diff options
Diffstat (limited to 'fs/ceph/xattr.c')
| -rw-r--r-- | fs/ceph/xattr.c | 844 | 
1 files changed, 844 insertions, 0 deletions
| diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c new file mode 100644 index 000000000000..37d6ce645691 --- /dev/null +++ b/fs/ceph/xattr.c | |||
| @@ -0,0 +1,844 @@ | |||
| 1 | #include "ceph_debug.h" | ||
| 2 | #include "super.h" | ||
| 3 | #include "decode.h" | ||
| 4 | |||
| 5 | #include <linux/xattr.h> | ||
| 6 | |||
| 7 | static bool ceph_is_valid_xattr(const char *name) | ||
| 8 | { | ||
| 9 | return !strncmp(name, XATTR_SECURITY_PREFIX, | ||
| 10 | XATTR_SECURITY_PREFIX_LEN) || | ||
| 11 | !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) || | ||
| 12 | !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN); | ||
| 13 | } | ||
| 14 | |||
| 15 | /* | ||
| 16 | * These define virtual xattrs exposing the recursive directory | ||
| 17 | * statistics and layout metadata. | ||
| 18 | */ | ||
| 19 | struct ceph_vxattr_cb { | ||
| 20 | bool readonly; | ||
| 21 | char *name; | ||
| 22 | size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val, | ||
| 23 | size_t size); | ||
| 24 | }; | ||
| 25 | |||
| 26 | /* directories */ | ||
| 27 | |||
| 28 | static size_t ceph_vxattrcb_entries(struct ceph_inode_info *ci, char *val, | ||
| 29 | size_t size) | ||
| 30 | { | ||
| 31 | return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs); | ||
| 32 | } | ||
| 33 | |||
| 34 | static size_t ceph_vxattrcb_files(struct ceph_inode_info *ci, char *val, | ||
| 35 | size_t size) | ||
| 36 | { | ||
| 37 | return snprintf(val, size, "%lld", ci->i_files); | ||
| 38 | } | ||
| 39 | |||
| 40 | static size_t ceph_vxattrcb_subdirs(struct ceph_inode_info *ci, char *val, | ||
| 41 | size_t size) | ||
| 42 | { | ||
| 43 | return snprintf(val, size, "%lld", ci->i_subdirs); | ||
| 44 | } | ||
| 45 | |||
| 46 | static size_t ceph_vxattrcb_rentries(struct ceph_inode_info *ci, char *val, | ||
| 47 | size_t size) | ||
| 48 | { | ||
| 49 | return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs); | ||
| 50 | } | ||
| 51 | |||
| 52 | static size_t ceph_vxattrcb_rfiles(struct ceph_inode_info *ci, char *val, | ||
| 53 | size_t size) | ||
| 54 | { | ||
| 55 | return snprintf(val, size, "%lld", ci->i_rfiles); | ||
| 56 | } | ||
| 57 | |||
| 58 | static size_t ceph_vxattrcb_rsubdirs(struct ceph_inode_info *ci, char *val, | ||
| 59 | size_t size) | ||
| 60 | { | ||
| 61 | return snprintf(val, size, "%lld", ci->i_rsubdirs); | ||
| 62 | } | ||
| 63 | |||
| 64 | static size_t ceph_vxattrcb_rbytes(struct ceph_inode_info *ci, char *val, | ||
| 65 | size_t size) | ||
| 66 | { | ||
| 67 | return snprintf(val, size, "%lld", ci->i_rbytes); | ||
| 68 | } | ||
| 69 | |||
| 70 | static size_t ceph_vxattrcb_rctime(struct ceph_inode_info *ci, char *val, | ||
| 71 | size_t size) | ||
| 72 | { | ||
| 73 | return snprintf(val, size, "%ld.%ld", (long)ci->i_rctime.tv_sec, | ||
| 74 | (long)ci->i_rctime.tv_nsec); | ||
| 75 | } | ||
| 76 | |||
| 77 | static struct ceph_vxattr_cb ceph_dir_vxattrs[] = { | ||
| 78 | { true, "user.ceph.dir.entries", ceph_vxattrcb_entries}, | ||
| 79 | { true, "user.ceph.dir.files", ceph_vxattrcb_files}, | ||
| 80 | { true, "user.ceph.dir.subdirs", ceph_vxattrcb_subdirs}, | ||
| 81 | { true, "user.ceph.dir.rentries", ceph_vxattrcb_rentries}, | ||
| 82 | { true, "user.ceph.dir.rfiles", ceph_vxattrcb_rfiles}, | ||
| 83 | { true, "user.ceph.dir.rsubdirs", ceph_vxattrcb_rsubdirs}, | ||
| 84 | { true, "user.ceph.dir.rbytes", ceph_vxattrcb_rbytes}, | ||
| 85 | { true, "user.ceph.dir.rctime", ceph_vxattrcb_rctime}, | ||
| 86 | { true, NULL, NULL } | ||
| 87 | }; | ||
| 88 | |||
| 89 | /* files */ | ||
| 90 | |||
| 91 | static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val, | ||
| 92 | size_t size) | ||
| 93 | { | ||
| 94 | int ret; | ||
| 95 | |||
| 96 | ret = snprintf(val, size, | ||
| 97 | "chunk_bytes=%lld\nstripe_count=%lld\nobject_size=%lld\n", | ||
| 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 | if (ceph_file_layout_pg_preferred(ci->i_layout)) | ||
| 102 | ret += snprintf(val + ret, size, "preferred_osd=%lld\n", | ||
| 103 | (unsigned long long)ceph_file_layout_pg_preferred( | ||
| 104 | ci->i_layout)); | ||
| 105 | return ret; | ||
| 106 | } | ||
| 107 | |||
| 108 | static struct ceph_vxattr_cb ceph_file_vxattrs[] = { | ||
| 109 | { true, "user.ceph.layout", ceph_vxattrcb_layout}, | ||
| 110 | { NULL, NULL } | ||
| 111 | }; | ||
| 112 | |||
| 113 | static struct ceph_vxattr_cb *ceph_inode_vxattrs(struct inode *inode) | ||
| 114 | { | ||
| 115 | if (S_ISDIR(inode->i_mode)) | ||
| 116 | return ceph_dir_vxattrs; | ||
| 117 | else if (S_ISREG(inode->i_mode)) | ||
| 118 | return ceph_file_vxattrs; | ||
| 119 | return NULL; | ||
| 120 | } | ||
| 121 | |||
| 122 | static struct ceph_vxattr_cb *ceph_match_vxattr(struct ceph_vxattr_cb *vxattr, | ||
| 123 | const char *name) | ||
| 124 | { | ||
| 125 | do { | ||
| 126 | if (strcmp(vxattr->name, name) == 0) | ||
| 127 | return vxattr; | ||
| 128 | vxattr++; | ||
| 129 | } while (vxattr->name); | ||
| 130 | return NULL; | ||
| 131 | } | ||
| 132 | |||
| 133 | static int __set_xattr(struct ceph_inode_info *ci, | ||
| 134 | const char *name, int name_len, | ||
| 135 | const char *val, int val_len, | ||
| 136 | int dirty, | ||
| 137 | int should_free_name, int should_free_val, | ||
| 138 | struct ceph_inode_xattr **newxattr) | ||
| 139 | { | ||
| 140 | struct rb_node **p; | ||
| 141 | struct rb_node *parent = NULL; | ||
| 142 | struct ceph_inode_xattr *xattr = NULL; | ||
| 143 | int c; | ||
| 144 | int new = 0; | ||
| 145 | |||
| 146 | p = &ci->i_xattrs.index.rb_node; | ||
| 147 | while (*p) { | ||
| 148 | parent = *p; | ||
| 149 | xattr = rb_entry(parent, struct ceph_inode_xattr, node); | ||
| 150 | c = strncmp(name, xattr->name, min(name_len, xattr->name_len)); | ||
| 151 | if (c < 0) | ||
| 152 | p = &(*p)->rb_left; | ||
| 153 | else if (c > 0) | ||
| 154 | p = &(*p)->rb_right; | ||
| 155 | else { | ||
| 156 | if (name_len == xattr->name_len) | ||
| 157 | break; | ||
| 158 | else if (name_len < xattr->name_len) | ||
| 159 | p = &(*p)->rb_left; | ||
| 160 | else | ||
| 161 | p = &(*p)->rb_right; | ||
| 162 | } | ||
| 163 | xattr = NULL; | ||
| 164 | } | ||
| 165 | |||
| 166 | if (!xattr) { | ||
| 167 | new = 1; | ||
| 168 | xattr = *newxattr; | ||
| 169 | xattr->name = name; | ||
| 170 | xattr->name_len = name_len; | ||
| 171 | xattr->should_free_name = should_free_name; | ||
| 172 | |||
| 173 | ci->i_xattrs.count++; | ||
| 174 | dout("__set_xattr count=%d\n", ci->i_xattrs.count); | ||
| 175 | } else { | ||
| 176 | kfree(*newxattr); | ||
| 177 | *newxattr = NULL; | ||
| 178 | if (xattr->should_free_val) | ||
| 179 | kfree((void *)xattr->val); | ||
| 180 | |||
| 181 | if (should_free_name) { | ||
| 182 | kfree((void *)name); | ||
| 183 | name = xattr->name; | ||
| 184 | } | ||
| 185 | ci->i_xattrs.names_size -= xattr->name_len; | ||
| 186 | ci->i_xattrs.vals_size -= xattr->val_len; | ||
| 187 | } | ||
| 188 | if (!xattr) { | ||
| 189 | pr_err("__set_xattr ENOMEM on %p %llx.%llx xattr %s=%s\n", | ||
| 190 | &ci->vfs_inode, ceph_vinop(&ci->vfs_inode), name, | ||
| 191 | xattr->val); | ||
| 192 | return -ENOMEM; | ||
| 193 | } | ||
| 194 | ci->i_xattrs.names_size += name_len; | ||
| 195 | ci->i_xattrs.vals_size += val_len; | ||
| 196 | if (val) | ||
| 197 | xattr->val = val; | ||
| 198 | else | ||
| 199 | xattr->val = ""; | ||
| 200 | |||
| 201 | xattr->val_len = val_len; | ||
| 202 | xattr->dirty = dirty; | ||
| 203 | xattr->should_free_val = (val && should_free_val); | ||
| 204 | |||
| 205 | if (new) { | ||
| 206 | rb_link_node(&xattr->node, parent, p); | ||
| 207 | rb_insert_color(&xattr->node, &ci->i_xattrs.index); | ||
| 208 | dout("__set_xattr_val p=%p\n", p); | ||
| 209 | } | ||
| 210 | |||
| 211 | dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n", | ||
| 212 | ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val); | ||
| 213 | |||
| 214 | return 0; | ||
| 215 | } | ||
| 216 | |||
| 217 | static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci, | ||
| 218 | const char *name) | ||
| 219 | { | ||
| 220 | struct rb_node **p; | ||
| 221 | struct rb_node *parent = NULL; | ||
| 222 | struct ceph_inode_xattr *xattr = NULL; | ||
| 223 | int c; | ||
| 224 | |||
| 225 | p = &ci->i_xattrs.index.rb_node; | ||
| 226 | while (*p) { | ||
| 227 | parent = *p; | ||
| 228 | xattr = rb_entry(parent, struct ceph_inode_xattr, node); | ||
| 229 | c = strncmp(name, xattr->name, xattr->name_len); | ||
| 230 | if (c < 0) | ||
| 231 | p = &(*p)->rb_left; | ||
| 232 | else if (c > 0) | ||
| 233 | p = &(*p)->rb_right; | ||
| 234 | else { | ||
| 235 | dout("__get_xattr %s: found %.*s\n", name, | ||
| 236 | xattr->val_len, xattr->val); | ||
| 237 | return xattr; | ||
| 238 | } | ||
| 239 | } | ||
| 240 | |||
| 241 | dout("__get_xattr %s: not found\n", name); | ||
| 242 | |||
| 243 | return NULL; | ||
| 244 | } | ||
| 245 | |||
| 246 | static void __free_xattr(struct ceph_inode_xattr *xattr) | ||
| 247 | { | ||
| 248 | BUG_ON(!xattr); | ||
| 249 | |||
| 250 | if (xattr->should_free_name) | ||
| 251 | kfree((void *)xattr->name); | ||
| 252 | if (xattr->should_free_val) | ||
| 253 | kfree((void *)xattr->val); | ||
| 254 | |||
| 255 | kfree(xattr); | ||
| 256 | } | ||
| 257 | |||
| 258 | static int __remove_xattr(struct ceph_inode_info *ci, | ||
| 259 | struct ceph_inode_xattr *xattr) | ||
| 260 | { | ||
| 261 | if (!xattr) | ||
| 262 | return -EOPNOTSUPP; | ||
| 263 | |||
| 264 | rb_erase(&xattr->node, &ci->i_xattrs.index); | ||
| 265 | |||
| 266 | if (xattr->should_free_name) | ||
| 267 | kfree((void *)xattr->name); | ||
| 268 | if (xattr->should_free_val) | ||
| 269 | kfree((void *)xattr->val); | ||
| 270 | |||
| 271 | ci->i_xattrs.names_size -= xattr->name_len; | ||
| 272 | ci->i_xattrs.vals_size -= xattr->val_len; | ||
| 273 | ci->i_xattrs.count--; | ||
| 274 | kfree(xattr); | ||
| 275 | |||
| 276 | return 0; | ||
| 277 | } | ||
| 278 | |||
| 279 | static int __remove_xattr_by_name(struct ceph_inode_info *ci, | ||
| 280 | const char *name) | ||
| 281 | { | ||
| 282 | struct rb_node **p; | ||
| 283 | struct ceph_inode_xattr *xattr; | ||
| 284 | int err; | ||
| 285 | |||
| 286 | p = &ci->i_xattrs.index.rb_node; | ||
| 287 | xattr = __get_xattr(ci, name); | ||
| 288 | err = __remove_xattr(ci, xattr); | ||
| 289 | return err; | ||
| 290 | } | ||
| 291 | |||
| 292 | static char *__copy_xattr_names(struct ceph_inode_info *ci, | ||
| 293 | char *dest) | ||
| 294 | { | ||
| 295 | struct rb_node *p; | ||
| 296 | struct ceph_inode_xattr *xattr = NULL; | ||
| 297 | |||
| 298 | p = rb_first(&ci->i_xattrs.index); | ||
| 299 | dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count); | ||
| 300 | |||
| 301 | while (p) { | ||
| 302 | xattr = rb_entry(p, struct ceph_inode_xattr, node); | ||
| 303 | memcpy(dest, xattr->name, xattr->name_len); | ||
| 304 | dest[xattr->name_len] = '\0'; | ||
| 305 | |||
| 306 | dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name, | ||
| 307 | xattr->name_len, ci->i_xattrs.names_size); | ||
| 308 | |||
| 309 | dest += xattr->name_len + 1; | ||
| 310 | p = rb_next(p); | ||
| 311 | } | ||
| 312 | |||
| 313 | return dest; | ||
| 314 | } | ||
| 315 | |||
| 316 | void __ceph_destroy_xattrs(struct ceph_inode_info *ci) | ||
| 317 | { | ||
| 318 | struct rb_node *p, *tmp; | ||
| 319 | struct ceph_inode_xattr *xattr = NULL; | ||
| 320 | |||
| 321 | p = rb_first(&ci->i_xattrs.index); | ||
| 322 | |||
| 323 | dout("__ceph_destroy_xattrs p=%p\n", p); | ||
| 324 | |||
| 325 | while (p) { | ||
| 326 | xattr = rb_entry(p, struct ceph_inode_xattr, node); | ||
| 327 | tmp = p; | ||
| 328 | p = rb_next(tmp); | ||
| 329 | dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p, | ||
| 330 | xattr->name_len, xattr->name); | ||
| 331 | rb_erase(tmp, &ci->i_xattrs.index); | ||
| 332 | |||
| 333 | __free_xattr(xattr); | ||
| 334 | } | ||
| 335 | |||
| 336 | ci->i_xattrs.names_size = 0; | ||
| 337 | ci->i_xattrs.vals_size = 0; | ||
| 338 | ci->i_xattrs.index_version = 0; | ||
| 339 | ci->i_xattrs.count = 0; | ||
| 340 | ci->i_xattrs.index = RB_ROOT; | ||
| 341 | } | ||
| 342 | |||
| 343 | static int __build_xattrs(struct inode *inode) | ||
| 344 | { | ||
| 345 | u32 namelen; | ||
| 346 | u32 numattr = 0; | ||
| 347 | void *p, *end; | ||
| 348 | u32 len; | ||
| 349 | const char *name, *val; | ||
| 350 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
| 351 | int xattr_version; | ||
| 352 | struct ceph_inode_xattr **xattrs = NULL; | ||
| 353 | int err = 0; | ||
| 354 | int i; | ||
| 355 | |||
| 356 | dout("__build_xattrs() len=%d\n", | ||
| 357 | ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0); | ||
| 358 | |||
| 359 | if (ci->i_xattrs.index_version >= ci->i_xattrs.version) | ||
| 360 | return 0; /* already built */ | ||
| 361 | |||
| 362 | __ceph_destroy_xattrs(ci); | ||
| 363 | |||
| 364 | start: | ||
| 365 | /* updated internal xattr rb tree */ | ||
| 366 | if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) { | ||
| 367 | p = ci->i_xattrs.blob->vec.iov_base; | ||
| 368 | end = p + ci->i_xattrs.blob->vec.iov_len; | ||
| 369 | ceph_decode_32_safe(&p, end, numattr, bad); | ||
| 370 | xattr_version = ci->i_xattrs.version; | ||
| 371 | spin_unlock(&inode->i_lock); | ||
| 372 | |||
| 373 | xattrs = kcalloc(numattr, sizeof(struct ceph_xattr *), | ||
| 374 | GFP_NOFS); | ||
| 375 | err = -ENOMEM; | ||
| 376 | if (!xattrs) | ||
| 377 | goto bad_lock; | ||
| 378 | memset(xattrs, 0, numattr*sizeof(struct ceph_xattr *)); | ||
| 379 | for (i = 0; i < numattr; i++) { | ||
| 380 | xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr), | ||
| 381 | GFP_NOFS); | ||
| 382 | if (!xattrs[i]) | ||
| 383 | goto bad_lock; | ||
| 384 | } | ||
| 385 | |||
| 386 | spin_lock(&inode->i_lock); | ||
| 387 | if (ci->i_xattrs.version != xattr_version) { | ||
| 388 | /* lost a race, retry */ | ||
| 389 | for (i = 0; i < numattr; i++) | ||
| 390 | kfree(xattrs[i]); | ||
| 391 | kfree(xattrs); | ||
| 392 | goto start; | ||
| 393 | } | ||
| 394 | err = -EIO; | ||
| 395 | while (numattr--) { | ||
| 396 | ceph_decode_32_safe(&p, end, len, bad); | ||
| 397 | namelen = len; | ||
| 398 | name = p; | ||
| 399 | p += len; | ||
| 400 | ceph_decode_32_safe(&p, end, len, bad); | ||
| 401 | val = p; | ||
| 402 | p += len; | ||
| 403 | |||
| 404 | err = __set_xattr(ci, name, namelen, val, len, | ||
| 405 | 0, 0, 0, &xattrs[numattr]); | ||
| 406 | |||
| 407 | if (err < 0) | ||
| 408 | goto bad; | ||
| 409 | } | ||
| 410 | kfree(xattrs); | ||
| 411 | } | ||
| 412 | ci->i_xattrs.index_version = ci->i_xattrs.version; | ||
| 413 | ci->i_xattrs.dirty = false; | ||
| 414 | |||
| 415 | return err; | ||
| 416 | bad_lock: | ||
| 417 | spin_lock(&inode->i_lock); | ||
| 418 | bad: | ||
| 419 | if (xattrs) { | ||
| 420 | for (i = 0; i < numattr; i++) | ||
| 421 | kfree(xattrs[i]); | ||
| 422 | kfree(xattrs); | ||
| 423 | } | ||
| 424 | ci->i_xattrs.names_size = 0; | ||
| 425 | return err; | ||
| 426 | } | ||
| 427 | |||
| 428 | static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size, | ||
| 429 | int val_size) | ||
| 430 | { | ||
| 431 | /* | ||
| 432 | * 4 bytes for the length, and additional 4 bytes per each xattr name, | ||
| 433 | * 4 bytes per each value | ||
| 434 | */ | ||
| 435 | int size = 4 + ci->i_xattrs.count*(4 + 4) + | ||
| 436 | ci->i_xattrs.names_size + | ||
| 437 | ci->i_xattrs.vals_size; | ||
| 438 | dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n", | ||
| 439 | ci->i_xattrs.count, ci->i_xattrs.names_size, | ||
| 440 | ci->i_xattrs.vals_size); | ||
| 441 | |||
| 442 | if (name_size) | ||
| 443 | size += 4 + 4 + name_size + val_size; | ||
| 444 | |||
| 445 | return size; | ||
| 446 | } | ||
| 447 | |||
| 448 | /* | ||
| 449 | * If there are dirty xattrs, reencode xattrs into the prealloc_blob | ||
| 450 | * and swap into place. | ||
| 451 | */ | ||
| 452 | void __ceph_build_xattrs_blob(struct ceph_inode_info *ci) | ||
| 453 | { | ||
| 454 | struct rb_node *p; | ||
| 455 | struct ceph_inode_xattr *xattr = NULL; | ||
| 456 | void *dest; | ||
| 457 | |||
| 458 | dout("__build_xattrs_blob %p\n", &ci->vfs_inode); | ||
| 459 | if (ci->i_xattrs.dirty) { | ||
| 460 | int need = __get_required_blob_size(ci, 0, 0); | ||
| 461 | |||
| 462 | BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len); | ||
| 463 | |||
| 464 | p = rb_first(&ci->i_xattrs.index); | ||
| 465 | dest = ci->i_xattrs.prealloc_blob->vec.iov_base; | ||
| 466 | |||
| 467 | ceph_encode_32(&dest, ci->i_xattrs.count); | ||
| 468 | while (p) { | ||
| 469 | xattr = rb_entry(p, struct ceph_inode_xattr, node); | ||
| 470 | |||
| 471 | ceph_encode_32(&dest, xattr->name_len); | ||
| 472 | memcpy(dest, xattr->name, xattr->name_len); | ||
| 473 | dest += xattr->name_len; | ||
| 474 | ceph_encode_32(&dest, xattr->val_len); | ||
| 475 | memcpy(dest, xattr->val, xattr->val_len); | ||
| 476 | dest += xattr->val_len; | ||
| 477 | |||
| 478 | p = rb_next(p); | ||
| 479 | } | ||
| 480 | |||
| 481 | /* adjust buffer len; it may be larger than we need */ | ||
| 482 | ci->i_xattrs.prealloc_blob->vec.iov_len = | ||
| 483 | dest - ci->i_xattrs.prealloc_blob->vec.iov_base; | ||
| 484 | |||
| 485 | if (ci->i_xattrs.blob) | ||
| 486 | ceph_buffer_put(ci->i_xattrs.blob); | ||
| 487 | ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob; | ||
| 488 | ci->i_xattrs.prealloc_blob = NULL; | ||
| 489 | ci->i_xattrs.dirty = false; | ||
| 490 | } | ||
| 491 | } | ||
| 492 | |||
| 493 | ssize_t ceph_getxattr(struct dentry *dentry, const char *name, void *value, | ||
| 494 | size_t size) | ||
| 495 | { | ||
| 496 | struct inode *inode = dentry->d_inode; | ||
| 497 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
| 498 | struct ceph_vxattr_cb *vxattrs = ceph_inode_vxattrs(inode); | ||
| 499 | int err; | ||
| 500 | struct ceph_inode_xattr *xattr; | ||
| 501 | struct ceph_vxattr_cb *vxattr = NULL; | ||
| 502 | |||
| 503 | if (!ceph_is_valid_xattr(name)) | ||
| 504 | return -ENODATA; | ||
| 505 | |||
| 506 | /* let's see if a virtual xattr was requested */ | ||
| 507 | if (vxattrs) | ||
| 508 | vxattr = ceph_match_vxattr(vxattrs, name); | ||
| 509 | |||
| 510 | spin_lock(&inode->i_lock); | ||
| 511 | dout("getxattr %p ver=%lld index_ver=%lld\n", inode, | ||
| 512 | ci->i_xattrs.version, ci->i_xattrs.index_version); | ||
| 513 | |||
| 514 | if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) && | ||
| 515 | (ci->i_xattrs.index_version >= ci->i_xattrs.version)) { | ||
| 516 | goto get_xattr; | ||
| 517 | } else { | ||
| 518 | spin_unlock(&inode->i_lock); | ||
| 519 | /* get xattrs from mds (if we don't already have them) */ | ||
| 520 | err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR); | ||
| 521 | if (err) | ||
| 522 | return err; | ||
| 523 | } | ||
| 524 | |||
| 525 | spin_lock(&inode->i_lock); | ||
| 526 | |||
| 527 | if (vxattr && vxattr->readonly) { | ||
| 528 | err = vxattr->getxattr_cb(ci, value, size); | ||
| 529 | goto out; | ||
| 530 | } | ||
| 531 | |||
| 532 | err = __build_xattrs(inode); | ||
| 533 | if (err < 0) | ||
| 534 | goto out; | ||
| 535 | |||
| 536 | get_xattr: | ||
| 537 | err = -ENODATA; /* == ENOATTR */ | ||
| 538 | xattr = __get_xattr(ci, name); | ||
| 539 | if (!xattr) { | ||
| 540 | if (vxattr) | ||
| 541 | err = vxattr->getxattr_cb(ci, value, size); | ||
| 542 | goto out; | ||
| 543 | } | ||
| 544 | |||
| 545 | err = -ERANGE; | ||
| 546 | if (size && size < xattr->val_len) | ||
| 547 | goto out; | ||
| 548 | |||
| 549 | err = xattr->val_len; | ||
| 550 | if (size == 0) | ||
| 551 | goto out; | ||
| 552 | |||
| 553 | memcpy(value, xattr->val, xattr->val_len); | ||
| 554 | |||
| 555 | out: | ||
| 556 | spin_unlock(&inode->i_lock); | ||
| 557 | return err; | ||
| 558 | } | ||
| 559 | |||
| 560 | ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size) | ||
| 561 | { | ||
| 562 | struct inode *inode = dentry->d_inode; | ||
| 563 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
| 564 | struct ceph_vxattr_cb *vxattrs = ceph_inode_vxattrs(inode); | ||
| 565 | u32 vir_namelen = 0; | ||
| 566 | u32 namelen; | ||
| 567 | int err; | ||
| 568 | u32 len; | ||
| 569 | int i; | ||
| 570 | |||
| 571 | spin_lock(&inode->i_lock); | ||
| 572 | dout("listxattr %p ver=%lld index_ver=%lld\n", inode, | ||
| 573 | ci->i_xattrs.version, ci->i_xattrs.index_version); | ||
| 574 | |||
| 575 | if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) && | ||
| 576 | (ci->i_xattrs.index_version > ci->i_xattrs.version)) { | ||
| 577 | goto list_xattr; | ||
| 578 | } else { | ||
| 579 | spin_unlock(&inode->i_lock); | ||
| 580 | err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR); | ||
| 581 | if (err) | ||
| 582 | return err; | ||
| 583 | } | ||
| 584 | |||
| 585 | spin_lock(&inode->i_lock); | ||
| 586 | |||
| 587 | err = __build_xattrs(inode); | ||
| 588 | if (err < 0) | ||
| 589 | goto out; | ||
| 590 | |||
| 591 | list_xattr: | ||
| 592 | vir_namelen = 0; | ||
| 593 | /* include virtual dir xattrs */ | ||
| 594 | if (vxattrs) | ||
| 595 | for (i = 0; vxattrs[i].name; i++) | ||
| 596 | vir_namelen += strlen(vxattrs[i].name) + 1; | ||
| 597 | /* adding 1 byte per each variable due to the null termination */ | ||
| 598 | namelen = vir_namelen + ci->i_xattrs.names_size + ci->i_xattrs.count; | ||
| 599 | err = -ERANGE; | ||
| 600 | if (size && namelen > size) | ||
| 601 | goto out; | ||
| 602 | |||
| 603 | err = namelen; | ||
| 604 | if (size == 0) | ||
| 605 | goto out; | ||
| 606 | |||
| 607 | names = __copy_xattr_names(ci, names); | ||
| 608 | |||
| 609 | /* virtual xattr names, too */ | ||
| 610 | if (vxattrs) | ||
| 611 | for (i = 0; vxattrs[i].name; i++) { | ||
| 612 | len = sprintf(names, "%s", vxattrs[i].name); | ||
| 613 | names += len + 1; | ||
| 614 | } | ||
| 615 | |||
| 616 | out: | ||
| 617 | spin_unlock(&inode->i_lock); | ||
| 618 | return err; | ||
| 619 | } | ||
| 620 | |||
| 621 | static int ceph_sync_setxattr(struct dentry *dentry, const char *name, | ||
| 622 | const char *value, size_t size, int flags) | ||
| 623 | { | ||
| 624 | struct ceph_client *client = ceph_client(dentry->d_sb); | ||
| 625 | struct inode *inode = dentry->d_inode; | ||
| 626 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
| 627 | struct inode *parent_inode = dentry->d_parent->d_inode; | ||
| 628 | struct ceph_mds_request *req; | ||
| 629 | struct ceph_mds_client *mdsc = &client->mdsc; | ||
| 630 | int err; | ||
| 631 | int i, nr_pages; | ||
| 632 | struct page **pages = NULL; | ||
| 633 | void *kaddr; | ||
| 634 | |||
| 635 | /* copy value into some pages */ | ||
| 636 | nr_pages = calc_pages_for(0, size); | ||
| 637 | if (nr_pages) { | ||
| 638 | pages = kmalloc(sizeof(pages[0])*nr_pages, GFP_NOFS); | ||
| 639 | if (!pages) | ||
| 640 | return -ENOMEM; | ||
| 641 | err = -ENOMEM; | ||
| 642 | for (i = 0; i < nr_pages; i++) { | ||
| 643 | pages[i] = alloc_page(GFP_NOFS); | ||
| 644 | if (!pages[i]) { | ||
| 645 | nr_pages = i; | ||
| 646 | goto out; | ||
| 647 | } | ||
| 648 | kaddr = kmap(pages[i]); | ||
| 649 | memcpy(kaddr, value + i*PAGE_CACHE_SIZE, | ||
| 650 | min(PAGE_CACHE_SIZE, size-i*PAGE_CACHE_SIZE)); | ||
| 651 | } | ||
| 652 | } | ||
| 653 | |||
| 654 | dout("setxattr value=%.*s\n", (int)size, value); | ||
| 655 | |||
| 656 | /* do request */ | ||
| 657 | req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETXATTR, | ||
| 658 | USE_AUTH_MDS); | ||
| 659 | if (IS_ERR(req)) { | ||
| 660 | err = PTR_ERR(req); | ||
| 661 | goto out; | ||
| 662 | } | ||
| 663 | req->r_inode = igrab(inode); | ||
| 664 | req->r_inode_drop = CEPH_CAP_XATTR_SHARED; | ||
| 665 | req->r_num_caps = 1; | ||
| 666 | req->r_args.setxattr.flags = cpu_to_le32(flags); | ||
| 667 | req->r_path2 = kstrdup(name, GFP_NOFS); | ||
| 668 | |||
| 669 | req->r_pages = pages; | ||
| 670 | req->r_num_pages = nr_pages; | ||
| 671 | req->r_data_len = size; | ||
| 672 | |||
| 673 | dout("xattr.ver (before): %lld\n", ci->i_xattrs.version); | ||
| 674 | err = ceph_mdsc_do_request(mdsc, parent_inode, req); | ||
| 675 | ceph_mdsc_put_request(req); | ||
| 676 | dout("xattr.ver (after): %lld\n", ci->i_xattrs.version); | ||
| 677 | |||
| 678 | out: | ||
| 679 | if (pages) { | ||
| 680 | for (i = 0; i < nr_pages; i++) | ||
| 681 | __free_page(pages[i]); | ||
| 682 | kfree(pages); | ||
| 683 | } | ||
| 684 | return err; | ||
| 685 | } | ||
| 686 | |||
| 687 | int ceph_setxattr(struct dentry *dentry, const char *name, | ||
| 688 | const void *value, size_t size, int flags) | ||
| 689 | { | ||
| 690 | struct inode *inode = dentry->d_inode; | ||
| 691 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
| 692 | struct ceph_vxattr_cb *vxattrs = ceph_inode_vxattrs(inode); | ||
| 693 | int err; | ||
| 694 | int name_len = strlen(name); | ||
| 695 | int val_len = size; | ||
| 696 | char *newname = NULL; | ||
| 697 | char *newval = NULL; | ||
| 698 | struct ceph_inode_xattr *xattr = NULL; | ||
| 699 | int issued; | ||
| 700 | int required_blob_size; | ||
| 701 | |||
| 702 | if (ceph_snap(inode) != CEPH_NOSNAP) | ||
| 703 | return -EROFS; | ||
| 704 | |||
| 705 | if (!ceph_is_valid_xattr(name)) | ||
| 706 | return -EOPNOTSUPP; | ||
| 707 | |||
| 708 | if (vxattrs) { | ||
| 709 | struct ceph_vxattr_cb *vxattr = | ||
| 710 | ceph_match_vxattr(vxattrs, name); | ||
| 711 | if (vxattr && vxattr->readonly) | ||
| 712 | return -EOPNOTSUPP; | ||
| 713 | } | ||
| 714 | |||
| 715 | /* preallocate memory for xattr name, value, index node */ | ||
| 716 | err = -ENOMEM; | ||
| 717 | newname = kmalloc(name_len + 1, GFP_NOFS); | ||
| 718 | if (!newname) | ||
| 719 | goto out; | ||
| 720 | memcpy(newname, name, name_len + 1); | ||
| 721 | |||
| 722 | if (val_len) { | ||
| 723 | newval = kmalloc(val_len + 1, GFP_NOFS); | ||
| 724 | if (!newval) | ||
| 725 | goto out; | ||
| 726 | memcpy(newval, value, val_len); | ||
| 727 | newval[val_len] = '\0'; | ||
| 728 | } | ||
| 729 | |||
| 730 | xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS); | ||
| 731 | if (!xattr) | ||
| 732 | goto out; | ||
| 733 | |||
| 734 | spin_lock(&inode->i_lock); | ||
| 735 | retry: | ||
| 736 | issued = __ceph_caps_issued(ci, NULL); | ||
| 737 | if (!(issued & CEPH_CAP_XATTR_EXCL)) | ||
| 738 | goto do_sync; | ||
| 739 | __build_xattrs(inode); | ||
| 740 | |||
| 741 | required_blob_size = __get_required_blob_size(ci, name_len, val_len); | ||
| 742 | |||
| 743 | if (!ci->i_xattrs.prealloc_blob || | ||
| 744 | required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) { | ||
| 745 | struct ceph_buffer *blob = NULL; | ||
| 746 | |||
| 747 | spin_unlock(&inode->i_lock); | ||
| 748 | dout(" preaallocating new blob size=%d\n", required_blob_size); | ||
| 749 | blob = ceph_buffer_new(required_blob_size, GFP_NOFS); | ||
| 750 | if (!blob) | ||
| 751 | goto out; | ||
| 752 | spin_lock(&inode->i_lock); | ||
| 753 | if (ci->i_xattrs.prealloc_blob) | ||
| 754 | ceph_buffer_put(ci->i_xattrs.prealloc_blob); | ||
| 755 | ci->i_xattrs.prealloc_blob = blob; | ||
| 756 | goto retry; | ||
| 757 | } | ||
| 758 | |||
| 759 | dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued)); | ||
| 760 | err = __set_xattr(ci, newname, name_len, newval, | ||
| 761 | val_len, 1, 1, 1, &xattr); | ||
| 762 | __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL); | ||
| 763 | ci->i_xattrs.dirty = true; | ||
| 764 | inode->i_ctime = CURRENT_TIME; | ||
| 765 | spin_unlock(&inode->i_lock); | ||
| 766 | |||
| 767 | return err; | ||
| 768 | |||
| 769 | do_sync: | ||
| 770 | spin_unlock(&inode->i_lock); | ||
| 771 | err = ceph_sync_setxattr(dentry, name, value, size, flags); | ||
| 772 | out: | ||
| 773 | kfree(newname); | ||
| 774 | kfree(newval); | ||
| 775 | kfree(xattr); | ||
| 776 | return err; | ||
| 777 | } | ||
| 778 | |||
| 779 | static int ceph_send_removexattr(struct dentry *dentry, const char *name) | ||
| 780 | { | ||
| 781 | struct ceph_client *client = ceph_client(dentry->d_sb); | ||
| 782 | struct ceph_mds_client *mdsc = &client->mdsc; | ||
| 783 | struct inode *inode = dentry->d_inode; | ||
| 784 | struct inode *parent_inode = dentry->d_parent->d_inode; | ||
| 785 | struct ceph_mds_request *req; | ||
| 786 | int err; | ||
| 787 | |||
| 788 | req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_RMXATTR, | ||
| 789 | USE_AUTH_MDS); | ||
| 790 | if (IS_ERR(req)) | ||
| 791 | return PTR_ERR(req); | ||
| 792 | req->r_inode = igrab(inode); | ||
| 793 | req->r_inode_drop = CEPH_CAP_XATTR_SHARED; | ||
| 794 | req->r_num_caps = 1; | ||
| 795 | req->r_path2 = kstrdup(name, GFP_NOFS); | ||
| 796 | |||
| 797 | err = ceph_mdsc_do_request(mdsc, parent_inode, req); | ||
| 798 | ceph_mdsc_put_request(req); | ||
| 799 | return err; | ||
| 800 | } | ||
| 801 | |||
| 802 | int ceph_removexattr(struct dentry *dentry, const char *name) | ||
| 803 | { | ||
| 804 | struct inode *inode = dentry->d_inode; | ||
| 805 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
| 806 | struct ceph_vxattr_cb *vxattrs = ceph_inode_vxattrs(inode); | ||
| 807 | int issued; | ||
| 808 | int err; | ||
| 809 | |||
| 810 | if (ceph_snap(inode) != CEPH_NOSNAP) | ||
| 811 | return -EROFS; | ||
| 812 | |||
| 813 | if (!ceph_is_valid_xattr(name)) | ||
| 814 | return -EOPNOTSUPP; | ||
| 815 | |||
| 816 | if (vxattrs) { | ||
| 817 | struct ceph_vxattr_cb *vxattr = | ||
| 818 | ceph_match_vxattr(vxattrs, name); | ||
| 819 | if (vxattr && vxattr->readonly) | ||
| 820 | return -EOPNOTSUPP; | ||
| 821 | } | ||
| 822 | |||
| 823 | spin_lock(&inode->i_lock); | ||
| 824 | __build_xattrs(inode); | ||
| 825 | issued = __ceph_caps_issued(ci, NULL); | ||
| 826 | dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued)); | ||
| 827 | |||
| 828 | if (!(issued & CEPH_CAP_XATTR_EXCL)) | ||
| 829 | goto do_sync; | ||
| 830 | |||
| 831 | err = __remove_xattr_by_name(ceph_inode(inode), name); | ||
| 832 | __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL); | ||
| 833 | ci->i_xattrs.dirty = true; | ||
| 834 | inode->i_ctime = CURRENT_TIME; | ||
| 835 | |||
| 836 | spin_unlock(&inode->i_lock); | ||
| 837 | |||
| 838 | return err; | ||
| 839 | do_sync: | ||
| 840 | spin_unlock(&inode->i_lock); | ||
| 841 | err = ceph_send_removexattr(dentry, name); | ||
| 842 | return err; | ||
| 843 | } | ||
| 844 | |||
