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)
45 #define VXATTR_FLAG_DIRSTAT (1<<3)
49 static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
51 struct ceph_file_layout *fl = &ci->i_layout;
52 return (fl->stripe_unit > 0 || fl->stripe_count > 0 ||
53 fl->object_size > 0 || fl->pool_id >= 0 ||
54 rcu_dereference_raw(fl->pool_ns) != NULL);
57 static ssize_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
60 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
61 struct ceph_osd_client *osdc = &fsc->client->osdc;
62 struct ceph_string *pool_ns;
63 s64 pool = ci->i_layout.pool_id;
64 const char *pool_name;
65 const char *ns_field = " pool_namespace=";
67 size_t len, total_len = 0;
70 pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
72 dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
73 down_read(&osdc->lock);
74 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
76 len = snprintf(buf, sizeof(buf),
77 "stripe_unit=%u stripe_count=%u object_size=%u pool=",
78 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
79 ci->i_layout.object_size);
80 total_len = len + strlen(pool_name);
82 len = snprintf(buf, sizeof(buf),
83 "stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
84 ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
85 ci->i_layout.object_size, pool);
90 total_len += strlen(ns_field) + pool_ns->len;
93 if (size >= total_len) {
94 memcpy(val, buf, len);
97 len = strlen(pool_name);
98 memcpy(val + ret, pool_name, len);
102 len = strlen(ns_field);
103 memcpy(val + ret, ns_field, len);
105 memcpy(val + ret, pool_ns->str, pool_ns->len);
109 up_read(&osdc->lock);
110 ceph_put_string(pool_ns);
115 * The convention with strings in xattrs is that they should not be NULL
116 * terminated, since we're returning the length with them. snprintf always
117 * NULL terminates however, so call it on a temporary buffer and then memcpy
118 * the result into place.
120 static __printf(3, 4)
121 int ceph_fmt_xattr(char *val, size_t size, const char *fmt, ...)
125 char buf[96]; /* NB: reevaluate size if new vxattrs are added */
128 ret = vsnprintf(buf, size ? sizeof(buf) : 0, fmt, args);
132 if (size && ret + 1 > sizeof(buf)) {
133 WARN_ONCE(true, "Returned length too big (%d)", ret);
138 memcpy(val, buf, ret);
142 static ssize_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
143 char *val, size_t size)
145 return ceph_fmt_xattr(val, size, "%u", ci->i_layout.stripe_unit);
148 static ssize_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
149 char *val, size_t size)
151 return ceph_fmt_xattr(val, size, "%u", ci->i_layout.stripe_count);
154 static ssize_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
155 char *val, size_t size)
157 return ceph_fmt_xattr(val, size, "%u", ci->i_layout.object_size);
160 static ssize_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
161 char *val, size_t size)
164 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
165 struct ceph_osd_client *osdc = &fsc->client->osdc;
166 s64 pool = ci->i_layout.pool_id;
167 const char *pool_name;
169 down_read(&osdc->lock);
170 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
172 ret = strlen(pool_name);
174 memcpy(val, pool_name, ret);
176 ret = ceph_fmt_xattr(val, size, "%lld", pool);
178 up_read(&osdc->lock);
182 static ssize_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info *ci,
183 char *val, size_t size)
186 struct ceph_string *ns = ceph_try_get_string(ci->i_layout.pool_ns);
191 memcpy(val, ns->str, ret);
199 static ssize_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
202 return ceph_fmt_xattr(val, size, "%lld", ci->i_files + ci->i_subdirs);
205 static ssize_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
208 return ceph_fmt_xattr(val, size, "%lld", ci->i_files);
211 static ssize_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
214 return ceph_fmt_xattr(val, size, "%lld", ci->i_subdirs);
217 static ssize_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
220 return ceph_fmt_xattr(val, size, "%lld",
221 ci->i_rfiles + ci->i_rsubdirs);
224 static ssize_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
227 return ceph_fmt_xattr(val, size, "%lld", ci->i_rfiles);
230 static ssize_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
233 return ceph_fmt_xattr(val, size, "%lld", ci->i_rsubdirs);
236 static ssize_t ceph_vxattrcb_dir_rsnaps(struct ceph_inode_info *ci, char *val,
239 return ceph_fmt_xattr(val, size, "%lld", ci->i_rsnaps);
242 static ssize_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
245 return ceph_fmt_xattr(val, size, "%lld", ci->i_rbytes);
248 static ssize_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
251 return ceph_fmt_xattr(val, size, "%lld.%09ld", ci->i_rctime.tv_sec,
252 ci->i_rctime.tv_nsec);
256 static bool ceph_vxattrcb_dir_pin_exists(struct ceph_inode_info *ci)
258 return ci->i_dir_pin != -ENODATA;
261 static ssize_t ceph_vxattrcb_dir_pin(struct ceph_inode_info *ci, char *val,
264 return ceph_fmt_xattr(val, size, "%d", (int)ci->i_dir_pin);
268 static bool ceph_vxattrcb_quota_exists(struct ceph_inode_info *ci)
271 spin_lock(&ci->i_ceph_lock);
272 if ((ci->i_max_files || ci->i_max_bytes) &&
273 ci->i_vino.snap == CEPH_NOSNAP &&
275 ci->i_snap_realm->ino == ci->i_vino.ino)
277 spin_unlock(&ci->i_ceph_lock);
281 static ssize_t ceph_vxattrcb_quota(struct ceph_inode_info *ci, char *val,
284 return ceph_fmt_xattr(val, size, "max_bytes=%llu max_files=%llu",
285 ci->i_max_bytes, ci->i_max_files);
288 static ssize_t ceph_vxattrcb_quota_max_bytes(struct ceph_inode_info *ci,
289 char *val, size_t size)
291 return ceph_fmt_xattr(val, size, "%llu", ci->i_max_bytes);
294 static ssize_t ceph_vxattrcb_quota_max_files(struct ceph_inode_info *ci,
295 char *val, size_t size)
297 return ceph_fmt_xattr(val, size, "%llu", ci->i_max_files);
301 static bool ceph_vxattrcb_snap_btime_exists(struct ceph_inode_info *ci)
303 return (ci->i_snap_btime.tv_sec != 0 || ci->i_snap_btime.tv_nsec != 0);
306 static ssize_t ceph_vxattrcb_snap_btime(struct ceph_inode_info *ci, char *val,
309 return ceph_fmt_xattr(val, size, "%lld.%09ld", ci->i_snap_btime.tv_sec,
310 ci->i_snap_btime.tv_nsec);
313 static ssize_t ceph_vxattrcb_cluster_fsid(struct ceph_inode_info *ci,
314 char *val, size_t size)
316 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
318 return ceph_fmt_xattr(val, size, "%pU", &fsc->client->fsid);
321 static ssize_t ceph_vxattrcb_client_id(struct ceph_inode_info *ci,
322 char *val, size_t size)
324 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
326 return ceph_fmt_xattr(val, size, "client%lld",
327 ceph_client_gid(fsc->client));
330 static ssize_t ceph_vxattrcb_caps(struct ceph_inode_info *ci, char *val,
335 spin_lock(&ci->i_ceph_lock);
336 issued = __ceph_caps_issued(ci, NULL);
337 spin_unlock(&ci->i_ceph_lock);
339 return ceph_fmt_xattr(val, size, "%s/0x%x",
340 ceph_cap_string(issued), issued);
343 #define CEPH_XATTR_NAME(_type, _name) XATTR_CEPH_PREFIX #_type "." #_name
344 #define CEPH_XATTR_NAME2(_type, _name, _name2) \
345 XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
347 #define XATTR_NAME_CEPH(_type, _name, _flags) \
349 .name = CEPH_XATTR_NAME(_type, _name), \
350 .name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
351 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
353 .flags = (VXATTR_FLAG_READONLY | _flags), \
355 #define XATTR_RSTAT_FIELD(_type, _name) \
356 XATTR_NAME_CEPH(_type, _name, VXATTR_FLAG_RSTAT)
357 #define XATTR_LAYOUT_FIELD(_type, _name, _field) \
359 .name = CEPH_XATTR_NAME2(_type, _name, _field), \
360 .name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
361 .getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
362 .exists_cb = ceph_vxattrcb_layout_exists, \
363 .flags = VXATTR_FLAG_HIDDEN, \
365 #define XATTR_QUOTA_FIELD(_type, _name) \
367 .name = CEPH_XATTR_NAME(_type, _name), \
368 .name_size = sizeof(CEPH_XATTR_NAME(_type, _name)), \
369 .getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
370 .exists_cb = ceph_vxattrcb_quota_exists, \
371 .flags = VXATTR_FLAG_HIDDEN, \
374 static struct ceph_vxattr ceph_dir_vxattrs[] = {
376 .name = "ceph.dir.layout",
377 .name_size = sizeof("ceph.dir.layout"),
378 .getxattr_cb = ceph_vxattrcb_layout,
379 .exists_cb = ceph_vxattrcb_layout_exists,
380 .flags = VXATTR_FLAG_HIDDEN,
382 XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
383 XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
384 XATTR_LAYOUT_FIELD(dir, layout, object_size),
385 XATTR_LAYOUT_FIELD(dir, layout, pool),
386 XATTR_LAYOUT_FIELD(dir, layout, pool_namespace),
387 XATTR_NAME_CEPH(dir, entries, VXATTR_FLAG_DIRSTAT),
388 XATTR_NAME_CEPH(dir, files, VXATTR_FLAG_DIRSTAT),
389 XATTR_NAME_CEPH(dir, subdirs, VXATTR_FLAG_DIRSTAT),
390 XATTR_RSTAT_FIELD(dir, rentries),
391 XATTR_RSTAT_FIELD(dir, rfiles),
392 XATTR_RSTAT_FIELD(dir, rsubdirs),
393 XATTR_RSTAT_FIELD(dir, rsnaps),
394 XATTR_RSTAT_FIELD(dir, rbytes),
395 XATTR_RSTAT_FIELD(dir, rctime),
397 .name = "ceph.dir.pin",
398 .name_size = sizeof("ceph.dir.pin"),
399 .getxattr_cb = ceph_vxattrcb_dir_pin,
400 .exists_cb = ceph_vxattrcb_dir_pin_exists,
401 .flags = VXATTR_FLAG_HIDDEN,
404 .name = "ceph.quota",
405 .name_size = sizeof("ceph.quota"),
406 .getxattr_cb = ceph_vxattrcb_quota,
407 .exists_cb = ceph_vxattrcb_quota_exists,
408 .flags = VXATTR_FLAG_HIDDEN,
410 XATTR_QUOTA_FIELD(quota, max_bytes),
411 XATTR_QUOTA_FIELD(quota, max_files),
413 .name = "ceph.snap.btime",
414 .name_size = sizeof("ceph.snap.btime"),
415 .getxattr_cb = ceph_vxattrcb_snap_btime,
416 .exists_cb = ceph_vxattrcb_snap_btime_exists,
417 .flags = VXATTR_FLAG_READONLY,
421 .name_size = sizeof("ceph.caps"),
422 .getxattr_cb = ceph_vxattrcb_caps,
424 .flags = VXATTR_FLAG_HIDDEN,
426 { .name = NULL, 0 } /* Required table terminator */
431 static struct ceph_vxattr ceph_file_vxattrs[] = {
433 .name = "ceph.file.layout",
434 .name_size = sizeof("ceph.file.layout"),
435 .getxattr_cb = ceph_vxattrcb_layout,
436 .exists_cb = ceph_vxattrcb_layout_exists,
437 .flags = VXATTR_FLAG_HIDDEN,
439 XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
440 XATTR_LAYOUT_FIELD(file, layout, stripe_count),
441 XATTR_LAYOUT_FIELD(file, layout, object_size),
442 XATTR_LAYOUT_FIELD(file, layout, pool),
443 XATTR_LAYOUT_FIELD(file, layout, pool_namespace),
445 .name = "ceph.snap.btime",
446 .name_size = sizeof("ceph.snap.btime"),
447 .getxattr_cb = ceph_vxattrcb_snap_btime,
448 .exists_cb = ceph_vxattrcb_snap_btime_exists,
449 .flags = VXATTR_FLAG_READONLY,
453 .name_size = sizeof("ceph.caps"),
454 .getxattr_cb = ceph_vxattrcb_caps,
456 .flags = VXATTR_FLAG_HIDDEN,
458 { .name = NULL, 0 } /* Required table terminator */
461 static struct ceph_vxattr ceph_common_vxattrs[] = {
463 .name = "ceph.cluster_fsid",
464 .name_size = sizeof("ceph.cluster_fsid"),
465 .getxattr_cb = ceph_vxattrcb_cluster_fsid,
467 .flags = VXATTR_FLAG_READONLY,
470 .name = "ceph.client_id",
471 .name_size = sizeof("ceph.client_id"),
472 .getxattr_cb = ceph_vxattrcb_client_id,
474 .flags = VXATTR_FLAG_READONLY,
476 { .name = NULL, 0 } /* Required table terminator */
479 static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
481 if (S_ISDIR(inode->i_mode))
482 return ceph_dir_vxattrs;
483 else if (S_ISREG(inode->i_mode))
484 return ceph_file_vxattrs;
488 static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
491 struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
494 while (vxattr->name) {
495 if (!strcmp(vxattr->name, name))
501 vxattr = ceph_common_vxattrs;
502 while (vxattr->name) {
503 if (!strcmp(vxattr->name, name))
511 static int __set_xattr(struct ceph_inode_info *ci,
512 const char *name, int name_len,
513 const char *val, int val_len,
514 int flags, int update_xattr,
515 struct ceph_inode_xattr **newxattr)
518 struct rb_node *parent = NULL;
519 struct ceph_inode_xattr *xattr = NULL;
523 p = &ci->i_xattrs.index.rb_node;
526 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
527 c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
533 if (name_len == xattr->name_len)
535 else if (name_len < xattr->name_len)
546 if (xattr && (flags & XATTR_CREATE))
548 else if (!xattr && (flags & XATTR_REPLACE))
556 if (update_xattr < 0) {
558 __remove_xattr(ci, xattr);
569 xattr->name_len = name_len;
570 xattr->should_free_name = update_xattr;
572 ci->i_xattrs.count++;
573 dout("__set_xattr count=%d\n", ci->i_xattrs.count);
577 if (xattr->should_free_val)
584 ci->i_xattrs.names_size -= xattr->name_len;
585 ci->i_xattrs.vals_size -= xattr->val_len;
587 ci->i_xattrs.names_size += name_len;
588 ci->i_xattrs.vals_size += val_len;
594 xattr->val_len = val_len;
595 xattr->dirty = update_xattr;
596 xattr->should_free_val = (val && update_xattr);
599 rb_link_node(&xattr->node, parent, p);
600 rb_insert_color(&xattr->node, &ci->i_xattrs.index);
601 dout("__set_xattr_val p=%p\n", p);
604 dout("__set_xattr_val added %llx.%llx xattr %p %.*s=%.*s\n",
605 ceph_vinop(&ci->vfs_inode), xattr, name_len, name, val_len, val);
610 static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
614 struct rb_node *parent = NULL;
615 struct ceph_inode_xattr *xattr = NULL;
616 int name_len = strlen(name);
619 p = &ci->i_xattrs.index.rb_node;
622 xattr = rb_entry(parent, struct ceph_inode_xattr, node);
623 c = strncmp(name, xattr->name, xattr->name_len);
624 if (c == 0 && name_len > xattr->name_len)
631 dout("__get_xattr %s: found %.*s\n", name,
632 xattr->val_len, xattr->val);
637 dout("__get_xattr %s: not found\n", name);
642 static void __free_xattr(struct ceph_inode_xattr *xattr)
646 if (xattr->should_free_name)
648 if (xattr->should_free_val)
654 static int __remove_xattr(struct ceph_inode_info *ci,
655 struct ceph_inode_xattr *xattr)
660 rb_erase(&xattr->node, &ci->i_xattrs.index);
662 if (xattr->should_free_name)
664 if (xattr->should_free_val)
667 ci->i_xattrs.names_size -= xattr->name_len;
668 ci->i_xattrs.vals_size -= xattr->val_len;
669 ci->i_xattrs.count--;
675 static char *__copy_xattr_names(struct ceph_inode_info *ci,
679 struct ceph_inode_xattr *xattr = NULL;
681 p = rb_first(&ci->i_xattrs.index);
682 dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
685 xattr = rb_entry(p, struct ceph_inode_xattr, node);
686 memcpy(dest, xattr->name, xattr->name_len);
687 dest[xattr->name_len] = '\0';
689 dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
690 xattr->name_len, ci->i_xattrs.names_size);
692 dest += xattr->name_len + 1;
699 void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
701 struct rb_node *p, *tmp;
702 struct ceph_inode_xattr *xattr = NULL;
704 p = rb_first(&ci->i_xattrs.index);
706 dout("__ceph_destroy_xattrs p=%p\n", p);
709 xattr = rb_entry(p, struct ceph_inode_xattr, node);
712 dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
713 xattr->name_len, xattr->name);
714 rb_erase(tmp, &ci->i_xattrs.index);
719 ci->i_xattrs.names_size = 0;
720 ci->i_xattrs.vals_size = 0;
721 ci->i_xattrs.index_version = 0;
722 ci->i_xattrs.count = 0;
723 ci->i_xattrs.index = RB_ROOT;
726 static int __build_xattrs(struct inode *inode)
727 __releases(ci->i_ceph_lock)
728 __acquires(ci->i_ceph_lock)
734 const char *name, *val;
735 struct ceph_inode_info *ci = ceph_inode(inode);
737 struct ceph_inode_xattr **xattrs = NULL;
741 dout("__build_xattrs() len=%d\n",
742 ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
744 if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
745 return 0; /* already built */
747 __ceph_destroy_xattrs(ci);
750 /* updated internal xattr rb tree */
751 if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
752 p = ci->i_xattrs.blob->vec.iov_base;
753 end = p + ci->i_xattrs.blob->vec.iov_len;
754 ceph_decode_32_safe(&p, end, numattr, bad);
755 xattr_version = ci->i_xattrs.version;
756 spin_unlock(&ci->i_ceph_lock);
758 xattrs = kcalloc(numattr, sizeof(struct ceph_inode_xattr *),
764 for (i = 0; i < numattr; i++) {
765 xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
771 spin_lock(&ci->i_ceph_lock);
772 if (ci->i_xattrs.version != xattr_version) {
773 /* lost a race, retry */
774 for (i = 0; i < numattr; i++)
782 ceph_decode_32_safe(&p, end, len, bad);
786 ceph_decode_32_safe(&p, end, len, bad);
790 err = __set_xattr(ci, name, namelen, val, len,
791 0, 0, &xattrs[numattr]);
798 ci->i_xattrs.index_version = ci->i_xattrs.version;
799 ci->i_xattrs.dirty = false;
803 spin_lock(&ci->i_ceph_lock);
806 for (i = 0; i < numattr; i++)
810 ci->i_xattrs.names_size = 0;
814 static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
818 * 4 bytes for the length, and additional 4 bytes per each xattr name,
819 * 4 bytes per each value
821 int size = 4 + ci->i_xattrs.count*(4 + 4) +
822 ci->i_xattrs.names_size +
823 ci->i_xattrs.vals_size;
824 dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
825 ci->i_xattrs.count, ci->i_xattrs.names_size,
826 ci->i_xattrs.vals_size);
829 size += 4 + 4 + name_size + val_size;
835 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
836 * and swap into place. It returns the old i_xattrs.blob (or NULL) so
837 * that it can be freed by the caller as the i_ceph_lock is likely to be
840 struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci)
843 struct ceph_inode_xattr *xattr = NULL;
844 struct ceph_buffer *old_blob = NULL;
847 dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
848 if (ci->i_xattrs.dirty) {
849 int need = __get_required_blob_size(ci, 0, 0);
851 BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
853 p = rb_first(&ci->i_xattrs.index);
854 dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
856 ceph_encode_32(&dest, ci->i_xattrs.count);
858 xattr = rb_entry(p, struct ceph_inode_xattr, node);
860 ceph_encode_32(&dest, xattr->name_len);
861 memcpy(dest, xattr->name, xattr->name_len);
862 dest += xattr->name_len;
863 ceph_encode_32(&dest, xattr->val_len);
864 memcpy(dest, xattr->val, xattr->val_len);
865 dest += xattr->val_len;
870 /* adjust buffer len; it may be larger than we need */
871 ci->i_xattrs.prealloc_blob->vec.iov_len =
872 dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
874 if (ci->i_xattrs.blob)
875 old_blob = ci->i_xattrs.blob;
876 ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
877 ci->i_xattrs.prealloc_blob = NULL;
878 ci->i_xattrs.dirty = false;
879 ci->i_xattrs.version++;
885 static inline int __get_request_mask(struct inode *in) {
886 struct ceph_mds_request *req = current->journal_info;
888 if (req && req->r_target_inode == in) {
889 if (req->r_op == CEPH_MDS_OP_LOOKUP ||
890 req->r_op == CEPH_MDS_OP_LOOKUPINO ||
891 req->r_op == CEPH_MDS_OP_LOOKUPPARENT ||
892 req->r_op == CEPH_MDS_OP_GETATTR) {
893 mask = le32_to_cpu(req->r_args.getattr.mask);
894 } else if (req->r_op == CEPH_MDS_OP_OPEN ||
895 req->r_op == CEPH_MDS_OP_CREATE) {
896 mask = le32_to_cpu(req->r_args.open.mask);
902 ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
905 struct ceph_inode_info *ci = ceph_inode(inode);
906 struct ceph_inode_xattr *xattr;
907 struct ceph_vxattr *vxattr = NULL;
911 /* let's see if a virtual xattr was requested */
912 vxattr = ceph_match_vxattr(inode, name);
915 if (vxattr->flags & VXATTR_FLAG_RSTAT)
916 mask |= CEPH_STAT_RSTAT;
917 if (vxattr->flags & VXATTR_FLAG_DIRSTAT)
918 mask |= CEPH_CAP_FILE_SHARED;
919 err = ceph_do_getattr(inode, mask, true);
923 if (!(vxattr->exists_cb && !vxattr->exists_cb(ci))) {
924 err = vxattr->getxattr_cb(ci, value, size);
925 if (size && size < err)
931 req_mask = __get_request_mask(inode);
933 spin_lock(&ci->i_ceph_lock);
934 dout("getxattr %p name '%s' ver=%lld index_ver=%lld\n", inode, name,
935 ci->i_xattrs.version, ci->i_xattrs.index_version);
937 if (ci->i_xattrs.version == 0 ||
938 !((req_mask & CEPH_CAP_XATTR_SHARED) ||
939 __ceph_caps_issued_mask_metric(ci, CEPH_CAP_XATTR_SHARED, 1))) {
940 spin_unlock(&ci->i_ceph_lock);
942 /* security module gets xattr while filling trace */
943 if (current->journal_info) {
944 pr_warn_ratelimited("sync getxattr %p "
945 "during filling trace\n", inode);
949 /* get xattrs from mds (if we don't already have them) */
950 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
953 spin_lock(&ci->i_ceph_lock);
956 err = __build_xattrs(inode);
960 err = -ENODATA; /* == ENOATTR */
961 xattr = __get_xattr(ci, name);
966 if (size && size < xattr->val_len)
969 err = xattr->val_len;
973 memcpy(value, xattr->val, xattr->val_len);
975 if (current->journal_info &&
976 !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) &&
977 security_ismaclabel(name + XATTR_SECURITY_PREFIX_LEN))
978 ci->i_ceph_flags |= CEPH_I_SEC_INITED;
980 spin_unlock(&ci->i_ceph_lock);
984 ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
986 struct inode *inode = d_inode(dentry);
987 struct ceph_inode_info *ci = ceph_inode(inode);
988 bool len_only = (size == 0);
992 spin_lock(&ci->i_ceph_lock);
993 dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
994 ci->i_xattrs.version, ci->i_xattrs.index_version);
996 if (ci->i_xattrs.version == 0 ||
997 !__ceph_caps_issued_mask_metric(ci, CEPH_CAP_XATTR_SHARED, 1)) {
998 spin_unlock(&ci->i_ceph_lock);
999 err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
1002 spin_lock(&ci->i_ceph_lock);
1005 err = __build_xattrs(inode);
1009 /* add 1 byte for each xattr due to the null termination */
1010 namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
1012 if (namelen > size) {
1016 names = __copy_xattr_names(ci, names);
1021 spin_unlock(&ci->i_ceph_lock);
1025 static int ceph_sync_setxattr(struct inode *inode, const char *name,
1026 const char *value, size_t size, int flags)
1028 struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
1029 struct ceph_inode_info *ci = ceph_inode(inode);
1030 struct ceph_mds_request *req;
1031 struct ceph_mds_client *mdsc = fsc->mdsc;
1032 struct ceph_osd_client *osdc = &fsc->client->osdc;
1033 struct ceph_pagelist *pagelist = NULL;
1034 int op = CEPH_MDS_OP_SETXATTR;
1038 /* copy value into pagelist */
1039 pagelist = ceph_pagelist_alloc(GFP_NOFS);
1043 err = ceph_pagelist_append(pagelist, value, size);
1046 } else if (!value) {
1047 if (flags & CEPH_XATTR_REPLACE)
1048 op = CEPH_MDS_OP_RMXATTR;
1050 flags |= CEPH_XATTR_REMOVE;
1053 dout("setxattr value=%.*s\n", (int)size, value);
1056 req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
1062 req->r_path2 = kstrdup(name, GFP_NOFS);
1063 if (!req->r_path2) {
1064 ceph_mdsc_put_request(req);
1069 if (op == CEPH_MDS_OP_SETXATTR) {
1070 req->r_args.setxattr.flags = cpu_to_le32(flags);
1071 req->r_args.setxattr.osdmap_epoch =
1072 cpu_to_le32(osdc->osdmap->epoch);
1073 req->r_pagelist = pagelist;
1077 req->r_inode = inode;
1079 req->r_num_caps = 1;
1080 req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
1082 dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
1083 err = ceph_mdsc_do_request(mdsc, NULL, req);
1084 ceph_mdsc_put_request(req);
1085 dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
1089 ceph_pagelist_release(pagelist);
1093 int __ceph_setxattr(struct inode *inode, const char *name,
1094 const void *value, size_t size, int flags)
1096 struct ceph_vxattr *vxattr;
1097 struct ceph_inode_info *ci = ceph_inode(inode);
1098 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
1099 struct ceph_cap_flush *prealloc_cf = NULL;
1100 struct ceph_buffer *old_blob = NULL;
1104 int name_len = strlen(name);
1106 char *newname = NULL;
1107 char *newval = NULL;
1108 struct ceph_inode_xattr *xattr = NULL;
1109 int required_blob_size;
1110 bool check_realm = false;
1111 bool lock_snap_rwsem = false;
1113 if (ceph_snap(inode) != CEPH_NOSNAP)
1116 vxattr = ceph_match_vxattr(inode, name);
1118 if (vxattr->flags & VXATTR_FLAG_READONLY)
1120 if (value && !strncmp(vxattr->name, "ceph.quota", 10))
1124 /* pass any unhandled ceph.* xattrs through to the MDS */
1125 if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
1126 goto do_sync_unlocked;
1128 /* preallocate memory for xattr name, value, index node */
1130 newname = kmemdup(name, name_len + 1, GFP_NOFS);
1135 newval = kmemdup(value, val_len, GFP_NOFS);
1140 xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
1144 prealloc_cf = ceph_alloc_cap_flush();
1148 spin_lock(&ci->i_ceph_lock);
1150 issued = __ceph_caps_issued(ci, NULL);
1151 if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
1154 if (!lock_snap_rwsem && !ci->i_head_snapc) {
1155 lock_snap_rwsem = true;
1156 if (!down_read_trylock(&mdsc->snap_rwsem)) {
1157 spin_unlock(&ci->i_ceph_lock);
1158 down_read(&mdsc->snap_rwsem);
1159 spin_lock(&ci->i_ceph_lock);
1164 dout("setxattr %p name '%s' issued %s\n", inode, name,
1165 ceph_cap_string(issued));
1166 __build_xattrs(inode);
1168 required_blob_size = __get_required_blob_size(ci, name_len, val_len);
1170 if (!ci->i_xattrs.prealloc_blob ||
1171 required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1172 struct ceph_buffer *blob;
1174 spin_unlock(&ci->i_ceph_lock);
1175 ceph_buffer_put(old_blob); /* Shouldn't be required */
1176 dout(" pre-allocating new blob size=%d\n", required_blob_size);
1177 blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1179 goto do_sync_unlocked;
1180 spin_lock(&ci->i_ceph_lock);
1181 /* prealloc_blob can't be released while holding i_ceph_lock */
1182 if (ci->i_xattrs.prealloc_blob)
1183 old_blob = ci->i_xattrs.prealloc_blob;
1184 ci->i_xattrs.prealloc_blob = blob;
1188 err = __set_xattr(ci, newname, name_len, newval, val_len,
1189 flags, value ? 1 : -1, &xattr);
1192 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL,
1194 ci->i_xattrs.dirty = true;
1195 inode->i_ctime = current_time(inode);
1198 spin_unlock(&ci->i_ceph_lock);
1199 ceph_buffer_put(old_blob);
1200 if (lock_snap_rwsem)
1201 up_read(&mdsc->snap_rwsem);
1203 __mark_inode_dirty(inode, dirty);
1204 ceph_free_cap_flush(prealloc_cf);
1208 spin_unlock(&ci->i_ceph_lock);
1210 if (lock_snap_rwsem)
1211 up_read(&mdsc->snap_rwsem);
1213 /* security module set xattr while filling trace */
1214 if (current->journal_info) {
1215 pr_warn_ratelimited("sync setxattr %p "
1216 "during filling trace\n", inode);
1219 err = ceph_sync_setxattr(inode, name, value, size, flags);
1220 if (err >= 0 && check_realm) {
1221 /* check if snaprealm was created for quota inode */
1222 spin_lock(&ci->i_ceph_lock);
1223 if ((ci->i_max_files || ci->i_max_bytes) &&
1224 !(ci->i_snap_realm &&
1225 ci->i_snap_realm->ino == ci->i_vino.ino))
1227 spin_unlock(&ci->i_ceph_lock);
1231 ceph_free_cap_flush(prealloc_cf);
1238 static int ceph_get_xattr_handler(const struct xattr_handler *handler,
1239 struct dentry *dentry, struct inode *inode,
1240 const char *name, void *value, size_t size)
1242 if (!ceph_is_valid_xattr(name))
1244 return __ceph_getxattr(inode, name, value, size);
1247 static int ceph_set_xattr_handler(const struct xattr_handler *handler,
1248 struct user_namespace *mnt_userns,
1249 struct dentry *unused, struct inode *inode,
1250 const char *name, const void *value,
1251 size_t size, int flags)
1253 if (!ceph_is_valid_xattr(name))
1255 return __ceph_setxattr(inode, name, value, size, flags);
1258 static const struct xattr_handler ceph_other_xattr_handler = {
1259 .prefix = "", /* match any name => handlers called with full name */
1260 .get = ceph_get_xattr_handler,
1261 .set = ceph_set_xattr_handler,
1264 #ifdef CONFIG_SECURITY
1265 bool ceph_security_xattr_wanted(struct inode *in)
1267 return in->i_security != NULL;
1270 bool ceph_security_xattr_deadlock(struct inode *in)
1272 struct ceph_inode_info *ci;
1274 if (!in->i_security)
1276 ci = ceph_inode(in);
1277 spin_lock(&ci->i_ceph_lock);
1278 ret = !(ci->i_ceph_flags & CEPH_I_SEC_INITED) &&
1279 !(ci->i_xattrs.version > 0 &&
1280 __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 0));
1281 spin_unlock(&ci->i_ceph_lock);
1285 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1286 int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1287 struct ceph_acl_sec_ctx *as_ctx)
1289 struct ceph_pagelist *pagelist = as_ctx->pagelist;
1294 err = security_dentry_init_security(dentry, mode, &dentry->d_name,
1296 &as_ctx->sec_ctxlen);
1298 WARN_ON_ONCE(err != -EOPNOTSUPP);
1299 err = 0; /* do nothing */
1305 pagelist = ceph_pagelist_alloc(GFP_KERNEL);
1308 err = ceph_pagelist_reserve(pagelist, PAGE_SIZE);
1311 ceph_pagelist_encode_32(pagelist, 1);
1315 * FIXME: Make security_dentry_init_security() generic. Currently
1316 * It only supports single security module and only selinux has
1317 * dentry_init_security hook.
1319 name = XATTR_NAME_SELINUX;
1320 name_len = strlen(name);
1321 err = ceph_pagelist_reserve(pagelist,
1322 4 * 2 + name_len + as_ctx->sec_ctxlen);
1326 if (as_ctx->pagelist) {
1327 /* update count of KV pairs */
1328 BUG_ON(pagelist->length <= sizeof(__le32));
1329 if (list_is_singular(&pagelist->head)) {
1330 le32_add_cpu((__le32*)pagelist->mapped_tail, 1);
1332 struct page *page = list_first_entry(&pagelist->head,
1334 void *addr = kmap_atomic(page);
1335 le32_add_cpu((__le32*)addr, 1);
1336 kunmap_atomic(addr);
1339 as_ctx->pagelist = pagelist;
1342 ceph_pagelist_encode_32(pagelist, name_len);
1343 ceph_pagelist_append(pagelist, name, name_len);
1345 ceph_pagelist_encode_32(pagelist, as_ctx->sec_ctxlen);
1346 ceph_pagelist_append(pagelist, as_ctx->sec_ctx, as_ctx->sec_ctxlen);
1350 if (pagelist && !as_ctx->pagelist)
1351 ceph_pagelist_release(pagelist);
1354 #endif /* CONFIG_CEPH_FS_SECURITY_LABEL */
1355 #endif /* CONFIG_SECURITY */
1357 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx)
1359 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1360 posix_acl_release(as_ctx->acl);
1361 posix_acl_release(as_ctx->default_acl);
1363 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1364 security_release_secctx(as_ctx->sec_ctx, as_ctx->sec_ctxlen);
1366 if (as_ctx->pagelist)
1367 ceph_pagelist_release(as_ctx->pagelist);
1371 * List of handlers for synthetic system.* attributes. Other
1372 * attributes are handled directly.
1374 const struct xattr_handler *ceph_xattr_handlers[] = {
1375 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1376 &posix_acl_access_xattr_handler,
1377 &posix_acl_default_xattr_handler,
1379 &ceph_other_xattr_handler,