dout("__get_xattr %s: found %.*s\n", name,
302 xattr->val_len, xattr->val);
303 return xattr;
304 }
305 }
306
307 dout("__get_xattr %s: not found\n", name);
308
309 return NULL;
310}
311
312static void __free_xattr(struct ceph_inode_xattr *xattr)
313{
314 BUG_ON(!xattr);
315
316 if (xattr->should_free_name)
317 kfree((void *)xattr->name);
318 if (xattr->should_free_val)
319 kfree((void *)xattr->val);
320
321 kfree(xattr);
322}
323
324static int __remove_xattr(struct ceph_inode_info *ci,
325 struct ceph_inode_xattr *xattr)
326{
327 if (!xattr)
328 return -EOPNOTSUPP;
329
330 rb_erase(&xattr->node, &ci->i_xattrs.index);
331
332 if (xattr->should_free_name)
333 kfree((void *)xattr->name);
334 if (xattr->should_free_val)
335 kfree((void *)xattr->val);
336
337 ci->i_xattrs.names_size -= xattr->name_len;
338 ci->i_xattrs.vals_size -= xattr->val_len;
339 ci->i_xattrs.count--;
340 kfree(xattr);
341
342 return 0;
343}
344
345static int __remove_xattr_by_name(struct ceph_inode_info *ci,
346 const char *name)
347{
348 struct rb_node **p;
349 struct ceph_inode_xattr *xattr;
350 int err;
351
352 p = &ci->i_xattrs.index.rb_node;
353 xattr = __get_xattr(ci, name);
354 err = __remove_xattr(ci, xattr);
355 return err;
356}
357
358static char *__copy_xattr_names(struct ceph_inode_info *ci,
359 char *dest)
360{
361 struct rb_node *p;
362 struct ceph_inode_xattr *xattr = NULL;
363
364 p = rb_first(&ci->i_xattrs.index);
365 dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
366
367 while (p) {
368 xattr = rb_entry(p, struct ceph_inode_xattr, node);
369 memcpy(dest, xattr->name, xattr->name_len);
370 dest[xattr->name_len] = '\0';
371
372 dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
373 xattr->name_len, ci->i_xattrs.names_size);
374
375 dest += xattr->name_len + 1;
376 p = rb_next(p);
377 }
378
379 return dest;
380}
381
382void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
383{
384 struct rb_node *p, *tmp;
385 struct ceph_inode_xattr *xattr = NULL;
386
387 p = rb_first(&ci->i_xattrs.index);
388
389 dout("__ceph_destroy_xattrs p=%p\n", p);
390
391 while (p) {
392 xattr = rb_entry(p, struct ceph_inode_xattr, node);
393 tmp = p;
394 p = rb_next(tmp);
395 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
396 xattr->name_len, xattr->name);
397 rb_erase(tmp, &ci->i_xattrs.index);
398
399 __free_xattr(xattr);
400 }
401
402 ci->i_xattrs.names_size = 0;
403 ci->i_xattrs.vals_size = 0;
404 ci->i_xattrs.index_version = 0;
405 ci->i_xattrs.count = 0;
406 ci->i_xattrs.index = RB_ROOT;
407}
408
409static int __build_xattrs(struct inode *inode)
be655596
SW
410 __releases(ci->i_ceph_lock)
411 __acquires(ci->i_ceph_lock)
355da1eb
SW
412{
413 u32 namelen;
414 u32 numattr = 0;
415 void *p, *end;
416 u32 len;
417 const char *name, *val;
418 struct ceph_inode_info *ci = ceph_inode(inode);
419 int xattr_version;
420 struct ceph_inode_xattr **xattrs = NULL;
63ff78b2 421 int err = 0;
355da1eb
SW
422 int i;
423
424 dout("__build_xattrs() len=%d\n",
425 ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
426
427 if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
428 return 0; /* already built */
429
430 __ceph_destroy_xattrs(ci);
431
432start:
433 /* updated internal xattr rb tree */
434 if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
435 p = ci->i_xattrs.blob->vec.iov_base;
436 end = p + ci->i_xattrs.blob->vec.iov_len;
437 ceph_decode_32_safe(&p, end, numattr, bad);
438 xattr_version = ci->i_xattrs.version;
be655596 439 spin_unlock(&ci->i_ceph_lock);
355da1eb
SW
440
441 xattrs = kcalloc(numattr, sizeof(struct ceph_xattr *),
442 GFP_NOFS);
443 err = -ENOMEM;
444 if (!xattrs)
445 goto bad_lock;
446 memset(xattrs, 0, numattr*sizeof(struct ceph_xattr *));
447 for (i = 0; i < numattr; i++) {
448 xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
449 GFP_NOFS);
450 if (!xattrs[i])
451 goto bad_lock;
452 }
453
be655596 454 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
455 if (ci->i_xattrs.version != xattr_version) {
456 /* lost a race, retry */
457 for (i = 0; i < numattr; i++)
458 kfree(xattrs[i]);
459 kfree(xattrs);
21ec6ffa 460 xattrs = NULL;
355da1eb
SW
461 goto start;
462 }
463 err = -EIO;
464 while (numattr--) {
465 ceph_decode_32_safe(&p, end, len, bad);
466 namelen = len;
467 name = p;
468 p += len;
469 ceph_decode_32_safe(&p, end, len, bad);
470 val = p;
471 p += len;
472
473 err = __set_xattr(ci, name, namelen, val, len,
474 0, 0, 0, &xattrs[numattr]);
475
476 if (err < 0)
477 goto bad;
478 }
479 kfree(xattrs);
480 }
481 ci->i_xattrs.index_version = ci->i_xattrs.version;
482 ci->i_xattrs.dirty = false;
483
484 return err;
485bad_lock:
be655596 486 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
487bad:
488 if (xattrs) {
489 for (i = 0; i < numattr; i++)
490 kfree(xattrs[i]);
491 kfree(xattrs);
492 }
493 ci->i_xattrs.names_size = 0;
494 return err;
495}
496
497static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
498 int val_size)
499{
500 /*
501 * 4 bytes for the length, and additional 4 bytes per each xattr name,
502 * 4 bytes per each value
503 */
504 int size = 4 + ci->i_xattrs.count*(4 + 4) +
505 ci->i_xattrs.names_size +
506 ci->i_xattrs.vals_size;
507 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
508 ci->i_xattrs.count, ci->i_xattrs.names_size,
509 ci->i_xattrs.vals_size);
510
511 if (name_size)
512 size += 4 + 4 + name_size + val_size;
513
514 return size;
515}
516
517/*
518 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
519 * and swap into place.
520 */
521void __ceph_build_xattrs_blob(struct ceph_inode_info *ci)
522{
523 struct rb_node *p;
524 struct ceph_inode_xattr *xattr = NULL;
525 void *dest;
526
527 dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
528 if (ci->i_xattrs.dirty) {
529 int need = __get_required_blob_size(ci, 0, 0);
530
531 BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
532
533 p = rb_first(&ci->i_xattrs.index);
534 dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
535
536 ceph_encode_32(&dest, ci->i_xattrs.count);
537 while (p) {
538 xattr = rb_entry(p, struct ceph_inode_xattr, node);
539
540 ceph_encode_32(&dest, xattr->name_len);
541 memcpy(dest, xattr->name, xattr->name_len);
542 dest += xattr->name_len;
543 ceph_encode_32(&dest, xattr->val_len);
544 memcpy(dest, xattr->val, xattr->val_len);
545 dest += xattr->val_len;
546
547 p = rb_next(p);
548 }
549
550 /* adjust buffer len; it may be larger than we need */
551 ci->i_xattrs.prealloc_blob->vec.iov_len =
552 dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
553
b6c1d5b8
SW
554 if (ci->i_xattrs.blob)
555 ceph_buffer_put(ci->i_xattrs.blob);
355da1eb
SW
556 ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
557 ci->i_xattrs.prealloc_blob = NULL;
558 ci->i_xattrs.dirty = false;
4a625be4 559 ci->i_xattrs.version++;
355da1eb
SW
560 }
561}
562
563ssize_t ceph_getxattr(struct dentry *dentry, const char *name, void *value,
564 size_t size)
565{
566 struct inode *inode = dentry->d_inode;
567 struct ceph_inode_info *ci = ceph_inode(inode);
355da1eb
SW
568 int err;
569 struct ceph_inode_xattr *xattr;
881a5fa2 570 struct ceph_vxattr *vxattr = NULL;
355da1eb
SW
571
572 if (!ceph_is_valid_xattr(name))
573 return -ENODATA;
574
575 /* let's see if a virtual xattr was requested */
06476a69 576 vxattr = ceph_match_vxattr(inode, name);
355da1eb 577
be655596 578 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
579 dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
580 ci->i_xattrs.version, ci->i_xattrs.index_version);
581
582 if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) &&
583 (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
584 goto get_xattr;
585 } else {
be655596 586 spin_unlock(&ci->i_ceph_lock);
355da1eb
SW
587 /* get xattrs from mds (if we don't already have them) */
588 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
589 if (err)
590 return err;
591 }
592
be655596 593 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
594
595 if (vxattr && vxattr->readonly) {
596 err = vxattr->getxattr_cb(ci, value, size);
597 goto out;
598 }
599
600 err = __build_xattrs(inode);
601 if (err < 0)
602 goto out;
603
604get_xattr:
605 err = -ENODATA; /* == ENOATTR */
606 xattr = __get_xattr(ci, name);
607 if (!xattr) {
608 if (vxattr)
609 err = vxattr->getxattr_cb(ci, value, size);
610 goto out;
611 }
612
613 err = -ERANGE;
614 if (size && size < xattr->val_len)
615 goto out;
616
617 err = xattr->val_len;
618 if (size == 0)
619 goto out;
620
621 memcpy(value, xattr->val, xattr->val_len);
622
623out:
be655596 624 spin_unlock(&ci->i_ceph_lock);
355da1eb
SW
625 return err;
626}
627
628ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
629{
630 struct inode *inode = dentry->d_inode;
631 struct ceph_inode_info *ci = ceph_inode(inode);
881a5fa2 632 struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
355da1eb
SW
633 u32 vir_namelen = 0;
634 u32 namelen;
635 int err;
636 u32 len;
637 int i;
638
be655596 639 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
640 dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
641 ci->i_xattrs.version, ci->i_xattrs.index_version);
642
643 if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) &&
bddfa3cc 644 (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
355da1eb
SW
645 goto list_xattr;
646 } else {
be655596 647 spin_unlock(&ci->i_ceph_lock);
355da1eb
SW
648 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
649 if (err)
650 return err;
651 }
652
be655596 653 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
654
655 err = __build_xattrs(inode);
656 if (err < 0)
657 goto out;
658
659list_xattr:
3ce6cd12
AE
660 /*
661 * Start with virtual dir xattr names (if any) (including
662 * terminating '\0' characters for each).
663 */
664 vir_namelen = ceph_vxattrs_name_size(vxattrs);
665
355da1eb
SW
666 /* adding 1 byte per each variable due to the null termination */
667 namelen = vir_namelen + ci->i_xattrs.names_size + ci->i_xattrs.count;
668 err = -ERANGE;
669 if (size && namelen > size)
670 goto out;
671
672 err = namelen;
673 if (size == 0)
674 goto out;
675
676 names = __copy_xattr_names(ci, names);
677
678 /* virtual xattr names, too */
679 if (vxattrs)
680 for (i = 0; vxattrs[i].name; i++) {
681 len = sprintf(names, "%s", vxattrs[i].name);
682 names += len + 1;
683 }
684
685out:
be655596 686 spin_unlock(&ci->i_ceph_lock);
355da1eb
SW
687 return err;
688}
689
690static int ceph_sync_setxattr(struct dentry *dentry, const char *name,
691 const char *value, size_t size, int flags)
692{
3d14c5d2 693 struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
355da1eb
SW
694 struct inode *inode = dentry->d_inode;
695 struct ceph_inode_info *ci = ceph_inode(inode);
5f21c96d 696 struct inode *parent_inode;
355da1eb 697 struct ceph_mds_request *req;
3d14c5d2 698 struct ceph_mds_client *mdsc = fsc->mdsc;
355da1eb
SW
699 int err;
700 int i, nr_pages;
701 struct page **pages = NULL;
702 void *kaddr;
703
704 /* copy value into some pages */
705 nr_pages = calc_pages_for(0, size);
706 if (nr_pages) {
707 pages = kmalloc(sizeof(pages[0])*nr_pages, GFP_NOFS);
708 if (!pages)
709 return -ENOMEM;
710 err = -ENOMEM;
711 for (i = 0; i < nr_pages; i++) {
31459fe4 712 pages[i] = __page_cache_alloc(GFP_NOFS);
355da1eb
SW
713 if (!pages[i]) {
714 nr_pages = i;
715 goto out;
716 }
717 kaddr = kmap(pages[i]);
718 memcpy(kaddr, value + i*PAGE_CACHE_SIZE,
719 min(PAGE_CACHE_SIZE, size-i*PAGE_CACHE_SIZE));
720 }
721 }
722
723 dout("setxattr value=%.*s\n", (int)size, value);
724
725 /* do request */
726 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETXATTR,
727 USE_AUTH_MDS);
60d87733
JL
728 if (IS_ERR(req)) {
729 err = PTR_ERR(req);
730 goto out;
731 }
70b666c3
SW
732 req->r_inode = inode;
733 ihold(inode);
355da1eb
SW
734 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
735 req->r_num_caps = 1;
736 req->r_args.setxattr.flags = cpu_to_le32(flags);
737 req->r_path2 = kstrdup(name, GFP_NOFS);
738
739 req->r_pages = pages;
740 req->r_num_pages = nr_pages;
741 req->r_data_len = size;
742
743 dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
5f21c96d 744 parent_inode = ceph_get_dentry_parent_inode(dentry);
355da1eb 745 err = ceph_mdsc_do_request(mdsc, parent_inode, req);
5f21c96d 746 iput(parent_inode);
355da1eb
SW
747 ceph_mdsc_put_request(req);
748 dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
749
750out:
751 if (pages) {
752 for (i = 0; i < nr_pages; i++)
753 __free_page(pages[i]);
754 kfree(pages);
755 }
756 return err;
757}
758
759int ceph_setxattr(struct dentry *dentry, const char *name,
760 const void *value, size_t size, int flags)
761{
762 struct inode *inode = dentry->d_inode;
881a5fa2 763 struct ceph_vxattr *vxattr;
355da1eb 764 struct ceph_inode_info *ci = ceph_inode(inode);
18fa8b3f 765 int issued;
355da1eb 766 int err;
18fa8b3f 767 int dirty;
355da1eb
SW
768 int name_len = strlen(name);
769 int val_len = size;
770 char *newname = NULL;
771 char *newval = NULL;
772 struct ceph_inode_xattr *xattr = NULL;
355da1eb
SW
773 int required_blob_size;
774
775 if (ceph_snap(inode) != CEPH_NOSNAP)
776 return -EROFS;
777
778 if (!ceph_is_valid_xattr(name))
779 return -EOPNOTSUPP;
780
06476a69
AE
781 vxattr = ceph_match_vxattr(inode, name);
782 if (vxattr && vxattr->readonly)
783 return -EOPNOTSUPP;
355da1eb
SW
784
785 /* preallocate memory for xattr name, value, index node */
786 err = -ENOMEM;
61413c2f 787 newname = kmemdup(name, name_len + 1, GFP_NOFS);
355da1eb
SW
788 if (!newname)
789 goto out;
355da1eb
SW
790
791 if (val_len) {
b829c195 792 newval = kmemdup(value, val_len, GFP_NOFS);
355da1eb
SW
793 if (!newval)
794 goto out;
355da1eb
SW
795 }
796
797 xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
798 if (!xattr)
799 goto out;
800
be655596 801 spin_lock(&ci->i_ceph_lock);
355da1eb
SW
802retry:
803 issued = __ceph_caps_issued(ci, NULL);
18fa8b3f 804 dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
355da1eb
SW
805 if (!(issued & CEPH_CAP_XATTR_EXCL))
806 goto do_sync;
807 __build_xattrs(inode);
808
809 required_blob_size = __get_required_blob_size(ci, name_len, val_len);
810
811 if (!ci->i_xattrs.prealloc_blob ||
812 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
18fa8b3f 813 struct ceph_buffer *blob;
355da1eb 814
be655596 815 spin_unlock(&ci->i_ceph_lock);
355da1eb 816 dout(" preaallocating new blob size=%d\n", required_blob_size);
b6c1d5b8 817 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
355da1eb
SW
818 if (!blob)
819 goto out;
be655596 820 spin_lock(&ci->i_ceph_lock);
b6c1d5b8
SW
821 if (ci->i_xattrs.prealloc_blob)
822 ceph_buffer_put(ci->i_xattrs.prealloc_blob);
355da1eb
SW
823 ci->i_xattrs.prealloc_blob = blob;
824 goto retry;
825 }
826
355da1eb
SW
827 err = __set_xattr(ci, newname, name_len, newval,
828 val_len, 1, 1, 1, &xattr);
18fa8b3f 829
fca65b4a 830 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
355da1eb
SW
831 ci->i_xattrs.dirty = true;
832 inode->i_ctime = CURRENT_TIME;
18fa8b3f 833
be655596 834 spin_unlock(&ci->i_ceph_lock);
fca65b4a
SW
835 if (dirty)
836 __mark_inode_dirty(inode, dirty);
355da1eb
SW
837 return err;
838
839do_sync:
be655596 840 spin_unlock(&ci->i_ceph_lock);
355da1eb
SW
841 err = ceph_sync_setxattr(dentry, name, value, size, flags);
842out:
843 kfree(newname);
844 kfree(newval);
845 kfree(xattr);
846 return err;
847}
848
849static int ceph_send_removexattr(struct dentry *dentry, const char *name)
850{
3d14c5d2
YS
851 struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
852 struct ceph_mds_client *mdsc = fsc->mdsc;
355da1eb 853 struct inode *inode = dentry->d_inode;
5f21c96d 854 struct inode *parent_inode;
355da1eb
SW
855 struct ceph_mds_request *req;
856 int err;
857
858 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_RMXATTR,
859 USE_AUTH_MDS);
860 if (IS_ERR(req))
861 return PTR_ERR(req);
70b666c3
SW
862 req->r_inode = inode;
863 ihold(inode);
355da1eb
SW
864 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
865 req->r_num_caps = 1;
866 req->r_path2 = kstrdup(name, GFP_NOFS);
867
5f21c96d 868 parent_inode = ceph_get_dentry_parent_inode(dentry);
355da1eb 869 err = ceph_mdsc_do_request(mdsc, parent_inode, req);
5f21c96d 870 iput(parent_inode);
355da1eb
SW
871 ceph_mdsc_put_request(req);
872 return err;
873}
874
875int ceph_removexattr(struct dentry *dentry, const char *name)
876{
877 struct inode *inode = dentry->d_inode;
881a5fa2 878 struct ceph_vxattr *vxattr;
355da1eb 879 struct ceph_inode_info *ci = ceph_inode(inode);
355da1eb
SW
880 int issued;
881 int err;
83eb26af 882 int required_blob_size;
fca65b4a 883 int dirty;
355da1eb
SW
884
885 if (ceph_snap(inode) != CEPH_NOSNAP)
886 return -EROFS;
887
888 if (!ceph_is_valid_xattr(name))
889 return -EOPNOTSUPP;
890
06476a69
AE
891 vxattr = ceph_match_vxattr(inode, name);
892 if (vxattr && vxattr->readonly)
893 return -EOPNOTSUPP;
355da1eb 894
83eb26af 895 err = -ENOMEM;
be655596 896 spin_lock(&ci->i_ceph_lock);
83eb26af 897retry:
355da1eb
SW
898 issued = __ceph_caps_issued(ci, NULL);
899 dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued));
900
901 if (!(issued & CEPH_CAP_XATTR_EXCL))
902 goto do_sync;
18fa8b3f 903 __build_xattrs(inode);
355da1eb 904
83eb26af
AE
905 required_blob_size = __get_required_blob_size(ci, 0, 0);
906
907 if (!ci->i_xattrs.prealloc_blob ||
908 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
909 struct ceph_buffer *blob;
910
911 spin_unlock(&ci->i_ceph_lock);
912 dout(" preaallocating new blob size=%d\n", required_blob_size);
913 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
914 if (!blob)
915 goto out;
916 spin_lock(&ci->i_ceph_lock);
917 if (ci->i_xattrs.prealloc_blob)
918 ceph_buffer_put(ci->i_xattrs.prealloc_blob);
919 ci->i_xattrs.prealloc_blob = blob;
920 goto retry;
921 }
922
355da1eb 923 err = __remove_xattr_by_name(ceph_inode(inode), name);
18fa8b3f 924
fca65b4a 925 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
355da1eb
SW
926 ci->i_xattrs.dirty = true;
927 inode->i_ctime = CURRENT_TIME;
be655596 928 spin_unlock(&ci->i_ceph_lock);
fca65b4a
SW
929 if (dirty)
930 __mark_inode_dirty(inode, dirty);
355da1eb
SW
931 return err;
932do_sync:
be655596 933 spin_unlock(&ci->i_ceph_lock);
355da1eb 934 err = ceph_send_removexattr(dentry, name);
83eb26af 935out:
355da1eb
SW
936 return err;
937}
938