1 // SPDX-License-Identifier: GPL-2.0 2 #include <linux/ceph/ceph_debug.h> 3 #include <linux/ceph/striper.h> 4 5 #include <linux/module.h> 6 #include <linux/sched.h> 7 #include <linux/slab.h> 8 #include <linux/file.h> 9 #include <linux/mount.h> 10 #include <linux/namei.h> 11 #include <linux/writeback.h> 12 #include <linux/falloc.h> 13 #include <linux/iversion.h> 14 #include <linux/ktime.h> 15 16 #include "super.h" 17 #include "mds_client.h" 18 #include "cache.h" 19 #include "io.h" 20 #include "metric.h" 21 22 static __le32 ceph_flags_sys2wire(u32 flags) 23 { 24 u32 wire_flags = 0; 25 26 switch (flags & O_ACCMODE) { 27 case O_RDONLY: 28 wire_flags |= CEPH_O_RDONLY; 29 break; 30 case O_WRONLY: 31 wire_flags |= CEPH_O_WRONLY; 32 break; 33 case O_RDWR: 34 wire_flags |= CEPH_O_RDWR; 35 break; 36 } 37 38 flags &= ~O_ACCMODE; 39 40 #define ceph_sys2wire(a) if (flags & a) { wire_flags |= CEPH_##a; flags &= ~a; } 41 42 ceph_sys2wire(O_CREAT); 43 ceph_sys2wire(O_EXCL); 44 ceph_sys2wire(O_TRUNC); 45 ceph_sys2wire(O_DIRECTORY); 46 ceph_sys2wire(O_NOFOLLOW); 47 48 #undef ceph_sys2wire 49 50 if (flags) 51 dout("unused open flags: %x\n", flags); 52 53 return cpu_to_le32(wire_flags); 54 } 55 56 /* 57 * Ceph file operations 58 * 59 * Implement basic open/close functionality, and implement 60 * read/write. 61 * 62 * We implement three modes of file I/O: 63 * - buffered uses the generic_file_aio_{read,write} helpers 64 * 65 * - synchronous is used when there is multi-client read/write 66 * sharing, avoids the page cache, and synchronously waits for an 67 * ack from the OSD. 68 * 69 * - direct io takes the variant of the sync path that references 70 * user pages directly. 71 * 72 * fsync() flushes and waits on dirty pages, but just queues metadata 73 * for writeback: since the MDS can recover size and mtime there is no 74 * need to wait for MDS acknowledgement. 75 */ 76 77 /* 78 * How many pages to get in one call to iov_iter_get_pages(). This 79 * determines the size of the on-stack array used as a buffer. 80 */ 81 #define ITER_GET_BVECS_PAGES 64 82 83 static ssize_t __iter_get_bvecs(struct iov_iter *iter, size_t maxsize, 84 struct bio_vec *bvecs) 85 { 86 size_t size = 0; 87 int bvec_idx = 0; 88 89 if (maxsize > iov_iter_count(iter)) 90 maxsize = iov_iter_count(iter); 91 92 while (size < maxsize) { 93 struct page *pages[ITER_GET_BVECS_PAGES]; 94 ssize_t bytes; 95 size_t start; 96 int idx = 0; 97 98 bytes = iov_iter_get_pages(iter, pages, maxsize - size, 99 ITER_GET_BVECS_PAGES, &start); 100 if (bytes < 0) 101 return size ?: bytes; 102 103 iov_iter_advance(iter, bytes); 104 size += bytes; 105 106 for ( ; bytes; idx++, bvec_idx++) { 107 struct bio_vec bv = { 108 .bv_page = pages[idx], 109 .bv_len = min_t(int, bytes, PAGE_SIZE - start), 110 .bv_offset = start, 111 }; 112 113 bvecs[bvec_idx] = bv; 114 bytes -= bv.bv_len; 115 start = 0; 116 } 117 } 118 119 return size; 120 } 121 122 /* 123 * iov_iter_get_pages() only considers one iov_iter segment, no matter 124 * what maxsize or maxpages are given. For ITER_BVEC that is a single 125 * page. 126 * 127 * Attempt to get up to @maxsize bytes worth of pages from @iter. 128 * Return the number of bytes in the created bio_vec array, or an error. 129 */ 130 static ssize_t iter_get_bvecs_alloc(struct iov_iter *iter, size_t maxsize, 131 struct bio_vec **bvecs, int *num_bvecs) 132 { 133 struct bio_vec *bv; 134 size_t orig_count = iov_iter_count(iter); 135 ssize_t bytes; 136 int npages; 137 138 iov_iter_truncate(iter, maxsize); 139 npages = iov_iter_npages(iter, INT_MAX); 140 iov_iter_reexpand(iter, orig_count); 141 142 /* 143 * __iter_get_bvecs() may populate only part of the array -- zero it 144 * out. 145 */ 146 bv = kvmalloc_array(npages, sizeof(*bv), GFP_KERNEL | __GFP_ZERO); 147 if (!bv) 148 return -ENOMEM; 149 150 bytes = __iter_get_bvecs(iter, maxsize, bv); 151 if (bytes < 0) { 152 /* 153 * No pages were pinned -- just free the array. 154 */ 155 kvfree(bv); 156 return bytes; 157 } 158 159 *bvecs = bv; 160 *num_bvecs = npages; 161 return bytes; 162 } 163 164 static void put_bvecs(struct bio_vec *bvecs, int num_bvecs, bool should_dirty) 165 { 166 int i; 167 168 for (i = 0; i < num_bvecs; i++) { 169 if (bvecs[i].bv_page) { 170 if (should_dirty) 171 set_page_dirty_lock(bvecs[i].bv_page); 172 put_page(bvecs[i].bv_page); 173 } 174 } 175 kvfree(bvecs); 176 } 177 178 /* 179 * Prepare an open request. Preallocate ceph_cap to avoid an 180 * inopportune ENOMEM later. 181 */ 182 static struct ceph_mds_request * 183 prepare_open_request(struct super_block *sb, int flags, int create_mode) 184 { 185 struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(sb); 186 struct ceph_mds_request *req; 187 int want_auth = USE_ANY_MDS; 188 int op = (flags & O_CREAT) ? CEPH_MDS_OP_CREATE : CEPH_MDS_OP_OPEN; 189 190 if (flags & (O_WRONLY|O_RDWR|O_CREAT|O_TRUNC)) 191 want_auth = USE_AUTH_MDS; 192 193 req = ceph_mdsc_create_request(mdsc, op, want_auth); 194 if (IS_ERR(req)) 195 goto out; 196 req->r_fmode = ceph_flags_to_mode(flags); 197 req->r_args.open.flags = ceph_flags_sys2wire(flags); 198 req->r_args.open.mode = cpu_to_le32(create_mode); 199 out: 200 return req; 201 } 202 203 static int ceph_init_file_info(struct inode *inode, struct file *file, 204 int fmode, bool isdir) 205 { 206 struct ceph_inode_info *ci = ceph_inode(inode); 207 struct ceph_file_info *fi; 208 209 dout("%s %p %p 0%o (%s)\n", __func__, inode, file, 210 inode->i_mode, isdir ? "dir" : "regular"); 211 BUG_ON(inode->i_fop->release != ceph_release); 212 213 if (isdir) { 214 struct ceph_dir_file_info *dfi = 215 kmem_cache_zalloc(ceph_dir_file_cachep, GFP_KERNEL); 216 if (!dfi) 217 return -ENOMEM; 218 219 file->private_data = dfi; 220 fi = &dfi->file_info; 221 dfi->next_offset = 2; 222 dfi->readdir_cache_idx = -1; 223 } else { 224 fi = kmem_cache_zalloc(ceph_file_cachep, GFP_KERNEL); 225 if (!fi) 226 return -ENOMEM; 227 228 file->private_data = fi; 229 } 230 231 ceph_get_fmode(ci, fmode, 1); 232 fi->fmode = fmode; 233 234 spin_lock_init(&fi->rw_contexts_lock); 235 INIT_LIST_HEAD(&fi->rw_contexts); 236 fi->meta_err = errseq_sample(&ci->i_meta_err); 237 fi->filp_gen = READ_ONCE(ceph_inode_to_client(inode)->filp_gen); 238 239 return 0; 240 } 241 242 /* 243 * initialize private struct file data. 244 * if we fail, clean up by dropping fmode reference on the ceph_inode 245 */ 246 static int ceph_init_file(struct inode *inode, struct file *file, int fmode) 247 { 248 int ret = 0; 249 250 switch (inode->i_mode & S_IFMT) { 251 case S_IFREG: 252 ceph_fscache_register_inode_cookie(inode); 253 ceph_fscache_file_set_cookie(inode, file); 254 fallthrough; 255 case S_IFDIR: 256 ret = ceph_init_file_info(inode, file, fmode, 257 S_ISDIR(inode->i_mode)); 258 break; 259 260 case S_IFLNK: 261 dout("init_file %p %p 0%o (symlink)\n", inode, file, 262 inode->i_mode); 263 break; 264 265 default: 266 dout("init_file %p %p 0%o (special)\n", inode, file, 267 inode->i_mode); 268 /* 269 * we need to drop the open ref now, since we don't 270 * have .release set to ceph_release. 271 */ 272 BUG_ON(inode->i_fop->release == ceph_release); 273 274 /* call the proper open fop */ 275 ret = inode->i_fop->open(inode, file); 276 } 277 return ret; 278 } 279 280 /* 281 * try renew caps after session gets killed. 282 */ 283 int ceph_renew_caps(struct inode *inode, int fmode) 284 { 285 struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb); 286 struct ceph_inode_info *ci = ceph_inode(inode); 287 struct ceph_mds_request *req; 288 int err, flags, wanted; 289 290 spin_lock(&ci->i_ceph_lock); 291 __ceph_touch_fmode(ci, mdsc, fmode); 292 wanted = __ceph_caps_file_wanted(ci); 293 if (__ceph_is_any_real_caps(ci) && 294 (!(wanted & CEPH_CAP_ANY_WR) || ci->i_auth_cap)) { 295 int issued = __ceph_caps_issued(ci, NULL); 296 spin_unlock(&ci->i_ceph_lock); 297 dout("renew caps %p want %s issued %s updating mds_wanted\n", 298 inode, ceph_cap_string(wanted), ceph_cap_string(issued)); 299 ceph_check_caps(ci, 0, NULL); 300 return 0; 301 } 302 spin_unlock(&ci->i_ceph_lock); 303 304 flags = 0; 305 if ((wanted & CEPH_CAP_FILE_RD) && (wanted & CEPH_CAP_FILE_WR)) 306 flags = O_RDWR; 307 else if (wanted & CEPH_CAP_FILE_RD) 308 flags = O_RDONLY; 309 else if (wanted & CEPH_CAP_FILE_WR) 310 flags = O_WRONLY; 311 #ifdef O_LAZY 312 if (wanted & CEPH_CAP_FILE_LAZYIO) 313 flags |= O_LAZY; 314 #endif 315 316 req = prepare_open_request(inode->i_sb, flags, 0); 317 if (IS_ERR(req)) { 318 err = PTR_ERR(req); 319 goto out; 320 } 321 322 req->r_inode = inode; 323 ihold(inode); 324 req->r_num_caps = 1; 325 326 err = ceph_mdsc_do_request(mdsc, NULL, req); 327 ceph_mdsc_put_request(req); 328 out: 329 dout("renew caps %p open result=%d\n", inode, err); 330 return err < 0 ? err : 0; 331 } 332 333 /* 334 * If we already have the requisite capabilities, we can satisfy 335 * the open request locally (no need to request new caps from the 336 * MDS). We do, however, need to inform the MDS (asynchronously) 337 * if our wanted caps set expands. 338 */ 339 int ceph_open(struct inode *inode, struct file *file) 340 { 341 struct ceph_inode_info *ci = ceph_inode(inode); 342 struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb); 343 struct ceph_mds_client *mdsc = fsc->mdsc; 344 struct ceph_mds_request *req; 345 struct ceph_file_info *fi = file->private_data; 346 int err; 347 int flags, fmode, wanted; 348 349 if (fi) { 350 dout("open file %p is already opened\n", file); 351 return 0; 352 } 353 354 /* filter out O_CREAT|O_EXCL; vfs did that already. yuck. */ 355 flags = file->f_flags & ~(O_CREAT|O_EXCL); 356 if (S_ISDIR(inode->i_mode)) 357 flags = O_DIRECTORY; /* mds likes to know */ 358 359 dout("open inode %p ino %llx.%llx file %p flags %d (%d)\n", inode, 360 ceph_vinop(inode), file, flags, file->f_flags); 361 fmode = ceph_flags_to_mode(flags); 362 wanted = ceph_caps_for_mode(fmode); 363 364 /* snapped files are read-only */ 365 if (ceph_snap(inode) != CEPH_NOSNAP && (file->f_mode & FMODE_WRITE)) 366 return -EROFS; 367 368 /* trivially open snapdir */ 369 if (ceph_snap(inode) == CEPH_SNAPDIR) { 370 return ceph_init_file(inode, file, fmode); 371 } 372 373 /* 374 * No need to block if we have caps on the auth MDS (for 375 * write) or any MDS (for read). Update wanted set 376 * asynchronously. 377 */ 378 spin_lock(&ci->i_ceph_lock); 379 if (__ceph_is_any_real_caps(ci) && 380 (((fmode & CEPH_FILE_MODE_WR) == 0) || ci->i_auth_cap)) { 381 int mds_wanted = __ceph_caps_mds_wanted(ci, true); 382 int issued = __ceph_caps_issued(ci, NULL); 383 384 dout("open %p fmode %d want %s issued %s using existing\n", 385 inode, fmode, ceph_cap_string(wanted), 386 ceph_cap_string(issued)); 387 __ceph_touch_fmode(ci, mdsc, fmode); 388 spin_unlock(&ci->i_ceph_lock); 389 390 /* adjust wanted? */ 391 if ((issued & wanted) != wanted && 392 (mds_wanted & wanted) != wanted && 393 ceph_snap(inode) != CEPH_SNAPDIR) 394 ceph_check_caps(ci, 0, NULL); 395 396 return ceph_init_file(inode, file, fmode); 397 } else if (ceph_snap(inode) != CEPH_NOSNAP && 398 (ci->i_snap_caps & wanted) == wanted) { 399 __ceph_touch_fmode(ci, mdsc, fmode); 400 spin_unlock(&ci->i_ceph_lock); 401 return ceph_init_file(inode, file, fmode); 402 } 403 404 spin_unlock(&ci->i_ceph_lock); 405 406 dout("open fmode %d wants %s\n", fmode, ceph_cap_string(wanted)); 407 req = prepare_open_request(inode->i_sb, flags, 0); 408 if (IS_ERR(req)) { 409 err = PTR_ERR(req); 410 goto out; 411 } 412 req->r_inode = inode; 413 ihold(inode); 414 415 req->r_num_caps = 1; 416 err = ceph_mdsc_do_request(mdsc, NULL, req); 417 if (!err) 418 err = ceph_init_file(inode, file, req->r_fmode); 419 ceph_mdsc_put_request(req); 420 dout("open result=%d on %llx.%llx\n", err, ceph_vinop(inode)); 421 out: 422 return err; 423 } 424 425 /* Clone the layout from a synchronous create, if the dir now has Dc caps */ 426 static void 427 cache_file_layout(struct inode *dst, struct inode *src) 428 { 429 struct ceph_inode_info *cdst = ceph_inode(dst); 430 struct ceph_inode_info *csrc = ceph_inode(src); 431 432 spin_lock(&cdst->i_ceph_lock); 433 if ((__ceph_caps_issued(cdst, NULL) & CEPH_CAP_DIR_CREATE) && 434 !ceph_file_layout_is_valid(&cdst->i_cached_layout)) { 435 memcpy(&cdst->i_cached_layout, &csrc->i_layout, 436 sizeof(cdst->i_cached_layout)); 437 rcu_assign_pointer(cdst->i_cached_layout.pool_ns, 438 ceph_try_get_string(csrc->i_layout.pool_ns)); 439 } 440 spin_unlock(&cdst->i_ceph_lock); 441 } 442 443 /* 444 * Try to set up an async create. We need caps, a file layout, and inode number, 445 * and either a lease on the dentry or complete dir info. If any of those 446 * criteria are not satisfied, then return false and the caller can go 447 * synchronous. 448 */ 449 static int try_prep_async_create(struct inode *dir, struct dentry *dentry, 450 struct ceph_file_layout *lo, u64 *pino) 451 { 452 struct ceph_inode_info *ci = ceph_inode(dir); 453 struct ceph_dentry_info *di = ceph_dentry(dentry); 454 int got = 0, want = CEPH_CAP_FILE_EXCL | CEPH_CAP_DIR_CREATE; 455 u64 ino; 456 457 spin_lock(&ci->i_ceph_lock); 458 /* No auth cap means no chance for Dc caps */ 459 if (!ci->i_auth_cap) 460 goto no_async; 461 462 /* Any delegated inos? */ 463 if (xa_empty(&ci->i_auth_cap->session->s_delegated_inos)) 464 goto no_async; 465 466 if (!ceph_file_layout_is_valid(&ci->i_cached_layout)) 467 goto no_async; 468 469 if ((__ceph_caps_issued(ci, NULL) & want) != want) 470 goto no_async; 471 472 if (d_in_lookup(dentry)) { 473 if (!__ceph_dir_is_complete(ci)) 474 goto no_async; 475 spin_lock(&dentry->d_lock); 476 di->lease_shared_gen = atomic_read(&ci->i_shared_gen); 477 spin_unlock(&dentry->d_lock); 478 } else if (atomic_read(&ci->i_shared_gen) != 479 READ_ONCE(di->lease_shared_gen)) { 480 goto no_async; 481 } 482 483 ino = ceph_get_deleg_ino(ci->i_auth_cap->session); 484 if (!ino) 485 goto no_async; 486 487 *pino = ino; 488 ceph_take_cap_refs(ci, want, false); 489 memcpy(lo, &ci->i_cached_layout, sizeof(*lo)); 490 rcu_assign_pointer(lo->pool_ns, 491 ceph_try_get_string(ci->i_cached_layout.pool_ns)); 492 got = want; 493 no_async: 494 spin_unlock(&ci->i_ceph_lock); 495 return got; 496 } 497 498 static void restore_deleg_ino(struct inode *dir, u64 ino) 499 { 500 struct ceph_inode_info *ci = ceph_inode(dir); 501 struct ceph_mds_session *s = NULL; 502 503 spin_lock(&ci->i_ceph_lock); 504 if (ci->i_auth_cap) 505 s = ceph_get_mds_session(ci->i_auth_cap->session); 506 spin_unlock(&ci->i_ceph_lock); 507 if (s) { 508 int err = ceph_restore_deleg_ino(s, ino); 509 if (err) 510 pr_warn("ceph: unable to restore delegated ino 0x%llx to session: %d\n", 511 ino, err); 512 ceph_put_mds_session(s); 513 } 514 } 515 516 static void ceph_async_create_cb(struct ceph_mds_client *mdsc, 517 struct ceph_mds_request *req) 518 { 519 int result = req->r_err ? req->r_err : 520 le32_to_cpu(req->r_reply_info.head->result); 521 522 if (result == -EJUKEBOX) 523 goto out; 524 525 mapping_set_error(req->r_parent->i_mapping, result); 526 527 if (result) { 528 struct dentry *dentry = req->r_dentry; 529 int pathlen = 0; 530 u64 base = 0; 531 char *path = ceph_mdsc_build_path(req->r_dentry, &pathlen, 532 &base, 0); 533 534 ceph_dir_clear_complete(req->r_parent); 535 if (!d_unhashed(dentry)) 536 d_drop(dentry); 537 538 /* FIXME: start returning I/O errors on all accesses? */ 539 pr_warn("ceph: async create failure path=(%llx)%s result=%d!\n", 540 base, IS_ERR(path) ? "<<bad>>" : path, result); 541 ceph_mdsc_free_path(path, pathlen); 542 } 543 544 if (req->r_target_inode) { 545 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode); 546 u64 ino = ceph_vino(req->r_target_inode).ino; 547 548 if (req->r_deleg_ino != ino) 549 pr_warn("%s: inode number mismatch! err=%d deleg_ino=0x%llx target=0x%llx\n", 550 __func__, req->r_err, req->r_deleg_ino, ino); 551 mapping_set_error(req->r_target_inode->i_mapping, result); 552 553 spin_lock(&ci->i_ceph_lock); 554 if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE) { 555 ci->i_ceph_flags &= ~CEPH_I_ASYNC_CREATE; 556 wake_up_bit(&ci->i_ceph_flags, CEPH_ASYNC_CREATE_BIT); 557 } 558 ceph_kick_flushing_inode_caps(req->r_session, ci); 559 spin_unlock(&ci->i_ceph_lock); 560 } else { 561 pr_warn("%s: no req->r_target_inode for 0x%llx\n", __func__, 562 req->r_deleg_ino); 563 } 564 out: 565 ceph_mdsc_release_dir_caps(req); 566 } 567 568 static int ceph_finish_async_create(struct inode *dir, struct dentry *dentry, 569 struct file *file, umode_t mode, 570 struct ceph_mds_request *req, 571 struct ceph_acl_sec_ctx *as_ctx, 572 struct ceph_file_layout *lo) 573 { 574 int ret; 575 char xattr_buf[4]; 576 struct ceph_mds_reply_inode in = { }; 577 struct ceph_mds_reply_info_in iinfo = { .in = &in }; 578 struct ceph_inode_info *ci = ceph_inode(dir); 579 struct inode *inode; 580 struct timespec64 now; 581 struct ceph_vino vino = { .ino = req->r_deleg_ino, 582 .snap = CEPH_NOSNAP }; 583 584 ktime_get_real_ts64(&now); 585 586 inode = ceph_get_inode(dentry->d_sb, vino); 587 if (IS_ERR(inode)) 588 return PTR_ERR(inode); 589 590 iinfo.inline_version = CEPH_INLINE_NONE; 591 iinfo.change_attr = 1; 592 ceph_encode_timespec64(&iinfo.btime, &now); 593 594 iinfo.xattr_len = ARRAY_SIZE(xattr_buf); 595 iinfo.xattr_data = xattr_buf; 596 memset(iinfo.xattr_data, 0, iinfo.xattr_len); 597 598 in.ino = cpu_to_le64(vino.ino); 599 in.snapid = cpu_to_le64(CEPH_NOSNAP); 600 in.version = cpu_to_le64(1); // ??? 601 in.cap.caps = in.cap.wanted = cpu_to_le32(CEPH_CAP_ALL_FILE); 602 in.cap.cap_id = cpu_to_le64(1); 603 in.cap.realm = cpu_to_le64(ci->i_snap_realm->ino); 604 in.cap.flags = CEPH_CAP_FLAG_AUTH; 605 in.ctime = in.mtime = in.atime = iinfo.btime; 606 in.mode = cpu_to_le32((u32)mode); 607 in.truncate_seq = cpu_to_le32(1); 608 in.truncate_size = cpu_to_le64(-1ULL); 609 in.xattr_version = cpu_to_le64(1); 610 in.uid = cpu_to_le32(from_kuid(&init_user_ns, current_fsuid())); 611 in.gid = cpu_to_le32(from_kgid(&init_user_ns, dir->i_mode & S_ISGID ? 612 dir->i_gid : current_fsgid())); 613 in.nlink = cpu_to_le32(1); 614 in.max_size = cpu_to_le64(lo->stripe_unit); 615 616 ceph_file_layout_to_legacy(lo, &in.layout); 617 618 ret = ceph_fill_inode(inode, NULL, &iinfo, NULL, req->r_session, 619 req->r_fmode, NULL); 620 if (ret) { 621 dout("%s failed to fill inode: %d\n", __func__, ret); 622 ceph_dir_clear_complete(dir); 623 if (!d_unhashed(dentry)) 624 d_drop(dentry); 625 if (inode->i_state & I_NEW) 626 discard_new_inode(inode); 627 } else { 628 struct dentry *dn; 629 630 dout("%s d_adding new inode 0x%llx to 0x%llx/%s\n", __func__, 631 vino.ino, ceph_ino(dir), dentry->d_name.name); 632 ceph_dir_clear_ordered(dir); 633 ceph_init_inode_acls(inode, as_ctx); 634 if (inode->i_state & I_NEW) { 635 /* 636 * If it's not I_NEW, then someone created this before 637 * we got here. Assume the server is aware of it at 638 * that point and don't worry about setting 639 * CEPH_I_ASYNC_CREATE. 640 */ 641 ceph_inode(inode)->i_ceph_flags = CEPH_I_ASYNC_CREATE; 642 unlock_new_inode(inode); 643 } 644 if (d_in_lookup(dentry) || d_really_is_negative(dentry)) { 645 if (!d_unhashed(dentry)) 646 d_drop(dentry); 647 dn = d_splice_alias(inode, dentry); 648 WARN_ON_ONCE(dn && dn != dentry); 649 } 650 file->f_mode |= FMODE_CREATED; 651 ret = finish_open(file, dentry, ceph_open); 652 } 653 return ret; 654 } 655 656 /* 657 * Do a lookup + open with a single request. If we get a non-existent 658 * file or symlink, return 1 so the VFS can retry. 659 */ 660 int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 661 struct file *file, unsigned flags, umode_t mode) 662 { 663 struct ceph_fs_client *fsc = ceph_sb_to_client(dir->i_sb); 664 struct ceph_mds_client *mdsc = fsc->mdsc; 665 struct ceph_mds_request *req; 666 struct dentry *dn; 667 struct ceph_acl_sec_ctx as_ctx = {}; 668 bool try_async = ceph_test_mount_opt(fsc, ASYNC_DIROPS); 669 int mask; 670 int err; 671 672 dout("atomic_open %p dentry %p '%pd' %s flags %d mode 0%o\n", 673 dir, dentry, dentry, 674 d_unhashed(dentry) ? "unhashed" : "hashed", flags, mode); 675 676 if (dentry->d_name.len > NAME_MAX) 677 return -ENAMETOOLONG; 678 679 if (flags & O_CREAT) { 680 if (ceph_quota_is_max_files_exceeded(dir)) 681 return -EDQUOT; 682 err = ceph_pre_init_acls(dir, &mode, &as_ctx); 683 if (err < 0) 684 return err; 685 err = ceph_security_init_secctx(dentry, mode, &as_ctx); 686 if (err < 0) 687 goto out_ctx; 688 } else if (!d_in_lookup(dentry)) { 689 /* If it's not being looked up, it's negative */ 690 return -ENOENT; 691 } 692 retry: 693 /* do the open */ 694 req = prepare_open_request(dir->i_sb, flags, mode); 695 if (IS_ERR(req)) { 696 err = PTR_ERR(req); 697 goto out_ctx; 698 } 699 req->r_dentry = dget(dentry); 700 req->r_num_caps = 2; 701 mask = CEPH_STAT_CAP_INODE | CEPH_CAP_AUTH_SHARED; 702 if (ceph_security_xattr_wanted(dir)) 703 mask |= CEPH_CAP_XATTR_SHARED; 704 req->r_args.open.mask = cpu_to_le32(mask); 705 req->r_parent = dir; 706 707 if (flags & O_CREAT) { 708 struct ceph_file_layout lo; 709 710 req->r_dentry_drop = CEPH_CAP_FILE_SHARED | CEPH_CAP_AUTH_EXCL; 711 req->r_dentry_unless = CEPH_CAP_FILE_EXCL; 712 if (as_ctx.pagelist) { 713 req->r_pagelist = as_ctx.pagelist; 714 as_ctx.pagelist = NULL; 715 } 716 if (try_async && 717 (req->r_dir_caps = 718 try_prep_async_create(dir, dentry, &lo, 719 &req->r_deleg_ino))) { 720 set_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags); 721 req->r_args.open.flags |= cpu_to_le32(CEPH_O_EXCL); 722 req->r_callback = ceph_async_create_cb; 723 err = ceph_mdsc_submit_request(mdsc, dir, req); 724 if (!err) { 725 err = ceph_finish_async_create(dir, dentry, 726 file, mode, req, 727 &as_ctx, &lo); 728 } else if (err == -EJUKEBOX) { 729 restore_deleg_ino(dir, req->r_deleg_ino); 730 ceph_mdsc_put_request(req); 731 try_async = false; 732 goto retry; 733 } 734 goto out_req; 735 } 736 } 737 738 set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags); 739 err = ceph_mdsc_do_request(mdsc, 740 (flags & (O_CREAT|O_TRUNC)) ? dir : NULL, 741 req); 742 err = ceph_handle_snapdir(req, dentry, err); 743 if (err) 744 goto out_req; 745 746 if ((flags & O_CREAT) && !req->r_reply_info.head->is_dentry) 747 err = ceph_handle_notrace_create(dir, dentry); 748 749 if (d_in_lookup(dentry)) { 750 dn = ceph_finish_lookup(req, dentry, err); 751 if (IS_ERR(dn)) 752 err = PTR_ERR(dn); 753 } else { 754 /* we were given a hashed negative dentry */ 755 dn = NULL; 756 } 757 if (err) 758 goto out_req; 759 if (dn || d_really_is_negative(dentry) || d_is_symlink(dentry)) { 760 /* make vfs retry on splice, ENOENT, or symlink */ 761 dout("atomic_open finish_no_open on dn %p\n", dn); 762 err = finish_no_open(file, dn); 763 } else { 764 dout("atomic_open finish_open on dn %p\n", dn); 765 if (req->r_op == CEPH_MDS_OP_CREATE && req->r_reply_info.has_create_ino) { 766 struct inode *newino = d_inode(dentry); 767 768 cache_file_layout(dir, newino); 769 ceph_init_inode_acls(newino, &as_ctx); 770 file->f_mode |= FMODE_CREATED; 771 } 772 err = finish_open(file, dentry, ceph_open); 773 } 774 out_req: 775 ceph_mdsc_put_request(req); 776 out_ctx: 777 ceph_release_acl_sec_ctx(&as_ctx); 778 dout("atomic_open result=%d\n", err); 779 return err; 780 } 781 782 int ceph_release(struct inode *inode, struct file *file) 783 { 784 struct ceph_inode_info *ci = ceph_inode(inode); 785 786 if (S_ISDIR(inode->i_mode)) { 787 struct ceph_dir_file_info *dfi = file->private_data; 788 dout("release inode %p dir file %p\n", inode, file); 789 WARN_ON(!list_empty(&dfi->file_info.rw_contexts)); 790 791 ceph_put_fmode(ci, dfi->file_info.fmode, 1); 792 793 if (dfi->last_readdir) 794 ceph_mdsc_put_request(dfi->last_readdir); 795 kfree(dfi->last_name); 796 kfree(dfi->dir_info); 797 kmem_cache_free(ceph_dir_file_cachep, dfi); 798 } else { 799 struct ceph_file_info *fi = file->private_data; 800 dout("release inode %p regular file %p\n", inode, file); 801 WARN_ON(!list_empty(&fi->rw_contexts)); 802 803 ceph_put_fmode(ci, fi->fmode, 1); 804 805 kmem_cache_free(ceph_file_cachep, fi); 806 } 807 808 /* wake up anyone waiting for caps on this inode */ 809 wake_up_all(&ci->i_cap_wq); 810 return 0; 811 } 812 813 enum { 814 HAVE_RETRIED = 1, 815 CHECK_EOF = 2, 816 READ_INLINE = 3, 817 }; 818 819 /* 820 * Completely synchronous read and write methods. Direct from __user 821 * buffer to osd, or directly to user pages (if O_DIRECT). 822 * 823 * If the read spans object boundary, just do multiple reads. (That's not 824 * atomic, but good enough for now.) 825 * 826 * If we get a short result from the OSD, check against i_size; we need to 827 * only return a short read to the caller if we hit EOF. 828 */ 829 static ssize_t ceph_sync_read(struct kiocb *iocb, struct iov_iter *to, 830 int *retry_op) 831 { 832 struct file *file = iocb->ki_filp; 833 struct inode *inode = file_inode(file); 834 struct ceph_inode_info *ci = ceph_inode(inode); 835 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 836 struct ceph_osd_client *osdc = &fsc->client->osdc; 837 ssize_t ret; 838 u64 off = iocb->ki_pos; 839 u64 len = iov_iter_count(to); 840 841 dout("sync_read on file %p %llu~%u %s\n", file, off, (unsigned)len, 842 (file->f_flags & O_DIRECT) ? "O_DIRECT" : ""); 843 844 if (!len) 845 return 0; 846 /* 847 * flush any page cache pages in this range. this 848 * will make concurrent normal and sync io slow, 849 * but it will at least behave sensibly when they are 850 * in sequence. 851 */ 852 ret = filemap_write_and_wait_range(inode->i_mapping, 853 off, off + len - 1); 854 if (ret < 0) 855 return ret; 856 857 ret = 0; 858 while ((len = iov_iter_count(to)) > 0) { 859 struct ceph_osd_request *req; 860 struct page **pages; 861 int num_pages; 862 size_t page_off; 863 u64 i_size; 864 bool more; 865 int idx; 866 size_t left; 867 868 req = ceph_osdc_new_request(osdc, &ci->i_layout, 869 ci->i_vino, off, &len, 0, 1, 870 CEPH_OSD_OP_READ, CEPH_OSD_FLAG_READ, 871 NULL, ci->i_truncate_seq, 872 ci->i_truncate_size, false); 873 if (IS_ERR(req)) { 874 ret = PTR_ERR(req); 875 break; 876 } 877 878 more = len < iov_iter_count(to); 879 880 num_pages = calc_pages_for(off, len); 881 page_off = off & ~PAGE_MASK; 882 pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 883 if (IS_ERR(pages)) { 884 ceph_osdc_put_request(req); 885 ret = PTR_ERR(pages); 886 break; 887 } 888 889 osd_req_op_extent_osd_data_pages(req, 0, pages, len, page_off, 890 false, false); 891 ret = ceph_osdc_start_request(osdc, req, false); 892 if (!ret) 893 ret = ceph_osdc_wait_request(osdc, req); 894 895 ceph_update_read_latency(&fsc->mdsc->metric, 896 req->r_start_latency, 897 req->r_end_latency, 898 ret); 899 900 ceph_osdc_put_request(req); 901 902 i_size = i_size_read(inode); 903 dout("sync_read %llu~%llu got %zd i_size %llu%s\n", 904 off, len, ret, i_size, (more ? " MORE" : "")); 905 906 if (ret == -ENOENT) 907 ret = 0; 908 if (ret >= 0 && ret < len && (off + ret < i_size)) { 909 int zlen = min(len - ret, i_size - off - ret); 910 int zoff = page_off + ret; 911 dout("sync_read zero gap %llu~%llu\n", 912 off + ret, off + ret + zlen); 913 ceph_zero_page_vector_range(zoff, zlen, pages); 914 ret += zlen; 915 } 916 917 idx = 0; 918 left = ret > 0 ? ret : 0; 919 while (left > 0) { 920 size_t len, copied; 921 page_off = off & ~PAGE_MASK; 922 len = min_t(size_t, left, PAGE_SIZE - page_off); 923 SetPageUptodate(pages[idx]); 924 copied = copy_page_to_iter(pages[idx++], 925 page_off, len, to); 926 off += copied; 927 left -= copied; 928 if (copied < len) { 929 ret = -EFAULT; 930 break; 931 } 932 } 933 ceph_release_page_vector(pages, num_pages); 934 935 if (ret < 0) { 936 if (ret == -EBLOCKLISTED) 937 fsc->blocklisted = true; 938 break; 939 } 940 941 if (off >= i_size || !more) 942 break; 943 } 944 945 if (off > iocb->ki_pos) { 946 if (ret >= 0 && 947 iov_iter_count(to) > 0 && off >= i_size_read(inode)) 948 *retry_op = CHECK_EOF; 949 ret = off - iocb->ki_pos; 950 iocb->ki_pos = off; 951 } 952 953 dout("sync_read result %zd retry_op %d\n", ret, *retry_op); 954 return ret; 955 } 956 957 struct ceph_aio_request { 958 struct kiocb *iocb; 959 size_t total_len; 960 bool write; 961 bool should_dirty; 962 int error; 963 struct list_head osd_reqs; 964 unsigned num_reqs; 965 atomic_t pending_reqs; 966 struct timespec64 mtime; 967 struct ceph_cap_flush *prealloc_cf; 968 }; 969 970 struct ceph_aio_work { 971 struct work_struct work; 972 struct ceph_osd_request *req; 973 }; 974 975 static void ceph_aio_retry_work(struct work_struct *work); 976 977 static void ceph_aio_complete(struct inode *inode, 978 struct ceph_aio_request *aio_req) 979 { 980 struct ceph_inode_info *ci = ceph_inode(inode); 981 int ret; 982 983 if (!atomic_dec_and_test(&aio_req->pending_reqs)) 984 return; 985 986 if (aio_req->iocb->ki_flags & IOCB_DIRECT) 987 inode_dio_end(inode); 988 989 ret = aio_req->error; 990 if (!ret) 991 ret = aio_req->total_len; 992 993 dout("ceph_aio_complete %p rc %d\n", inode, ret); 994 995 if (ret >= 0 && aio_req->write) { 996 int dirty; 997 998 loff_t endoff = aio_req->iocb->ki_pos + aio_req->total_len; 999 if (endoff > i_size_read(inode)) { 1000 if (ceph_inode_set_size(inode, endoff)) 1001 ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL); 1002 } 1003 1004 spin_lock(&ci->i_ceph_lock); 1005 ci->i_inline_version = CEPH_INLINE_NONE; 1006 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, 1007 &aio_req->prealloc_cf); 1008 spin_unlock(&ci->i_ceph_lock); 1009 if (dirty) 1010 __mark_inode_dirty(inode, dirty); 1011 1012 } 1013 1014 ceph_put_cap_refs(ci, (aio_req->write ? CEPH_CAP_FILE_WR : 1015 CEPH_CAP_FILE_RD)); 1016 1017 aio_req->iocb->ki_complete(aio_req->iocb, ret, 0); 1018 1019 ceph_free_cap_flush(aio_req->prealloc_cf); 1020 kfree(aio_req); 1021 } 1022 1023 static void ceph_aio_complete_req(struct ceph_osd_request *req) 1024 { 1025 int rc = req->r_result; 1026 struct inode *inode = req->r_inode; 1027 struct ceph_aio_request *aio_req = req->r_priv; 1028 struct ceph_osd_data *osd_data = osd_req_op_extent_osd_data(req, 0); 1029 struct ceph_client_metric *metric = &ceph_sb_to_mdsc(inode->i_sb)->metric; 1030 1031 BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_BVECS); 1032 BUG_ON(!osd_data->num_bvecs); 1033 1034 dout("ceph_aio_complete_req %p rc %d bytes %u\n", 1035 inode, rc, osd_data->bvec_pos.iter.bi_size); 1036 1037 /* r_start_latency == 0 means the request was not submitted */ 1038 if (req->r_start_latency) { 1039 if (aio_req->write) 1040 ceph_update_write_latency(metric, req->r_start_latency, 1041 req->r_end_latency, rc); 1042 else 1043 ceph_update_read_latency(metric, req->r_start_latency, 1044 req->r_end_latency, rc); 1045 } 1046 1047 if (rc == -EOLDSNAPC) { 1048 struct ceph_aio_work *aio_work; 1049 BUG_ON(!aio_req->write); 1050 1051 aio_work = kmalloc(sizeof(*aio_work), GFP_NOFS); 1052 if (aio_work) { 1053 INIT_WORK(&aio_work->work, ceph_aio_retry_work); 1054 aio_work->req = req; 1055 queue_work(ceph_inode_to_client(inode)->inode_wq, 1056 &aio_work->work); 1057 return; 1058 } 1059 rc = -ENOMEM; 1060 } else if (!aio_req->write) { 1061 if (rc == -ENOENT) 1062 rc = 0; 1063 if (rc >= 0 && osd_data->bvec_pos.iter.bi_size > rc) { 1064 struct iov_iter i; 1065 int zlen = osd_data->bvec_pos.iter.bi_size - rc; 1066 1067 /* 1068 * If read is satisfied by single OSD request, 1069 * it can pass EOF. Otherwise read is within 1070 * i_size. 1071 */ 1072 if (aio_req->num_reqs == 1) { 1073 loff_t i_size = i_size_read(inode); 1074 loff_t endoff = aio_req->iocb->ki_pos + rc; 1075 if (endoff < i_size) 1076 zlen = min_t(size_t, zlen, 1077 i_size - endoff); 1078 aio_req->total_len = rc + zlen; 1079 } 1080 1081 iov_iter_bvec(&i, READ, osd_data->bvec_pos.bvecs, 1082 osd_data->num_bvecs, 1083 osd_data->bvec_pos.iter.bi_size); 1084 iov_iter_advance(&i, rc); 1085 iov_iter_zero(zlen, &i); 1086 } 1087 } 1088 1089 put_bvecs(osd_data->bvec_pos.bvecs, osd_data->num_bvecs, 1090 aio_req->should_dirty); 1091 ceph_osdc_put_request(req); 1092 1093 if (rc < 0) 1094 cmpxchg(&aio_req->error, 0, rc); 1095 1096 ceph_aio_complete(inode, aio_req); 1097 return; 1098 } 1099 1100 static void ceph_aio_retry_work(struct work_struct *work) 1101 { 1102 struct ceph_aio_work *aio_work = 1103 container_of(work, struct ceph_aio_work, work); 1104 struct ceph_osd_request *orig_req = aio_work->req; 1105 struct ceph_aio_request *aio_req = orig_req->r_priv; 1106 struct inode *inode = orig_req->r_inode; 1107 struct ceph_inode_info *ci = ceph_inode(inode); 1108 struct ceph_snap_context *snapc; 1109 struct ceph_osd_request *req; 1110 int ret; 1111 1112 spin_lock(&ci->i_ceph_lock); 1113 if (__ceph_have_pending_cap_snap(ci)) { 1114 struct ceph_cap_snap *capsnap = 1115 list_last_entry(&ci->i_cap_snaps, 1116 struct ceph_cap_snap, 1117 ci_item); 1118 snapc = ceph_get_snap_context(capsnap->context); 1119 } else { 1120 BUG_ON(!ci->i_head_snapc); 1121 snapc = ceph_get_snap_context(ci->i_head_snapc); 1122 } 1123 spin_unlock(&ci->i_ceph_lock); 1124 1125 req = ceph_osdc_alloc_request(orig_req->r_osdc, snapc, 1, 1126 false, GFP_NOFS); 1127 if (!req) { 1128 ret = -ENOMEM; 1129 req = orig_req; 1130 goto out; 1131 } 1132 1133 req->r_flags = /* CEPH_OSD_FLAG_ORDERSNAP | */ CEPH_OSD_FLAG_WRITE; 1134 ceph_oloc_copy(&req->r_base_oloc, &orig_req->r_base_oloc); 1135 ceph_oid_copy(&req->r_base_oid, &orig_req->r_base_oid); 1136 1137 req->r_ops[0] = orig_req->r_ops[0]; 1138 1139 req->r_mtime = aio_req->mtime; 1140 req->r_data_offset = req->r_ops[0].extent.offset; 1141 1142 ret = ceph_osdc_alloc_messages(req, GFP_NOFS); 1143 if (ret) { 1144 ceph_osdc_put_request(req); 1145 req = orig_req; 1146 goto out; 1147 } 1148 1149 ceph_osdc_put_request(orig_req); 1150 1151 req->r_callback = ceph_aio_complete_req; 1152 req->r_inode = inode; 1153 req->r_priv = aio_req; 1154 1155 ret = ceph_osdc_start_request(req->r_osdc, req, false); 1156 out: 1157 if (ret < 0) { 1158 req->r_result = ret; 1159 ceph_aio_complete_req(req); 1160 } 1161 1162 ceph_put_snap_context(snapc); 1163 kfree(aio_work); 1164 } 1165 1166 static ssize_t 1167 ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter, 1168 struct ceph_snap_context *snapc, 1169 struct ceph_cap_flush **pcf) 1170 { 1171 struct file *file = iocb->ki_filp; 1172 struct inode *inode = file_inode(file); 1173 struct ceph_inode_info *ci = ceph_inode(inode); 1174 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1175 struct ceph_client_metric *metric = &fsc->mdsc->metric; 1176 struct ceph_vino vino; 1177 struct ceph_osd_request *req; 1178 struct bio_vec *bvecs; 1179 struct ceph_aio_request *aio_req = NULL; 1180 int num_pages = 0; 1181 int flags; 1182 int ret = 0; 1183 struct timespec64 mtime = current_time(inode); 1184 size_t count = iov_iter_count(iter); 1185 loff_t pos = iocb->ki_pos; 1186 bool write = iov_iter_rw(iter) == WRITE; 1187 bool should_dirty = !write && iter_is_iovec(iter); 1188 1189 if (write && ceph_snap(file_inode(file)) != CEPH_NOSNAP) 1190 return -EROFS; 1191 1192 dout("sync_direct_%s on file %p %lld~%u snapc %p seq %lld\n", 1193 (write ? "write" : "read"), file, pos, (unsigned)count, 1194 snapc, snapc ? snapc->seq : 0); 1195 1196 if (write) { 1197 int ret2 = invalidate_inode_pages2_range(inode->i_mapping, 1198 pos >> PAGE_SHIFT, 1199 (pos + count - 1) >> PAGE_SHIFT); 1200 if (ret2 < 0) 1201 dout("invalidate_inode_pages2_range returned %d\n", ret2); 1202 1203 flags = /* CEPH_OSD_FLAG_ORDERSNAP | */ CEPH_OSD_FLAG_WRITE; 1204 } else { 1205 flags = CEPH_OSD_FLAG_READ; 1206 } 1207 1208 while (iov_iter_count(iter) > 0) { 1209 u64 size = iov_iter_count(iter); 1210 ssize_t len; 1211 1212 if (write) 1213 size = min_t(u64, size, fsc->mount_options->wsize); 1214 else 1215 size = min_t(u64, size, fsc->mount_options->rsize); 1216 1217 vino = ceph_vino(inode); 1218 req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, 1219 vino, pos, &size, 0, 1220 1, 1221 write ? CEPH_OSD_OP_WRITE : 1222 CEPH_OSD_OP_READ, 1223 flags, snapc, 1224 ci->i_truncate_seq, 1225 ci->i_truncate_size, 1226 false); 1227 if (IS_ERR(req)) { 1228 ret = PTR_ERR(req); 1229 break; 1230 } 1231 1232 len = iter_get_bvecs_alloc(iter, size, &bvecs, &num_pages); 1233 if (len < 0) { 1234 ceph_osdc_put_request(req); 1235 ret = len; 1236 break; 1237 } 1238 if (len != size) 1239 osd_req_op_extent_update(req, 0, len); 1240 1241 /* 1242 * To simplify error handling, allow AIO when IO within i_size 1243 * or IO can be satisfied by single OSD request. 1244 */ 1245 if (pos == iocb->ki_pos && !is_sync_kiocb(iocb) && 1246 (len == count || pos + count <= i_size_read(inode))) { 1247 aio_req = kzalloc(sizeof(*aio_req), GFP_KERNEL); 1248 if (aio_req) { 1249 aio_req->iocb = iocb; 1250 aio_req->write = write; 1251 aio_req->should_dirty = should_dirty; 1252 INIT_LIST_HEAD(&aio_req->osd_reqs); 1253 if (write) { 1254 aio_req->mtime = mtime; 1255 swap(aio_req->prealloc_cf, *pcf); 1256 } 1257 } 1258 /* ignore error */ 1259 } 1260 1261 if (write) { 1262 /* 1263 * throw out any page cache pages in this range. this 1264 * may block. 1265 */ 1266 truncate_inode_pages_range(inode->i_mapping, pos, 1267 PAGE_ALIGN(pos + len) - 1); 1268 1269 req->r_mtime = mtime; 1270 } 1271 1272 osd_req_op_extent_osd_data_bvecs(req, 0, bvecs, num_pages, len); 1273 1274 if (aio_req) { 1275 aio_req->total_len += len; 1276 aio_req->num_reqs++; 1277 atomic_inc(&aio_req->pending_reqs); 1278 1279 req->r_callback = ceph_aio_complete_req; 1280 req->r_inode = inode; 1281 req->r_priv = aio_req; 1282 list_add_tail(&req->r_private_item, &aio_req->osd_reqs); 1283 1284 pos += len; 1285 continue; 1286 } 1287 1288 ret = ceph_osdc_start_request(req->r_osdc, req, false); 1289 if (!ret) 1290 ret = ceph_osdc_wait_request(&fsc->client->osdc, req); 1291 1292 if (write) 1293 ceph_update_write_latency(metric, req->r_start_latency, 1294 req->r_end_latency, ret); 1295 else 1296 ceph_update_read_latency(metric, req->r_start_latency, 1297 req->r_end_latency, ret); 1298 1299 size = i_size_read(inode); 1300 if (!write) { 1301 if (ret == -ENOENT) 1302 ret = 0; 1303 if (ret >= 0 && ret < len && pos + ret < size) { 1304 struct iov_iter i; 1305 int zlen = min_t(size_t, len - ret, 1306 size - pos - ret); 1307 1308 iov_iter_bvec(&i, READ, bvecs, num_pages, len); 1309 iov_iter_advance(&i, ret); 1310 iov_iter_zero(zlen, &i); 1311 ret += zlen; 1312 } 1313 if (ret >= 0) 1314 len = ret; 1315 } 1316 1317 put_bvecs(bvecs, num_pages, should_dirty); 1318 ceph_osdc_put_request(req); 1319 if (ret < 0) 1320 break; 1321 1322 pos += len; 1323 if (!write && pos >= size) 1324 break; 1325 1326 if (write && pos > size) { 1327 if (ceph_inode_set_size(inode, pos)) 1328 ceph_check_caps(ceph_inode(inode), 1329 CHECK_CAPS_AUTHONLY, 1330 NULL); 1331 } 1332 } 1333 1334 if (aio_req) { 1335 LIST_HEAD(osd_reqs); 1336 1337 if (aio_req->num_reqs == 0) { 1338 kfree(aio_req); 1339 return ret; 1340 } 1341 1342 ceph_get_cap_refs(ci, write ? CEPH_CAP_FILE_WR : 1343 CEPH_CAP_FILE_RD); 1344 1345 list_splice(&aio_req->osd_reqs, &osd_reqs); 1346 inode_dio_begin(inode); 1347 while (!list_empty(&osd_reqs)) { 1348 req = list_first_entry(&osd_reqs, 1349 struct ceph_osd_request, 1350 r_private_item); 1351 list_del_init(&req->r_private_item); 1352 if (ret >= 0) 1353 ret = ceph_osdc_start_request(req->r_osdc, 1354 req, false); 1355 if (ret < 0) { 1356 req->r_result = ret; 1357 ceph_aio_complete_req(req); 1358 } 1359 } 1360 return -EIOCBQUEUED; 1361 } 1362 1363 if (ret != -EOLDSNAPC && pos > iocb->ki_pos) { 1364 ret = pos - iocb->ki_pos; 1365 iocb->ki_pos = pos; 1366 } 1367 return ret; 1368 } 1369 1370 /* 1371 * Synchronous write, straight from __user pointer or user pages. 1372 * 1373 * If write spans object boundary, just do multiple writes. (For a 1374 * correct atomic write, we should e.g. take write locks on all 1375 * objects, rollback on failure, etc.) 1376 */ 1377 static ssize_t 1378 ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos, 1379 struct ceph_snap_context *snapc) 1380 { 1381 struct file *file = iocb->ki_filp; 1382 struct inode *inode = file_inode(file); 1383 struct ceph_inode_info *ci = ceph_inode(inode); 1384 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1385 struct ceph_vino vino; 1386 struct ceph_osd_request *req; 1387 struct page **pages; 1388 u64 len; 1389 int num_pages; 1390 int written = 0; 1391 int flags; 1392 int ret; 1393 bool check_caps = false; 1394 struct timespec64 mtime = current_time(inode); 1395 size_t count = iov_iter_count(from); 1396 1397 if (ceph_snap(file_inode(file)) != CEPH_NOSNAP) 1398 return -EROFS; 1399 1400 dout("sync_write on file %p %lld~%u snapc %p seq %lld\n", 1401 file, pos, (unsigned)count, snapc, snapc->seq); 1402 1403 ret = filemap_write_and_wait_range(inode->i_mapping, 1404 pos, pos + count - 1); 1405 if (ret < 0) 1406 return ret; 1407 1408 ret = invalidate_inode_pages2_range(inode->i_mapping, 1409 pos >> PAGE_SHIFT, 1410 (pos + count - 1) >> PAGE_SHIFT); 1411 if (ret < 0) 1412 dout("invalidate_inode_pages2_range returned %d\n", ret); 1413 1414 flags = /* CEPH_OSD_FLAG_ORDERSNAP | */ CEPH_OSD_FLAG_WRITE; 1415 1416 while ((len = iov_iter_count(from)) > 0) { 1417 size_t left; 1418 int n; 1419 1420 vino = ceph_vino(inode); 1421 req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, 1422 vino, pos, &len, 0, 1, 1423 CEPH_OSD_OP_WRITE, flags, snapc, 1424 ci->i_truncate_seq, 1425 ci->i_truncate_size, 1426 false); 1427 if (IS_ERR(req)) { 1428 ret = PTR_ERR(req); 1429 break; 1430 } 1431 1432 /* 1433 * write from beginning of first page, 1434 * regardless of io alignment 1435 */ 1436 num_pages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT; 1437 1438 pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 1439 if (IS_ERR(pages)) { 1440 ret = PTR_ERR(pages); 1441 goto out; 1442 } 1443 1444 left = len; 1445 for (n = 0; n < num_pages; n++) { 1446 size_t plen = min_t(size_t, left, PAGE_SIZE); 1447 ret = copy_page_from_iter(pages[n], 0, plen, from); 1448 if (ret != plen) { 1449 ret = -EFAULT; 1450 break; 1451 } 1452 left -= ret; 1453 } 1454 1455 if (ret < 0) { 1456 ceph_release_page_vector(pages, num_pages); 1457 goto out; 1458 } 1459 1460 req->r_inode = inode; 1461 1462 osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0, 1463 false, true); 1464 1465 req->r_mtime = mtime; 1466 ret = ceph_osdc_start_request(&fsc->client->osdc, req, false); 1467 if (!ret) 1468 ret = ceph_osdc_wait_request(&fsc->client->osdc, req); 1469 1470 ceph_update_write_latency(&fsc->mdsc->metric, req->r_start_latency, 1471 req->r_end_latency, ret); 1472 out: 1473 ceph_osdc_put_request(req); 1474 if (ret != 0) { 1475 ceph_set_error_write(ci); 1476 break; 1477 } 1478 1479 ceph_clear_error_write(ci); 1480 pos += len; 1481 written += len; 1482 if (pos > i_size_read(inode)) { 1483 check_caps = ceph_inode_set_size(inode, pos); 1484 if (check_caps) 1485 ceph_check_caps(ceph_inode(inode), 1486 CHECK_CAPS_AUTHONLY, 1487 NULL); 1488 } 1489 1490 } 1491 1492 if (ret != -EOLDSNAPC && written > 0) { 1493 ret = written; 1494 iocb->ki_pos = pos; 1495 } 1496 return ret; 1497 } 1498 1499 /* 1500 * Wrap generic_file_aio_read with checks for cap bits on the inode. 1501 * Atomically grab references, so that those bits are not released 1502 * back to the MDS mid-read. 1503 * 1504 * Hmm, the sync read case isn't actually async... should it be? 1505 */ 1506 static ssize_t ceph_read_iter(struct kiocb *iocb, struct iov_iter *to) 1507 { 1508 struct file *filp = iocb->ki_filp; 1509 struct ceph_file_info *fi = filp->private_data; 1510 size_t len = iov_iter_count(to); 1511 struct inode *inode = file_inode(filp); 1512 struct ceph_inode_info *ci = ceph_inode(inode); 1513 struct page *pinned_page = NULL; 1514 bool direct_lock = iocb->ki_flags & IOCB_DIRECT; 1515 ssize_t ret; 1516 int want, got = 0; 1517 int retry_op = 0, read = 0; 1518 1519 again: 1520 dout("aio_read %p %llx.%llx %llu~%u trying to get caps on %p\n", 1521 inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len, inode); 1522 1523 if (direct_lock) 1524 ceph_start_io_direct(inode); 1525 else 1526 ceph_start_io_read(inode); 1527 1528 if (fi->fmode & CEPH_FILE_MODE_LAZY) 1529 want = CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO; 1530 else 1531 want = CEPH_CAP_FILE_CACHE; 1532 ret = ceph_get_caps(filp, CEPH_CAP_FILE_RD, want, -1, 1533 &got, &pinned_page); 1534 if (ret < 0) { 1535 if (iocb->ki_flags & IOCB_DIRECT) 1536 ceph_end_io_direct(inode); 1537 else 1538 ceph_end_io_read(inode); 1539 return ret; 1540 } 1541 1542 if ((got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0 || 1543 (iocb->ki_flags & IOCB_DIRECT) || 1544 (fi->flags & CEPH_F_SYNC)) { 1545 1546 dout("aio_sync_read %p %llx.%llx %llu~%u got cap refs on %s\n", 1547 inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len, 1548 ceph_cap_string(got)); 1549 1550 if (ci->i_inline_version == CEPH_INLINE_NONE) { 1551 if (!retry_op && (iocb->ki_flags & IOCB_DIRECT)) { 1552 ret = ceph_direct_read_write(iocb, to, 1553 NULL, NULL); 1554 if (ret >= 0 && ret < len) 1555 retry_op = CHECK_EOF; 1556 } else { 1557 ret = ceph_sync_read(iocb, to, &retry_op); 1558 } 1559 } else { 1560 retry_op = READ_INLINE; 1561 } 1562 } else { 1563 CEPH_DEFINE_RW_CONTEXT(rw_ctx, got); 1564 dout("aio_read %p %llx.%llx %llu~%u got cap refs on %s\n", 1565 inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len, 1566 ceph_cap_string(got)); 1567 ceph_add_rw_context(fi, &rw_ctx); 1568 ret = generic_file_read_iter(iocb, to); 1569 ceph_del_rw_context(fi, &rw_ctx); 1570 } 1571 1572 dout("aio_read %p %llx.%llx dropping cap refs on %s = %d\n", 1573 inode, ceph_vinop(inode), ceph_cap_string(got), (int)ret); 1574 if (pinned_page) { 1575 put_page(pinned_page); 1576 pinned_page = NULL; 1577 } 1578 ceph_put_cap_refs(ci, got); 1579 1580 if (direct_lock) 1581 ceph_end_io_direct(inode); 1582 else 1583 ceph_end_io_read(inode); 1584 1585 if (retry_op > HAVE_RETRIED && ret >= 0) { 1586 int statret; 1587 struct page *page = NULL; 1588 loff_t i_size; 1589 if (retry_op == READ_INLINE) { 1590 page = __page_cache_alloc(GFP_KERNEL); 1591 if (!page) 1592 return -ENOMEM; 1593 } 1594 1595 statret = __ceph_do_getattr(inode, page, 1596 CEPH_STAT_CAP_INLINE_DATA, !!page); 1597 if (statret < 0) { 1598 if (page) 1599 __free_page(page); 1600 if (statret == -ENODATA) { 1601 BUG_ON(retry_op != READ_INLINE); 1602 goto again; 1603 } 1604 return statret; 1605 } 1606 1607 i_size = i_size_read(inode); 1608 if (retry_op == READ_INLINE) { 1609 BUG_ON(ret > 0 || read > 0); 1610 if (iocb->ki_pos < i_size && 1611 iocb->ki_pos < PAGE_SIZE) { 1612 loff_t end = min_t(loff_t, i_size, 1613 iocb->ki_pos + len); 1614 end = min_t(loff_t, end, PAGE_SIZE); 1615 if (statret < end) 1616 zero_user_segment(page, statret, end); 1617 ret = copy_page_to_iter(page, 1618 iocb->ki_pos & ~PAGE_MASK, 1619 end - iocb->ki_pos, to); 1620 iocb->ki_pos += ret; 1621 read += ret; 1622 } 1623 if (iocb->ki_pos < i_size && read < len) { 1624 size_t zlen = min_t(size_t, len - read, 1625 i_size - iocb->ki_pos); 1626 ret = iov_iter_zero(zlen, to); 1627 iocb->ki_pos += ret; 1628 read += ret; 1629 } 1630 __free_pages(page, 0); 1631 return read; 1632 } 1633 1634 /* hit EOF or hole? */ 1635 if (retry_op == CHECK_EOF && iocb->ki_pos < i_size && 1636 ret < len) { 1637 dout("sync_read hit hole, ppos %lld < size %lld" 1638 ", reading more\n", iocb->ki_pos, i_size); 1639 1640 read += ret; 1641 len -= ret; 1642 retry_op = HAVE_RETRIED; 1643 goto again; 1644 } 1645 } 1646 1647 if (ret >= 0) 1648 ret += read; 1649 1650 return ret; 1651 } 1652 1653 /* 1654 * Take cap references to avoid releasing caps to MDS mid-write. 1655 * 1656 * If we are synchronous, and write with an old snap context, the OSD 1657 * may return EOLDSNAPC. In that case, retry the write.. _after_ 1658 * dropping our cap refs and allowing the pending snap to logically 1659 * complete _before_ this write occurs. 1660 * 1661 * If we are near ENOSPC, write synchronously. 1662 */ 1663 static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) 1664 { 1665 struct file *file = iocb->ki_filp; 1666 struct ceph_file_info *fi = file->private_data; 1667 struct inode *inode = file_inode(file); 1668 struct ceph_inode_info *ci = ceph_inode(inode); 1669 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1670 struct ceph_osd_client *osdc = &fsc->client->osdc; 1671 struct ceph_cap_flush *prealloc_cf; 1672 ssize_t count, written = 0; 1673 int err, want, got; 1674 bool direct_lock = false; 1675 u32 map_flags; 1676 u64 pool_flags; 1677 loff_t pos; 1678 loff_t limit = max(i_size_read(inode), fsc->max_file_size); 1679 1680 if (ceph_snap(inode) != CEPH_NOSNAP) 1681 return -EROFS; 1682 1683 prealloc_cf = ceph_alloc_cap_flush(); 1684 if (!prealloc_cf) 1685 return -ENOMEM; 1686 1687 if ((iocb->ki_flags & (IOCB_DIRECT | IOCB_APPEND)) == IOCB_DIRECT) 1688 direct_lock = true; 1689 1690 retry_snap: 1691 if (direct_lock) 1692 ceph_start_io_direct(inode); 1693 else 1694 ceph_start_io_write(inode); 1695 1696 /* We can write back this queue in page reclaim */ 1697 current->backing_dev_info = inode_to_bdi(inode); 1698 1699 if (iocb->ki_flags & IOCB_APPEND) { 1700 err = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE, false); 1701 if (err < 0) 1702 goto out; 1703 } 1704 1705 err = generic_write_checks(iocb, from); 1706 if (err <= 0) 1707 goto out; 1708 1709 pos = iocb->ki_pos; 1710 if (unlikely(pos >= limit)) { 1711 err = -EFBIG; 1712 goto out; 1713 } else { 1714 iov_iter_truncate(from, limit - pos); 1715 } 1716 1717 count = iov_iter_count(from); 1718 if (ceph_quota_is_max_bytes_exceeded(inode, pos + count)) { 1719 err = -EDQUOT; 1720 goto out; 1721 } 1722 1723 err = file_remove_privs(file); 1724 if (err) 1725 goto out; 1726 1727 err = file_update_time(file); 1728 if (err) 1729 goto out; 1730 1731 inode_inc_iversion_raw(inode); 1732 1733 if (ci->i_inline_version != CEPH_INLINE_NONE) { 1734 err = ceph_uninline_data(file, NULL); 1735 if (err < 0) 1736 goto out; 1737 } 1738 1739 down_read(&osdc->lock); 1740 map_flags = osdc->osdmap->flags; 1741 pool_flags = ceph_pg_pool_flags(osdc->osdmap, ci->i_layout.pool_id); 1742 up_read(&osdc->lock); 1743 if ((map_flags & CEPH_OSDMAP_FULL) || 1744 (pool_flags & CEPH_POOL_FLAG_FULL)) { 1745 err = -ENOSPC; 1746 goto out; 1747 } 1748 1749 dout("aio_write %p %llx.%llx %llu~%zd getting caps. i_size %llu\n", 1750 inode, ceph_vinop(inode), pos, count, i_size_read(inode)); 1751 if (fi->fmode & CEPH_FILE_MODE_LAZY) 1752 want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO; 1753 else 1754 want = CEPH_CAP_FILE_BUFFER; 1755 got = 0; 1756 err = ceph_get_caps(file, CEPH_CAP_FILE_WR, want, pos + count, 1757 &got, NULL); 1758 if (err < 0) 1759 goto out; 1760 1761 dout("aio_write %p %llx.%llx %llu~%zd got cap refs on %s\n", 1762 inode, ceph_vinop(inode), pos, count, ceph_cap_string(got)); 1763 1764 if ((got & (CEPH_CAP_FILE_BUFFER|CEPH_CAP_FILE_LAZYIO)) == 0 || 1765 (iocb->ki_flags & IOCB_DIRECT) || (fi->flags & CEPH_F_SYNC) || 1766 (ci->i_ceph_flags & CEPH_I_ERROR_WRITE)) { 1767 struct ceph_snap_context *snapc; 1768 struct iov_iter data; 1769 1770 spin_lock(&ci->i_ceph_lock); 1771 if (__ceph_have_pending_cap_snap(ci)) { 1772 struct ceph_cap_snap *capsnap = 1773 list_last_entry(&ci->i_cap_snaps, 1774 struct ceph_cap_snap, 1775 ci_item); 1776 snapc = ceph_get_snap_context(capsnap->context); 1777 } else { 1778 BUG_ON(!ci->i_head_snapc); 1779 snapc = ceph_get_snap_context(ci->i_head_snapc); 1780 } 1781 spin_unlock(&ci->i_ceph_lock); 1782 1783 /* we might need to revert back to that point */ 1784 data = *from; 1785 if (iocb->ki_flags & IOCB_DIRECT) 1786 written = ceph_direct_read_write(iocb, &data, snapc, 1787 &prealloc_cf); 1788 else 1789 written = ceph_sync_write(iocb, &data, pos, snapc); 1790 if (direct_lock) 1791 ceph_end_io_direct(inode); 1792 else 1793 ceph_end_io_write(inode); 1794 if (written > 0) 1795 iov_iter_advance(from, written); 1796 ceph_put_snap_context(snapc); 1797 } else { 1798 /* 1799 * No need to acquire the i_truncate_mutex. Because 1800 * the MDS revokes Fwb caps before sending truncate 1801 * message to us. We can't get Fwb cap while there 1802 * are pending vmtruncate. So write and vmtruncate 1803 * can not run at the same time 1804 */ 1805 written = generic_perform_write(file, from, pos); 1806 if (likely(written >= 0)) 1807 iocb->ki_pos = pos + written; 1808 ceph_end_io_write(inode); 1809 } 1810 1811 if (written >= 0) { 1812 int dirty; 1813 1814 spin_lock(&ci->i_ceph_lock); 1815 ci->i_inline_version = CEPH_INLINE_NONE; 1816 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, 1817 &prealloc_cf); 1818 spin_unlock(&ci->i_ceph_lock); 1819 if (dirty) 1820 __mark_inode_dirty(inode, dirty); 1821 if (ceph_quota_is_max_bytes_approaching(inode, iocb->ki_pos)) 1822 ceph_check_caps(ci, 0, NULL); 1823 } 1824 1825 dout("aio_write %p %llx.%llx %llu~%u dropping cap refs on %s\n", 1826 inode, ceph_vinop(inode), pos, (unsigned)count, 1827 ceph_cap_string(got)); 1828 ceph_put_cap_refs(ci, got); 1829 1830 if (written == -EOLDSNAPC) { 1831 dout("aio_write %p %llx.%llx %llu~%u" "got EOLDSNAPC, retrying\n", 1832 inode, ceph_vinop(inode), pos, (unsigned)count); 1833 goto retry_snap; 1834 } 1835 1836 if (written >= 0) { 1837 if ((map_flags & CEPH_OSDMAP_NEARFULL) || 1838 (pool_flags & CEPH_POOL_FLAG_NEARFULL)) 1839 iocb->ki_flags |= IOCB_DSYNC; 1840 written = generic_write_sync(iocb, written); 1841 } 1842 1843 goto out_unlocked; 1844 out: 1845 if (direct_lock) 1846 ceph_end_io_direct(inode); 1847 else 1848 ceph_end_io_write(inode); 1849 out_unlocked: 1850 ceph_free_cap_flush(prealloc_cf); 1851 current->backing_dev_info = NULL; 1852 return written ? written : err; 1853 } 1854 1855 /* 1856 * llseek. be sure to verify file size on SEEK_END. 1857 */ 1858 static loff_t ceph_llseek(struct file *file, loff_t offset, int whence) 1859 { 1860 struct inode *inode = file->f_mapping->host; 1861 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1862 loff_t i_size; 1863 loff_t ret; 1864 1865 inode_lock(inode); 1866 1867 if (whence == SEEK_END || whence == SEEK_DATA || whence == SEEK_HOLE) { 1868 ret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE, false); 1869 if (ret < 0) 1870 goto out; 1871 } 1872 1873 i_size = i_size_read(inode); 1874 switch (whence) { 1875 case SEEK_END: 1876 offset += i_size; 1877 break; 1878 case SEEK_CUR: 1879 /* 1880 * Here we special-case the lseek(fd, 0, SEEK_CUR) 1881 * position-querying operation. Avoid rewriting the "same" 1882 * f_pos value back to the file because a concurrent read(), 1883 * write() or lseek() might have altered it 1884 */ 1885 if (offset == 0) { 1886 ret = file->f_pos; 1887 goto out; 1888 } 1889 offset += file->f_pos; 1890 break; 1891 case SEEK_DATA: 1892 if (offset < 0 || offset >= i_size) { 1893 ret = -ENXIO; 1894 goto out; 1895 } 1896 break; 1897 case SEEK_HOLE: 1898 if (offset < 0 || offset >= i_size) { 1899 ret = -ENXIO; 1900 goto out; 1901 } 1902 offset = i_size; 1903 break; 1904 } 1905 1906 ret = vfs_setpos(file, offset, max(i_size, fsc->max_file_size)); 1907 1908 out: 1909 inode_unlock(inode); 1910 return ret; 1911 } 1912 1913 static inline void ceph_zero_partial_page( 1914 struct inode *inode, loff_t offset, unsigned size) 1915 { 1916 struct page *page; 1917 pgoff_t index = offset >> PAGE_SHIFT; 1918 1919 page = find_lock_page(inode->i_mapping, index); 1920 if (page) { 1921 wait_on_page_writeback(page); 1922 zero_user(page, offset & (PAGE_SIZE - 1), size); 1923 unlock_page(page); 1924 put_page(page); 1925 } 1926 } 1927 1928 static void ceph_zero_pagecache_range(struct inode *inode, loff_t offset, 1929 loff_t length) 1930 { 1931 loff_t nearly = round_up(offset, PAGE_SIZE); 1932 if (offset < nearly) { 1933 loff_t size = nearly - offset; 1934 if (length < size) 1935 size = length; 1936 ceph_zero_partial_page(inode, offset, size); 1937 offset += size; 1938 length -= size; 1939 } 1940 if (length >= PAGE_SIZE) { 1941 loff_t size = round_down(length, PAGE_SIZE); 1942 truncate_pagecache_range(inode, offset, offset + size - 1); 1943 offset += size; 1944 length -= size; 1945 } 1946 if (length) 1947 ceph_zero_partial_page(inode, offset, length); 1948 } 1949 1950 static int ceph_zero_partial_object(struct inode *inode, 1951 loff_t offset, loff_t *length) 1952 { 1953 struct ceph_inode_info *ci = ceph_inode(inode); 1954 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1955 struct ceph_osd_request *req; 1956 int ret = 0; 1957 loff_t zero = 0; 1958 int op; 1959 1960 if (!length) { 1961 op = offset ? CEPH_OSD_OP_DELETE : CEPH_OSD_OP_TRUNCATE; 1962 length = &zero; 1963 } else { 1964 op = CEPH_OSD_OP_ZERO; 1965 } 1966 1967 req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, 1968 ceph_vino(inode), 1969 offset, length, 1970 0, 1, op, 1971 CEPH_OSD_FLAG_WRITE, 1972 NULL, 0, 0, false); 1973 if (IS_ERR(req)) { 1974 ret = PTR_ERR(req); 1975 goto out; 1976 } 1977 1978 req->r_mtime = inode->i_mtime; 1979 ret = ceph_osdc_start_request(&fsc->client->osdc, req, false); 1980 if (!ret) { 1981 ret = ceph_osdc_wait_request(&fsc->client->osdc, req); 1982 if (ret == -ENOENT) 1983 ret = 0; 1984 } 1985 ceph_osdc_put_request(req); 1986 1987 out: 1988 return ret; 1989 } 1990 1991 static int ceph_zero_objects(struct inode *inode, loff_t offset, loff_t length) 1992 { 1993 int ret = 0; 1994 struct ceph_inode_info *ci = ceph_inode(inode); 1995 s32 stripe_unit = ci->i_layout.stripe_unit; 1996 s32 stripe_count = ci->i_layout.stripe_count; 1997 s32 object_size = ci->i_layout.object_size; 1998 u64 object_set_size = object_size * stripe_count; 1999 u64 nearly, t; 2000 2001 /* round offset up to next period boundary */ 2002 nearly = offset + object_set_size - 1; 2003 t = nearly; 2004 nearly -= do_div(t, object_set_size); 2005 2006 while (length && offset < nearly) { 2007 loff_t size = length; 2008 ret = ceph_zero_partial_object(inode, offset, &size); 2009 if (ret < 0) 2010 return ret; 2011 offset += size; 2012 length -= size; 2013 } 2014 while (length >= object_set_size) { 2015 int i; 2016 loff_t pos = offset; 2017 for (i = 0; i < stripe_count; ++i) { 2018 ret = ceph_zero_partial_object(inode, pos, NULL); 2019 if (ret < 0) 2020 return ret; 2021 pos += stripe_unit; 2022 } 2023 offset += object_set_size; 2024 length -= object_set_size; 2025 } 2026 while (length) { 2027 loff_t size = length; 2028 ret = ceph_zero_partial_object(inode, offset, &size); 2029 if (ret < 0) 2030 return ret; 2031 offset += size; 2032 length -= size; 2033 } 2034 return ret; 2035 } 2036 2037 static long ceph_fallocate(struct file *file, int mode, 2038 loff_t offset, loff_t length) 2039 { 2040 struct ceph_file_info *fi = file->private_data; 2041 struct inode *inode = file_inode(file); 2042 struct ceph_inode_info *ci = ceph_inode(inode); 2043 struct ceph_cap_flush *prealloc_cf; 2044 int want, got = 0; 2045 int dirty; 2046 int ret = 0; 2047 loff_t endoff = 0; 2048 loff_t size; 2049 2050 if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) 2051 return -EOPNOTSUPP; 2052 2053 if (!S_ISREG(inode->i_mode)) 2054 return -EOPNOTSUPP; 2055 2056 prealloc_cf = ceph_alloc_cap_flush(); 2057 if (!prealloc_cf) 2058 return -ENOMEM; 2059 2060 inode_lock(inode); 2061 2062 if (ceph_snap(inode) != CEPH_NOSNAP) { 2063 ret = -EROFS; 2064 goto unlock; 2065 } 2066 2067 if (ci->i_inline_version != CEPH_INLINE_NONE) { 2068 ret = ceph_uninline_data(file, NULL); 2069 if (ret < 0) 2070 goto unlock; 2071 } 2072 2073 size = i_size_read(inode); 2074 2075 /* Are we punching a hole beyond EOF? */ 2076 if (offset >= size) 2077 goto unlock; 2078 if ((offset + length) > size) 2079 length = size - offset; 2080 2081 if (fi->fmode & CEPH_FILE_MODE_LAZY) 2082 want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO; 2083 else 2084 want = CEPH_CAP_FILE_BUFFER; 2085 2086 ret = ceph_get_caps(file, CEPH_CAP_FILE_WR, want, endoff, &got, NULL); 2087 if (ret < 0) 2088 goto unlock; 2089 2090 ceph_zero_pagecache_range(inode, offset, length); 2091 ret = ceph_zero_objects(inode, offset, length); 2092 2093 if (!ret) { 2094 spin_lock(&ci->i_ceph_lock); 2095 ci->i_inline_version = CEPH_INLINE_NONE; 2096 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, 2097 &prealloc_cf); 2098 spin_unlock(&ci->i_ceph_lock); 2099 if (dirty) 2100 __mark_inode_dirty(inode, dirty); 2101 } 2102 2103 ceph_put_cap_refs(ci, got); 2104 unlock: 2105 inode_unlock(inode); 2106 ceph_free_cap_flush(prealloc_cf); 2107 return ret; 2108 } 2109 2110 /* 2111 * This function tries to get FILE_WR capabilities for dst_ci and FILE_RD for 2112 * src_ci. Two attempts are made to obtain both caps, and an error is return if 2113 * this fails; zero is returned on success. 2114 */ 2115 static int get_rd_wr_caps(struct file *src_filp, int *src_got, 2116 struct file *dst_filp, 2117 loff_t dst_endoff, int *dst_got) 2118 { 2119 int ret = 0; 2120 bool retrying = false; 2121 2122 retry_caps: 2123 ret = ceph_get_caps(dst_filp, CEPH_CAP_FILE_WR, CEPH_CAP_FILE_BUFFER, 2124 dst_endoff, dst_got, NULL); 2125 if (ret < 0) 2126 return ret; 2127 2128 /* 2129 * Since we're already holding the FILE_WR capability for the dst file, 2130 * we would risk a deadlock by using ceph_get_caps. Thus, we'll do some 2131 * retry dance instead to try to get both capabilities. 2132 */ 2133 ret = ceph_try_get_caps(file_inode(src_filp), 2134 CEPH_CAP_FILE_RD, CEPH_CAP_FILE_SHARED, 2135 false, src_got); 2136 if (ret <= 0) { 2137 /* Start by dropping dst_ci caps and getting src_ci caps */ 2138 ceph_put_cap_refs(ceph_inode(file_inode(dst_filp)), *dst_got); 2139 if (retrying) { 2140 if (!ret) 2141 /* ceph_try_get_caps masks EAGAIN */ 2142 ret = -EAGAIN; 2143 return ret; 2144 } 2145 ret = ceph_get_caps(src_filp, CEPH_CAP_FILE_RD, 2146 CEPH_CAP_FILE_SHARED, -1, src_got, NULL); 2147 if (ret < 0) 2148 return ret; 2149 /*... drop src_ci caps too, and retry */ 2150 ceph_put_cap_refs(ceph_inode(file_inode(src_filp)), *src_got); 2151 retrying = true; 2152 goto retry_caps; 2153 } 2154 return ret; 2155 } 2156 2157 static void put_rd_wr_caps(struct ceph_inode_info *src_ci, int src_got, 2158 struct ceph_inode_info *dst_ci, int dst_got) 2159 { 2160 ceph_put_cap_refs(src_ci, src_got); 2161 ceph_put_cap_refs(dst_ci, dst_got); 2162 } 2163 2164 /* 2165 * This function does several size-related checks, returning an error if: 2166 * - source file is smaller than off+len 2167 * - destination file size is not OK (inode_newsize_ok()) 2168 * - max bytes quotas is exceeded 2169 */ 2170 static int is_file_size_ok(struct inode *src_inode, struct inode *dst_inode, 2171 loff_t src_off, loff_t dst_off, size_t len) 2172 { 2173 loff_t size, endoff; 2174 2175 size = i_size_read(src_inode); 2176 /* 2177 * Don't copy beyond source file EOF. Instead of simply setting length 2178 * to (size - src_off), just drop to VFS default implementation, as the 2179 * local i_size may be stale due to other clients writing to the source 2180 * inode. 2181 */ 2182 if (src_off + len > size) { 2183 dout("Copy beyond EOF (%llu + %zu > %llu)\n", 2184 src_off, len, size); 2185 return -EOPNOTSUPP; 2186 } 2187 size = i_size_read(dst_inode); 2188 2189 endoff = dst_off + len; 2190 if (inode_newsize_ok(dst_inode, endoff)) 2191 return -EOPNOTSUPP; 2192 2193 if (ceph_quota_is_max_bytes_exceeded(dst_inode, endoff)) 2194 return -EDQUOT; 2195 2196 return 0; 2197 } 2198 2199 static ssize_t ceph_do_objects_copy(struct ceph_inode_info *src_ci, u64 *src_off, 2200 struct ceph_inode_info *dst_ci, u64 *dst_off, 2201 struct ceph_fs_client *fsc, 2202 size_t len, unsigned int flags) 2203 { 2204 struct ceph_object_locator src_oloc, dst_oloc; 2205 struct ceph_object_id src_oid, dst_oid; 2206 size_t bytes = 0; 2207 u64 src_objnum, src_objoff, dst_objnum, dst_objoff; 2208 u32 src_objlen, dst_objlen; 2209 u32 object_size = src_ci->i_layout.object_size; 2210 int ret; 2211 2212 src_oloc.pool = src_ci->i_layout.pool_id; 2213 src_oloc.pool_ns = ceph_try_get_string(src_ci->i_layout.pool_ns); 2214 dst_oloc.pool = dst_ci->i_layout.pool_id; 2215 dst_oloc.pool_ns = ceph_try_get_string(dst_ci->i_layout.pool_ns); 2216 2217 while (len >= object_size) { 2218 ceph_calc_file_object_mapping(&src_ci->i_layout, *src_off, 2219 object_size, &src_objnum, 2220 &src_objoff, &src_objlen); 2221 ceph_calc_file_object_mapping(&dst_ci->i_layout, *dst_off, 2222 object_size, &dst_objnum, 2223 &dst_objoff, &dst_objlen); 2224 ceph_oid_init(&src_oid); 2225 ceph_oid_printf(&src_oid, "%llx.%08llx", 2226 src_ci->i_vino.ino, src_objnum); 2227 ceph_oid_init(&dst_oid); 2228 ceph_oid_printf(&dst_oid, "%llx.%08llx", 2229 dst_ci->i_vino.ino, dst_objnum); 2230 /* Do an object remote copy */ 2231 ret = ceph_osdc_copy_from(&fsc->client->osdc, 2232 src_ci->i_vino.snap, 0, 2233 &src_oid, &src_oloc, 2234 CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL | 2235 CEPH_OSD_OP_FLAG_FADVISE_NOCACHE, 2236 &dst_oid, &dst_oloc, 2237 CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL | 2238 CEPH_OSD_OP_FLAG_FADVISE_DONTNEED, 2239 dst_ci->i_truncate_seq, 2240 dst_ci->i_truncate_size, 2241 CEPH_OSD_COPY_FROM_FLAG_TRUNCATE_SEQ); 2242 if (ret) { 2243 if (ret == -EOPNOTSUPP) { 2244 fsc->have_copy_from2 = false; 2245 pr_notice("OSDs don't support copy-from2; disabling copy offload\n"); 2246 } 2247 dout("ceph_osdc_copy_from returned %d\n", ret); 2248 if (!bytes) 2249 bytes = ret; 2250 goto out; 2251 } 2252 len -= object_size; 2253 bytes += object_size; 2254 *src_off += object_size; 2255 *dst_off += object_size; 2256 } 2257 2258 out: 2259 ceph_oloc_destroy(&src_oloc); 2260 ceph_oloc_destroy(&dst_oloc); 2261 return bytes; 2262 } 2263 2264 static ssize_t __ceph_copy_file_range(struct file *src_file, loff_t src_off, 2265 struct file *dst_file, loff_t dst_off, 2266 size_t len, unsigned int flags) 2267 { 2268 struct inode *src_inode = file_inode(src_file); 2269 struct inode *dst_inode = file_inode(dst_file); 2270 struct ceph_inode_info *src_ci = ceph_inode(src_inode); 2271 struct ceph_inode_info *dst_ci = ceph_inode(dst_inode); 2272 struct ceph_cap_flush *prealloc_cf; 2273 struct ceph_fs_client *src_fsc = ceph_inode_to_client(src_inode); 2274 loff_t size; 2275 ssize_t ret = -EIO, bytes; 2276 u64 src_objnum, dst_objnum, src_objoff, dst_objoff; 2277 u32 src_objlen, dst_objlen; 2278 int src_got = 0, dst_got = 0, err, dirty; 2279 2280 if (src_inode->i_sb != dst_inode->i_sb) { 2281 struct ceph_fs_client *dst_fsc = ceph_inode_to_client(dst_inode); 2282 2283 if (ceph_fsid_compare(&src_fsc->client->fsid, 2284 &dst_fsc->client->fsid)) { 2285 dout("Copying files across clusters: src: %pU dst: %pU\n", 2286 &src_fsc->client->fsid, &dst_fsc->client->fsid); 2287 return -EXDEV; 2288 } 2289 } 2290 if (ceph_snap(dst_inode) != CEPH_NOSNAP) 2291 return -EROFS; 2292 2293 /* 2294 * Some of the checks below will return -EOPNOTSUPP, which will force a 2295 * fallback to the default VFS copy_file_range implementation. This is 2296 * desirable in several cases (for ex, the 'len' is smaller than the 2297 * size of the objects, or in cases where that would be more 2298 * efficient). 2299 */ 2300 2301 if (ceph_test_mount_opt(src_fsc, NOCOPYFROM)) 2302 return -EOPNOTSUPP; 2303 2304 if (!src_fsc->have_copy_from2) 2305 return -EOPNOTSUPP; 2306 2307 /* 2308 * Striped file layouts require that we copy partial objects, but the 2309 * OSD copy-from operation only supports full-object copies. Limit 2310 * this to non-striped file layouts for now. 2311 */ 2312 if ((src_ci->i_layout.stripe_unit != dst_ci->i_layout.stripe_unit) || 2313 (src_ci->i_layout.stripe_count != 1) || 2314 (dst_ci->i_layout.stripe_count != 1) || 2315 (src_ci->i_layout.object_size != dst_ci->i_layout.object_size)) { 2316 dout("Invalid src/dst files layout\n"); 2317 return -EOPNOTSUPP; 2318 } 2319 2320 if (len < src_ci->i_layout.object_size) 2321 return -EOPNOTSUPP; /* no remote copy will be done */ 2322 2323 prealloc_cf = ceph_alloc_cap_flush(); 2324 if (!prealloc_cf) 2325 return -ENOMEM; 2326 2327 /* Start by sync'ing the source and destination files */ 2328 ret = file_write_and_wait_range(src_file, src_off, (src_off + len)); 2329 if (ret < 0) { 2330 dout("failed to write src file (%zd)\n", ret); 2331 goto out; 2332 } 2333 ret = file_write_and_wait_range(dst_file, dst_off, (dst_off + len)); 2334 if (ret < 0) { 2335 dout("failed to write dst file (%zd)\n", ret); 2336 goto out; 2337 } 2338 2339 /* 2340 * We need FILE_WR caps for dst_ci and FILE_RD for src_ci as other 2341 * clients may have dirty data in their caches. And OSDs know nothing 2342 * about caps, so they can't safely do the remote object copies. 2343 */ 2344 err = get_rd_wr_caps(src_file, &src_got, 2345 dst_file, (dst_off + len), &dst_got); 2346 if (err < 0) { 2347 dout("get_rd_wr_caps returned %d\n", err); 2348 ret = -EOPNOTSUPP; 2349 goto out; 2350 } 2351 2352 ret = is_file_size_ok(src_inode, dst_inode, src_off, dst_off, len); 2353 if (ret < 0) 2354 goto out_caps; 2355 2356 /* Drop dst file cached pages */ 2357 ret = invalidate_inode_pages2_range(dst_inode->i_mapping, 2358 dst_off >> PAGE_SHIFT, 2359 (dst_off + len) >> PAGE_SHIFT); 2360 if (ret < 0) { 2361 dout("Failed to invalidate inode pages (%zd)\n", ret); 2362 ret = 0; /* XXX */ 2363 } 2364 ceph_calc_file_object_mapping(&src_ci->i_layout, src_off, 2365 src_ci->i_layout.object_size, 2366 &src_objnum, &src_objoff, &src_objlen); 2367 ceph_calc_file_object_mapping(&dst_ci->i_layout, dst_off, 2368 dst_ci->i_layout.object_size, 2369 &dst_objnum, &dst_objoff, &dst_objlen); 2370 /* object-level offsets need to the same */ 2371 if (src_objoff != dst_objoff) { 2372 ret = -EOPNOTSUPP; 2373 goto out_caps; 2374 } 2375 2376 /* 2377 * Do a manual copy if the object offset isn't object aligned. 2378 * 'src_objlen' contains the bytes left until the end of the object, 2379 * starting at the src_off 2380 */ 2381 if (src_objoff) { 2382 dout("Initial partial copy of %u bytes\n", src_objlen); 2383 2384 /* 2385 * we need to temporarily drop all caps as we'll be calling 2386 * {read,write}_iter, which will get caps again. 2387 */ 2388 put_rd_wr_caps(src_ci, src_got, dst_ci, dst_got); 2389 ret = do_splice_direct(src_file, &src_off, dst_file, 2390 &dst_off, src_objlen, flags); 2391 /* Abort on short copies or on error */ 2392 if (ret < src_objlen) { 2393 dout("Failed partial copy (%zd)\n", ret); 2394 goto out; 2395 } 2396 len -= ret; 2397 err = get_rd_wr_caps(src_file, &src_got, 2398 dst_file, (dst_off + len), &dst_got); 2399 if (err < 0) 2400 goto out; 2401 err = is_file_size_ok(src_inode, dst_inode, 2402 src_off, dst_off, len); 2403 if (err < 0) 2404 goto out_caps; 2405 } 2406 2407 size = i_size_read(dst_inode); 2408 bytes = ceph_do_objects_copy(src_ci, &src_off, dst_ci, &dst_off, 2409 src_fsc, len, flags); 2410 if (bytes <= 0) { 2411 if (!ret) 2412 ret = bytes; 2413 goto out_caps; 2414 } 2415 dout("Copied %zu bytes out of %zu\n", bytes, len); 2416 len -= bytes; 2417 ret += bytes; 2418 2419 file_update_time(dst_file); 2420 inode_inc_iversion_raw(dst_inode); 2421 2422 if (dst_off > size) { 2423 /* Let the MDS know about dst file size change */ 2424 if (ceph_inode_set_size(dst_inode, dst_off) || 2425 ceph_quota_is_max_bytes_approaching(dst_inode, dst_off)) 2426 ceph_check_caps(dst_ci, CHECK_CAPS_AUTHONLY, NULL); 2427 } 2428 /* Mark Fw dirty */ 2429 spin_lock(&dst_ci->i_ceph_lock); 2430 dst_ci->i_inline_version = CEPH_INLINE_NONE; 2431 dirty = __ceph_mark_dirty_caps(dst_ci, CEPH_CAP_FILE_WR, &prealloc_cf); 2432 spin_unlock(&dst_ci->i_ceph_lock); 2433 if (dirty) 2434 __mark_inode_dirty(dst_inode, dirty); 2435 2436 out_caps: 2437 put_rd_wr_caps(src_ci, src_got, dst_ci, dst_got); 2438 2439 /* 2440 * Do the final manual copy if we still have some bytes left, unless 2441 * there were errors in remote object copies (len >= object_size). 2442 */ 2443 if (len && (len < src_ci->i_layout.object_size)) { 2444 dout("Final partial copy of %zu bytes\n", len); 2445 bytes = do_splice_direct(src_file, &src_off, dst_file, 2446 &dst_off, len, flags); 2447 if (bytes > 0) 2448 ret += bytes; 2449 else 2450 dout("Failed partial copy (%zd)\n", bytes); 2451 } 2452 2453 out: 2454 ceph_free_cap_flush(prealloc_cf); 2455 2456 return ret; 2457 } 2458 2459 static ssize_t ceph_copy_file_range(struct file *src_file, loff_t src_off, 2460 struct file *dst_file, loff_t dst_off, 2461 size_t len, unsigned int flags) 2462 { 2463 ssize_t ret; 2464 2465 ret = __ceph_copy_file_range(src_file, src_off, dst_file, dst_off, 2466 len, flags); 2467 2468 if (ret == -EOPNOTSUPP || ret == -EXDEV) 2469 ret = generic_copy_file_range(src_file, src_off, dst_file, 2470 dst_off, len, flags); 2471 return ret; 2472 } 2473 2474 const struct file_operations ceph_file_fops = { 2475 .open = ceph_open, 2476 .release = ceph_release, 2477 .llseek = ceph_llseek, 2478 .read_iter = ceph_read_iter, 2479 .write_iter = ceph_write_iter, 2480 .mmap = ceph_mmap, 2481 .fsync = ceph_fsync, 2482 .lock = ceph_lock, 2483 .setlease = simple_nosetlease, 2484 .flock = ceph_flock, 2485 .splice_read = generic_file_splice_read, 2486 .splice_write = iter_file_splice_write, 2487 .unlocked_ioctl = ceph_ioctl, 2488 .compat_ioctl = compat_ptr_ioctl, 2489 .fallocate = ceph_fallocate, 2490 .copy_file_range = ceph_copy_file_range, 2491 }; 2492