1// SPDX-License-Identifier: GPL-2.0 2#include <linux/ceph/ceph_debug.h> 3 4#include <linux/exportfs.h> 5#include <linux/slab.h> 6#include <asm/unaligned.h> 7 8#include "super.h" 9#include "mds_client.h" 10#include "crypto.h" 11 12/* 13 * Basic fh 14 */ 15struct ceph_nfs_fh { 16 u64 ino; 17} __attribute__ ((packed)); 18 19/* 20 * Larger fh that includes parent ino. 21 */ 22struct ceph_nfs_confh { 23 u64 ino, parent_ino; 24} __attribute__ ((packed)); 25 26/* 27 * fh for snapped inode 28 */ 29struct ceph_nfs_snapfh { 30 u64 ino; 31 u64 snapid; 32 u64 parent_ino; 33 u32 hash; 34} __attribute__ ((packed)); 35 36static int ceph_encode_snapfh(struct inode *inode, u32 *rawfh, int *max_len, 37 struct inode *parent_inode) 38{ 39 static const int snap_handle_length = 40 sizeof(struct ceph_nfs_snapfh) >> 2; 41 struct ceph_nfs_snapfh *sfh = (void *)rawfh; 42 u64 snapid = ceph_snap(inode); 43 int ret; 44 bool no_parent = true; 45 46 if (*max_len < snap_handle_length) { 47 *max_len = snap_handle_length; 48 ret = FILEID_INVALID; 49 goto out; 50 } 51 52 ret = -EINVAL; 53 if (snapid != CEPH_SNAPDIR) { 54 struct inode *dir; 55 struct dentry *dentry = d_find_alias(inode); 56 if (!dentry) 57 goto out; 58 59 rcu_read_lock(); 60 dir = d_inode_rcu(dentry->d_parent); 61 if (ceph_snap(dir) != CEPH_SNAPDIR) { 62 sfh->parent_ino = ceph_ino(dir); 63 sfh->hash = ceph_dentry_hash(dir, dentry); 64 no_parent = false; 65 } 66 rcu_read_unlock(); 67 dput(dentry); 68 } 69 70 if (no_parent) { 71 if (!S_ISDIR(inode->i_mode)) 72 goto out; 73 sfh->parent_ino = sfh->ino; 74 sfh->hash = 0; 75 } 76 sfh->ino = ceph_ino(inode); 77 sfh->snapid = snapid; 78 79 *max_len = snap_handle_length; 80 ret = FILEID_BTRFS_WITH_PARENT; 81out: 82 dout("encode_snapfh %llx.%llx ret=%d\n", ceph_vinop(inode), ret); 83 return ret; 84} 85 86static int ceph_encode_fh(struct inode *inode, u32 *rawfh, int *max_len, 87 struct inode *parent_inode) 88{ 89 static const int handle_length = 90 sizeof(struct ceph_nfs_fh) >> 2; 91 static const int connected_handle_length = 92 sizeof(struct ceph_nfs_confh) >> 2; 93 int type; 94 95 if (ceph_snap(inode) != CEPH_NOSNAP) 96 return ceph_encode_snapfh(inode, rawfh, max_len, parent_inode); 97 98 if (parent_inode && (*max_len < connected_handle_length)) { 99 *max_len = connected_handle_length; 100 return FILEID_INVALID; 101 } else if (*max_len < handle_length) { 102 *max_len = handle_length; 103 return FILEID_INVALID; 104 } 105 106 if (parent_inode) { 107 struct ceph_nfs_confh *cfh = (void *)rawfh; 108 dout("encode_fh %llx with parent %llx\n", 109 ceph_ino(inode), ceph_ino(parent_inode)); 110 cfh->ino = ceph_ino(inode); 111 cfh->parent_ino = ceph_ino(parent_inode); 112 *max_len = connected_handle_length; 113 type = FILEID_INO32_GEN_PARENT; 114 } else { 115 struct ceph_nfs_fh *fh = (void *)rawfh; 116 dout("encode_fh %llx\n", ceph_ino(inode)); 117 fh->ino = ceph_ino(inode); 118 *max_len = handle_length; 119 type = FILEID_INO32_GEN; 120 } 121 return type; 122} 123 124static struct inode *__lookup_inode(struct super_block *sb, u64 ino) 125{ 126 struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc; 127 struct inode *inode; 128 struct ceph_vino vino; 129 int err; 130 131 vino.ino = ino; 132 vino.snap = CEPH_NOSNAP; 133 134 if (ceph_vino_is_reserved(vino)) 135 return ERR_PTR(-ESTALE); 136 137 inode = ceph_find_inode(sb, vino); 138 if (!inode) { 139 struct ceph_mds_request *req; 140 int mask; 141 142 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPINO, 143 USE_ANY_MDS); 144 if (IS_ERR(req)) 145 return ERR_CAST(req); 146 147 mask = CEPH_STAT_CAP_INODE; 148 if (ceph_security_xattr_wanted(d_inode(sb->s_root))) 149 mask |= CEPH_CAP_XATTR_SHARED; 150 req->r_args.lookupino.mask = cpu_to_le32(mask); 151 152 req->r_ino1 = vino; 153 req->r_num_caps = 1; 154 err = ceph_mdsc_do_request(mdsc, NULL, req); 155 inode = req->r_target_inode; 156 if (inode) 157 ihold(inode); 158 ceph_mdsc_put_request(req); 159 if (!inode) 160 return err < 0 ? ERR_PTR(err) : ERR_PTR(-ESTALE); 161 } else { 162 if (ceph_inode_is_shutdown(inode)) { 163 iput(inode); 164 return ERR_PTR(-ESTALE); 165 } 166 } 167 return inode; 168} 169 170struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino) 171{ 172 struct inode *inode = __lookup_inode(sb, ino); 173 if (IS_ERR(inode)) 174 return inode; 175 if (inode->i_nlink == 0) { 176 iput(inode); 177 return ERR_PTR(-ESTALE); 178 } 179 return inode; 180} 181 182static struct dentry *__fh_to_dentry(struct super_block *sb, u64 ino) 183{ 184 struct inode *inode = __lookup_inode(sb, ino); 185 struct ceph_inode_info *ci = ceph_inode(inode); 186 int err; 187 188 if (IS_ERR(inode)) 189 return ERR_CAST(inode); 190 /* We need LINK caps to reliably check i_nlink */ 191 err = ceph_do_getattr(inode, CEPH_CAP_LINK_SHARED, false); 192 if (err) { 193 iput(inode); 194 return ERR_PTR(err); 195 } 196 /* -ESTALE if inode as been unlinked and no file is open */ 197 if ((inode->i_nlink == 0) && !__ceph_is_file_opened(ci)) { 198 iput(inode); 199 return ERR_PTR(-ESTALE); 200 } 201 return d_obtain_alias(inode); 202} 203 204static struct dentry *__snapfh_to_dentry(struct super_block *sb, 205 struct ceph_nfs_snapfh *sfh, 206 bool want_parent) 207{ 208 struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc; 209 struct ceph_mds_request *req; 210 struct inode *inode; 211 struct ceph_vino vino; 212 int mask; 213 int err; 214 bool unlinked = false; 215 216 if (want_parent) { 217 vino.ino = sfh->parent_ino; 218 if (sfh->snapid == CEPH_SNAPDIR) 219 vino.snap = CEPH_NOSNAP; 220 else if (sfh->ino == sfh->parent_ino) 221 vino.snap = CEPH_SNAPDIR; 222 else 223 vino.snap = sfh->snapid; 224 } else { 225 vino.ino = sfh->ino; 226 vino.snap = sfh->snapid; 227 } 228 229 if (ceph_vino_is_reserved(vino)) 230 return ERR_PTR(-ESTALE); 231 232 inode = ceph_find_inode(sb, vino); 233 if (inode) { 234 if (ceph_inode_is_shutdown(inode)) { 235 iput(inode); 236 return ERR_PTR(-ESTALE); 237 } 238 return d_obtain_alias(inode); 239 } 240 241 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPINO, 242 USE_ANY_MDS); 243 if (IS_ERR(req)) 244 return ERR_CAST(req); 245 246 mask = CEPH_STAT_CAP_INODE; 247 if (ceph_security_xattr_wanted(d_inode(sb->s_root))) 248 mask |= CEPH_CAP_XATTR_SHARED; 249 req->r_args.lookupino.mask = cpu_to_le32(mask); 250 if (vino.snap < CEPH_NOSNAP) { 251 req->r_args.lookupino.snapid = cpu_to_le64(vino.snap); 252 if (!want_parent && sfh->ino != sfh->parent_ino) { 253 req->r_args.lookupino.parent = 254 cpu_to_le64(sfh->parent_ino); 255 req->r_args.lookupino.hash = 256 cpu_to_le32(sfh->hash); 257 } 258 } 259 260 req->r_ino1 = vino; 261 req->r_num_caps = 1; 262 err = ceph_mdsc_do_request(mdsc, NULL, req); 263 inode = req->r_target_inode; 264 if (inode) { 265 if (vino.snap == CEPH_SNAPDIR) { 266 if (inode->i_nlink == 0) 267 unlinked = true; 268 inode = ceph_get_snapdir(inode); 269 } else if (ceph_snap(inode) == vino.snap) { 270 ihold(inode); 271 } else { 272 /* mds does not support lookup snapped inode */ 273 inode = ERR_PTR(-EOPNOTSUPP); 274 } 275 } else { 276 inode = ERR_PTR(-ESTALE); 277 } 278 ceph_mdsc_put_request(req); 279 280 if (want_parent) { 281 dout("snapfh_to_parent %llx.%llx\n err=%d\n", 282 vino.ino, vino.snap, err); 283 } else { 284 dout("snapfh_to_dentry %llx.%llx parent %llx hash %x err=%d", 285 vino.ino, vino.snap, sfh->parent_ino, sfh->hash, err); 286 } 287 if (IS_ERR(inode)) 288 return ERR_CAST(inode); 289 /* see comments in ceph_get_parent() */ 290 return unlinked ? d_obtain_root(inode) : d_obtain_alias(inode); 291} 292 293/* 294 * convert regular fh to dentry 295 */ 296static struct dentry *ceph_fh_to_dentry(struct super_block *sb, 297 struct fid *fid, 298 int fh_len, int fh_type) 299{ 300 struct ceph_nfs_fh *fh = (void *)fid->raw; 301 302 if (fh_type == FILEID_BTRFS_WITH_PARENT) { 303 struct ceph_nfs_snapfh *sfh = (void *)fid->raw; 304 return __snapfh_to_dentry(sb, sfh, false); 305 } 306 307 if (fh_type != FILEID_INO32_GEN && 308 fh_type != FILEID_INO32_GEN_PARENT) 309 return NULL; 310 if (fh_len < sizeof(*fh) / 4) 311 return NULL; 312 313 dout("fh_to_dentry %llx\n", fh->ino); 314 return __fh_to_dentry(sb, fh->ino); 315} 316 317static struct dentry *__get_parent(struct super_block *sb, 318 struct dentry *child, u64 ino) 319{ 320 struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc; 321 struct ceph_mds_request *req; 322 struct inode *inode; 323 int mask; 324 int err; 325 326 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPPARENT, 327 USE_ANY_MDS); 328 if (IS_ERR(req)) 329 return ERR_CAST(req); 330 331 if (child) { 332 req->r_inode = d_inode(child); 333 ihold(d_inode(child)); 334 } else { 335 req->r_ino1 = (struct ceph_vino) { 336 .ino = ino, 337 .snap = CEPH_NOSNAP, 338 }; 339 } 340 341 mask = CEPH_STAT_CAP_INODE; 342 if (ceph_security_xattr_wanted(d_inode(sb->s_root))) 343 mask |= CEPH_CAP_XATTR_SHARED; 344 req->r_args.getattr.mask = cpu_to_le32(mask); 345 346 req->r_num_caps = 1; 347 err = ceph_mdsc_do_request(mdsc, NULL, req); 348 if (err) { 349 ceph_mdsc_put_request(req); 350 return ERR_PTR(err); 351 } 352 353 inode = req->r_target_inode; 354 if (inode) 355 ihold(inode); 356 ceph_mdsc_put_request(req); 357 if (!inode) 358 return ERR_PTR(-ENOENT); 359 360 return d_obtain_alias(inode); 361} 362 363static struct dentry *ceph_get_parent(struct dentry *child) 364{ 365 struct inode *inode = d_inode(child); 366 struct dentry *dn; 367 368 if (ceph_snap(inode) != CEPH_NOSNAP) { 369 struct inode* dir; 370 bool unlinked = false; 371 /* do not support non-directory */ 372 if (!d_is_dir(child)) { 373 dn = ERR_PTR(-EINVAL); 374 goto out; 375 } 376 dir = __lookup_inode(inode->i_sb, ceph_ino(inode)); 377 if (IS_ERR(dir)) { 378 dn = ERR_CAST(dir); 379 goto out; 380 } 381 /* There can be multiple paths to access snapped inode. 382 * For simplicity, treat snapdir of head inode as parent */ 383 if (ceph_snap(inode) != CEPH_SNAPDIR) { 384 struct inode *snapdir = ceph_get_snapdir(dir); 385 if (dir->i_nlink == 0) 386 unlinked = true; 387 iput(dir); 388 if (IS_ERR(snapdir)) { 389 dn = ERR_CAST(snapdir); 390 goto out; 391 } 392 dir = snapdir; 393 } 394 /* If directory has already been deleted, futher get_parent 395 * will fail. Do not mark snapdir dentry as disconnected, 396 * this prevent exportfs from doing futher get_parent. */ 397 if (unlinked) 398 dn = d_obtain_root(dir); 399 else 400 dn = d_obtain_alias(dir); 401 } else { 402 dn = __get_parent(child->d_sb, child, 0); 403 } 404out: 405 dout("get_parent %p ino %llx.%llx err=%ld\n", 406 child, ceph_vinop(inode), (long)PTR_ERR_OR_ZERO(dn)); 407 return dn; 408} 409 410/* 411 * convert regular fh to parent 412 */ 413static struct dentry *ceph_fh_to_parent(struct super_block *sb, 414 struct fid *fid, 415 int fh_len, int fh_type) 416{ 417 struct ceph_nfs_confh *cfh = (void *)fid->raw; 418 struct dentry *dentry; 419 420 if (fh_type == FILEID_BTRFS_WITH_PARENT) { 421 struct ceph_nfs_snapfh *sfh = (void *)fid->raw; 422 return __snapfh_to_dentry(sb, sfh, true); 423 } 424 425 if (fh_type != FILEID_INO32_GEN_PARENT) 426 return NULL; 427 if (fh_len < sizeof(*cfh) / 4) 428 return NULL; 429 430 dout("fh_to_parent %llx\n", cfh->parent_ino); 431 dentry = __get_parent(sb, NULL, cfh->ino); 432 if (unlikely(dentry == ERR_PTR(-ENOENT))) 433 dentry = __fh_to_dentry(sb, cfh->parent_ino); 434 return dentry; 435} 436 437static int __get_snap_name(struct dentry *parent, char *name, 438 struct dentry *child) 439{ 440 struct inode *inode = d_inode(child); 441 struct inode *dir = d_inode(parent); 442 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 443 struct ceph_mds_request *req = NULL; 444 char *last_name = NULL; 445 unsigned next_offset = 2; 446 int err = -EINVAL; 447 448 if (ceph_ino(inode) != ceph_ino(dir)) 449 goto out; 450 if (ceph_snap(inode) == CEPH_SNAPDIR) { 451 if (ceph_snap(dir) == CEPH_NOSNAP) { 452 strcpy(name, fsc->mount_options->snapdir_name); 453 err = 0; 454 } 455 goto out; 456 } 457 if (ceph_snap(dir) != CEPH_SNAPDIR) 458 goto out; 459 460 while (1) { 461 struct ceph_mds_reply_info_parsed *rinfo; 462 struct ceph_mds_reply_dir_entry *rde; 463 int i; 464 465 req = ceph_mdsc_create_request(fsc->mdsc, CEPH_MDS_OP_LSSNAP, 466 USE_AUTH_MDS); 467 if (IS_ERR(req)) { 468 err = PTR_ERR(req); 469 req = NULL; 470 goto out; 471 } 472 err = ceph_alloc_readdir_reply_buffer(req, inode); 473 if (err) 474 goto out; 475 476 req->r_direct_mode = USE_AUTH_MDS; 477 req->r_readdir_offset = next_offset; 478 req->r_args.readdir.flags = 479 cpu_to_le16(CEPH_READDIR_REPLY_BITFLAGS); 480 if (last_name) { 481 req->r_path2 = last_name; 482 last_name = NULL; 483 } 484 485 req->r_inode = dir; 486 ihold(dir); 487 req->r_dentry = dget(parent); 488 489 inode_lock(dir); 490 err = ceph_mdsc_do_request(fsc->mdsc, NULL, req); 491 inode_unlock(dir); 492 493 if (err < 0) 494 goto out; 495 496 rinfo = &req->r_reply_info; 497 for (i = 0; i < rinfo->dir_nr; i++) { 498 rde = rinfo->dir_entries + i; 499 BUG_ON(!rde->inode.in); 500 if (ceph_snap(inode) == 501 le64_to_cpu(rde->inode.in->snapid)) { 502 memcpy(name, rde->name, rde->name_len); 503 name[rde->name_len] = '\0'; 504 err = 0; 505 goto out; 506 } 507 } 508 509 if (rinfo->dir_end) 510 break; 511 512 BUG_ON(rinfo->dir_nr <= 0); 513 rde = rinfo->dir_entries + (rinfo->dir_nr - 1); 514 next_offset += rinfo->dir_nr; 515 last_name = kstrndup(rde->name, rde->name_len, GFP_KERNEL); 516 if (!last_name) { 517 err = -ENOMEM; 518 goto out; 519 } 520 521 ceph_mdsc_put_request(req); 522 req = NULL; 523 } 524 err = -ENOENT; 525out: 526 if (req) 527 ceph_mdsc_put_request(req); 528 kfree(last_name); 529 dout("get_snap_name %p ino %llx.%llx err=%d\n", 530 child, ceph_vinop(inode), err); 531 return err; 532} 533 534static int ceph_get_name(struct dentry *parent, char *name, 535 struct dentry *child) 536{ 537 struct ceph_mds_client *mdsc; 538 struct ceph_mds_request *req; 539 struct inode *dir = d_inode(parent); 540 struct inode *inode = d_inode(child); 541 struct ceph_mds_reply_info_parsed *rinfo; 542 int err; 543 544 if (ceph_snap(inode) != CEPH_NOSNAP) 545 return __get_snap_name(parent, name, child); 546 547 mdsc = ceph_inode_to_client(inode)->mdsc; 548 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPNAME, 549 USE_ANY_MDS); 550 if (IS_ERR(req)) 551 return PTR_ERR(req); 552 553 inode_lock(dir); 554 req->r_inode = inode; 555 ihold(inode); 556 req->r_ino2 = ceph_vino(d_inode(parent)); 557 req->r_parent = dir; 558 ihold(dir); 559 set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags); 560 req->r_num_caps = 2; 561 err = ceph_mdsc_do_request(mdsc, NULL, req); 562 inode_unlock(dir); 563 564 if (err) 565 goto out; 566 567 rinfo = &req->r_reply_info; 568 if (!IS_ENCRYPTED(dir)) { 569 memcpy(name, rinfo->dname, rinfo->dname_len); 570 name[rinfo->dname_len] = 0; 571 } else { 572 struct fscrypt_str oname = FSTR_INIT(NULL, 0); 573 struct ceph_fname fname = { .dir = dir, 574 .name = rinfo->dname, 575 .ctext = rinfo->altname, 576 .name_len = rinfo->dname_len, 577 .ctext_len = rinfo->altname_len }; 578 579 err = ceph_fname_alloc_buffer(dir, &oname); 580 if (err < 0) 581 goto out; 582 583 err = ceph_fname_to_usr(&fname, NULL, &oname, NULL); 584 if (!err) { 585 memcpy(name, oname.name, oname.len); 586 name[oname.len] = 0; 587 } 588 ceph_fname_free_buffer(dir, &oname); 589 } 590out: 591 dout("get_name %p ino %llx.%llx err %d %s%s\n", 592 child, ceph_vinop(inode), err, 593 err ? "" : "name ", err ? "" : name); 594 ceph_mdsc_put_request(req); 595 return err; 596} 597 598const struct export_operations ceph_export_ops = { 599 .encode_fh = ceph_encode_fh, 600 .fh_to_dentry = ceph_fh_to_dentry, 601 .fh_to_parent = ceph_fh_to_parent, 602 .get_parent = ceph_get_parent, 603 .get_name = ceph_get_name, 604}; 605