1 #include <linux/ceph/ceph_debug.h> 2 3 #include <linux/module.h> 4 #include <linux/sched.h> 5 #include <linux/slab.h> 6 #include <linux/file.h> 7 #include <linux/mount.h> 8 #include <linux/namei.h> 9 #include <linux/writeback.h> 10 #include <linux/falloc.h> 11 12 #include "super.h" 13 #include "mds_client.h" 14 #include "cache.h" 15 16 /* 17 * Ceph file operations 18 * 19 * Implement basic open/close functionality, and implement 20 * read/write. 21 * 22 * We implement three modes of file I/O: 23 * - buffered uses the generic_file_aio_{read,write} helpers 24 * 25 * - synchronous is used when there is multi-client read/write 26 * sharing, avoids the page cache, and synchronously waits for an 27 * ack from the OSD. 28 * 29 * - direct io takes the variant of the sync path that references 30 * user pages directly. 31 * 32 * fsync() flushes and waits on dirty pages, but just queues metadata 33 * for writeback: since the MDS can recover size and mtime there is no 34 * need to wait for MDS acknowledgement. 35 */ 36 37 /* 38 * Calculate the length sum of direct io vectors that can 39 * be combined into one page vector. 40 */ 41 static size_t dio_get_pagev_size(const struct iov_iter *it) 42 { 43 const struct iovec *iov = it->iov; 44 const struct iovec *iovend = iov + it->nr_segs; 45 size_t size; 46 47 size = iov->iov_len - it->iov_offset; 48 /* 49 * An iov can be page vectored when both the current tail 50 * and the next base are page aligned. 51 */ 52 while (PAGE_ALIGNED((iov->iov_base + iov->iov_len)) && 53 (++iov < iovend && PAGE_ALIGNED((iov->iov_base)))) { 54 size += iov->iov_len; 55 } 56 dout("dio_get_pagevlen len = %zu\n", size); 57 return size; 58 } 59 60 /* 61 * Allocate a page vector based on (@it, @nbytes). 62 * The return value is the tuple describing a page vector, 63 * that is (@pages, @page_align, @num_pages). 64 */ 65 static struct page ** 66 dio_get_pages_alloc(const struct iov_iter *it, size_t nbytes, 67 size_t *page_align, int *num_pages) 68 { 69 struct iov_iter tmp_it = *it; 70 size_t align; 71 struct page **pages; 72 int ret = 0, idx, npages; 73 74 align = (unsigned long)(it->iov->iov_base + it->iov_offset) & 75 (PAGE_SIZE - 1); 76 npages = calc_pages_for(align, nbytes); 77 pages = kmalloc(sizeof(*pages) * npages, GFP_KERNEL); 78 if (!pages) { 79 pages = vmalloc(sizeof(*pages) * npages); 80 if (!pages) 81 return ERR_PTR(-ENOMEM); 82 } 83 84 for (idx = 0; idx < npages; ) { 85 size_t start; 86 ret = iov_iter_get_pages(&tmp_it, pages + idx, nbytes, 87 npages - idx, &start); 88 if (ret < 0) 89 goto fail; 90 91 iov_iter_advance(&tmp_it, ret); 92 nbytes -= ret; 93 idx += (ret + start + PAGE_SIZE - 1) / PAGE_SIZE; 94 } 95 96 BUG_ON(nbytes != 0); 97 *num_pages = npages; 98 *page_align = align; 99 dout("dio_get_pages_alloc: got %d pages align %zu\n", npages, align); 100 return pages; 101 fail: 102 ceph_put_page_vector(pages, idx, false); 103 return ERR_PTR(ret); 104 } 105 106 /* 107 * Prepare an open request. Preallocate ceph_cap to avoid an 108 * inopportune ENOMEM later. 109 */ 110 static struct ceph_mds_request * 111 prepare_open_request(struct super_block *sb, int flags, int create_mode) 112 { 113 struct ceph_fs_client *fsc = ceph_sb_to_client(sb); 114 struct ceph_mds_client *mdsc = fsc->mdsc; 115 struct ceph_mds_request *req; 116 int want_auth = USE_ANY_MDS; 117 int op = (flags & O_CREAT) ? CEPH_MDS_OP_CREATE : CEPH_MDS_OP_OPEN; 118 119 if (flags & (O_WRONLY|O_RDWR|O_CREAT|O_TRUNC)) 120 want_auth = USE_AUTH_MDS; 121 122 req = ceph_mdsc_create_request(mdsc, op, want_auth); 123 if (IS_ERR(req)) 124 goto out; 125 req->r_fmode = ceph_flags_to_mode(flags); 126 req->r_args.open.flags = cpu_to_le32(flags); 127 req->r_args.open.mode = cpu_to_le32(create_mode); 128 out: 129 return req; 130 } 131 132 /* 133 * initialize private struct file data. 134 * if we fail, clean up by dropping fmode reference on the ceph_inode 135 */ 136 static int ceph_init_file(struct inode *inode, struct file *file, int fmode) 137 { 138 struct ceph_file_info *cf; 139 int ret = 0; 140 141 switch (inode->i_mode & S_IFMT) { 142 case S_IFREG: 143 ceph_fscache_register_inode_cookie(inode); 144 ceph_fscache_file_set_cookie(inode, file); 145 case S_IFDIR: 146 dout("init_file %p %p 0%o (regular)\n", inode, file, 147 inode->i_mode); 148 cf = kmem_cache_zalloc(ceph_file_cachep, GFP_KERNEL); 149 if (cf == NULL) { 150 ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ 151 return -ENOMEM; 152 } 153 cf->fmode = fmode; 154 cf->next_offset = 2; 155 cf->readdir_cache_idx = -1; 156 file->private_data = cf; 157 BUG_ON(inode->i_fop->release != ceph_release); 158 break; 159 160 case S_IFLNK: 161 dout("init_file %p %p 0%o (symlink)\n", inode, file, 162 inode->i_mode); 163 ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ 164 break; 165 166 default: 167 dout("init_file %p %p 0%o (special)\n", inode, file, 168 inode->i_mode); 169 /* 170 * we need to drop the open ref now, since we don't 171 * have .release set to ceph_release. 172 */ 173 ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ 174 BUG_ON(inode->i_fop->release == ceph_release); 175 176 /* call the proper open fop */ 177 ret = inode->i_fop->open(inode, file); 178 } 179 return ret; 180 } 181 182 /* 183 * try renew caps after session gets killed. 184 */ 185 int ceph_renew_caps(struct inode *inode) 186 { 187 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc; 188 struct ceph_inode_info *ci = ceph_inode(inode); 189 struct ceph_mds_request *req; 190 int err, flags, wanted; 191 192 spin_lock(&ci->i_ceph_lock); 193 wanted = __ceph_caps_file_wanted(ci); 194 if (__ceph_is_any_real_caps(ci) && 195 (!(wanted & CEPH_CAP_ANY_WR) == 0 || ci->i_auth_cap)) { 196 int issued = __ceph_caps_issued(ci, NULL); 197 spin_unlock(&ci->i_ceph_lock); 198 dout("renew caps %p want %s issued %s updating mds_wanted\n", 199 inode, ceph_cap_string(wanted), ceph_cap_string(issued)); 200 ceph_check_caps(ci, 0, NULL); 201 return 0; 202 } 203 spin_unlock(&ci->i_ceph_lock); 204 205 flags = 0; 206 if ((wanted & CEPH_CAP_FILE_RD) && (wanted & CEPH_CAP_FILE_WR)) 207 flags = O_RDWR; 208 else if (wanted & CEPH_CAP_FILE_RD) 209 flags = O_RDONLY; 210 else if (wanted & CEPH_CAP_FILE_WR) 211 flags = O_WRONLY; 212 #ifdef O_LAZY 213 if (wanted & CEPH_CAP_FILE_LAZYIO) 214 flags |= O_LAZY; 215 #endif 216 217 req = prepare_open_request(inode->i_sb, flags, 0); 218 if (IS_ERR(req)) { 219 err = PTR_ERR(req); 220 goto out; 221 } 222 223 req->r_inode = inode; 224 ihold(inode); 225 req->r_num_caps = 1; 226 req->r_fmode = -1; 227 228 err = ceph_mdsc_do_request(mdsc, NULL, req); 229 ceph_mdsc_put_request(req); 230 out: 231 dout("renew caps %p open result=%d\n", inode, err); 232 return err < 0 ? err : 0; 233 } 234 235 /* 236 * If we already have the requisite capabilities, we can satisfy 237 * the open request locally (no need to request new caps from the 238 * MDS). We do, however, need to inform the MDS (asynchronously) 239 * if our wanted caps set expands. 240 */ 241 int ceph_open(struct inode *inode, struct file *file) 242 { 243 struct ceph_inode_info *ci = ceph_inode(inode); 244 struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb); 245 struct ceph_mds_client *mdsc = fsc->mdsc; 246 struct ceph_mds_request *req; 247 struct ceph_file_info *cf = file->private_data; 248 int err; 249 int flags, fmode, wanted; 250 251 if (cf) { 252 dout("open file %p is already opened\n", file); 253 return 0; 254 } 255 256 /* filter out O_CREAT|O_EXCL; vfs did that already. yuck. */ 257 flags = file->f_flags & ~(O_CREAT|O_EXCL); 258 if (S_ISDIR(inode->i_mode)) 259 flags = O_DIRECTORY; /* mds likes to know */ 260 261 dout("open inode %p ino %llx.%llx file %p flags %d (%d)\n", inode, 262 ceph_vinop(inode), file, flags, file->f_flags); 263 fmode = ceph_flags_to_mode(flags); 264 wanted = ceph_caps_for_mode(fmode); 265 266 /* snapped files are read-only */ 267 if (ceph_snap(inode) != CEPH_NOSNAP && (file->f_mode & FMODE_WRITE)) 268 return -EROFS; 269 270 /* trivially open snapdir */ 271 if (ceph_snap(inode) == CEPH_SNAPDIR) { 272 spin_lock(&ci->i_ceph_lock); 273 __ceph_get_fmode(ci, fmode); 274 spin_unlock(&ci->i_ceph_lock); 275 return ceph_init_file(inode, file, fmode); 276 } 277 278 /* 279 * No need to block if we have caps on the auth MDS (for 280 * write) or any MDS (for read). Update wanted set 281 * asynchronously. 282 */ 283 spin_lock(&ci->i_ceph_lock); 284 if (__ceph_is_any_real_caps(ci) && 285 (((fmode & CEPH_FILE_MODE_WR) == 0) || ci->i_auth_cap)) { 286 int mds_wanted = __ceph_caps_mds_wanted(ci); 287 int issued = __ceph_caps_issued(ci, NULL); 288 289 dout("open %p fmode %d want %s issued %s using existing\n", 290 inode, fmode, ceph_cap_string(wanted), 291 ceph_cap_string(issued)); 292 __ceph_get_fmode(ci, fmode); 293 spin_unlock(&ci->i_ceph_lock); 294 295 /* adjust wanted? */ 296 if ((issued & wanted) != wanted && 297 (mds_wanted & wanted) != wanted && 298 ceph_snap(inode) != CEPH_SNAPDIR) 299 ceph_check_caps(ci, 0, NULL); 300 301 return ceph_init_file(inode, file, fmode); 302 } else if (ceph_snap(inode) != CEPH_NOSNAP && 303 (ci->i_snap_caps & wanted) == wanted) { 304 __ceph_get_fmode(ci, fmode); 305 spin_unlock(&ci->i_ceph_lock); 306 return ceph_init_file(inode, file, fmode); 307 } 308 309 spin_unlock(&ci->i_ceph_lock); 310 311 dout("open fmode %d wants %s\n", fmode, ceph_cap_string(wanted)); 312 req = prepare_open_request(inode->i_sb, flags, 0); 313 if (IS_ERR(req)) { 314 err = PTR_ERR(req); 315 goto out; 316 } 317 req->r_inode = inode; 318 ihold(inode); 319 320 req->r_num_caps = 1; 321 err = ceph_mdsc_do_request(mdsc, NULL, req); 322 if (!err) 323 err = ceph_init_file(inode, file, req->r_fmode); 324 ceph_mdsc_put_request(req); 325 dout("open result=%d on %llx.%llx\n", err, ceph_vinop(inode)); 326 out: 327 return err; 328 } 329 330 331 /* 332 * Do a lookup + open with a single request. If we get a non-existent 333 * file or symlink, return 1 so the VFS can retry. 334 */ 335 int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 336 struct file *file, unsigned flags, umode_t mode, 337 int *opened) 338 { 339 struct ceph_fs_client *fsc = ceph_sb_to_client(dir->i_sb); 340 struct ceph_mds_client *mdsc = fsc->mdsc; 341 struct ceph_mds_request *req; 342 struct dentry *dn; 343 struct ceph_acls_info acls = {}; 344 int mask; 345 int err; 346 347 dout("atomic_open %p dentry %p '%pd' %s flags %d mode 0%o\n", 348 dir, dentry, dentry, 349 d_unhashed(dentry) ? "unhashed" : "hashed", flags, mode); 350 351 if (dentry->d_name.len > NAME_MAX) 352 return -ENAMETOOLONG; 353 354 err = ceph_init_dentry(dentry); 355 if (err < 0) 356 return err; 357 358 if (flags & O_CREAT) { 359 err = ceph_pre_init_acls(dir, &mode, &acls); 360 if (err < 0) 361 return err; 362 } 363 364 /* do the open */ 365 req = prepare_open_request(dir->i_sb, flags, mode); 366 if (IS_ERR(req)) { 367 err = PTR_ERR(req); 368 goto out_acl; 369 } 370 req->r_dentry = dget(dentry); 371 req->r_num_caps = 2; 372 if (flags & O_CREAT) { 373 req->r_dentry_drop = CEPH_CAP_FILE_SHARED; 374 req->r_dentry_unless = CEPH_CAP_FILE_EXCL; 375 if (acls.pagelist) { 376 req->r_pagelist = acls.pagelist; 377 acls.pagelist = NULL; 378 } 379 } 380 381 mask = CEPH_STAT_CAP_INODE | CEPH_CAP_AUTH_SHARED; 382 if (ceph_security_xattr_wanted(dir)) 383 mask |= CEPH_CAP_XATTR_SHARED; 384 req->r_args.open.mask = cpu_to_le32(mask); 385 386 req->r_locked_dir = dir; /* caller holds dir->i_mutex */ 387 err = ceph_mdsc_do_request(mdsc, 388 (flags & (O_CREAT|O_TRUNC)) ? dir : NULL, 389 req); 390 err = ceph_handle_snapdir(req, dentry, err); 391 if (err) 392 goto out_req; 393 394 if ((flags & O_CREAT) && !req->r_reply_info.head->is_dentry) 395 err = ceph_handle_notrace_create(dir, dentry); 396 397 if (d_in_lookup(dentry)) { 398 dn = ceph_finish_lookup(req, dentry, err); 399 if (IS_ERR(dn)) 400 err = PTR_ERR(dn); 401 } else { 402 /* we were given a hashed negative dentry */ 403 dn = NULL; 404 } 405 if (err) 406 goto out_req; 407 if (dn || d_really_is_negative(dentry) || d_is_symlink(dentry)) { 408 /* make vfs retry on splice, ENOENT, or symlink */ 409 dout("atomic_open finish_no_open on dn %p\n", dn); 410 err = finish_no_open(file, dn); 411 } else { 412 dout("atomic_open finish_open on dn %p\n", dn); 413 if (req->r_op == CEPH_MDS_OP_CREATE && req->r_reply_info.has_create_ino) { 414 ceph_init_inode_acls(d_inode(dentry), &acls); 415 *opened |= FILE_CREATED; 416 } 417 err = finish_open(file, dentry, ceph_open, opened); 418 } 419 out_req: 420 if (!req->r_err && req->r_target_inode) 421 ceph_put_fmode(ceph_inode(req->r_target_inode), req->r_fmode); 422 ceph_mdsc_put_request(req); 423 out_acl: 424 ceph_release_acls_info(&acls); 425 dout("atomic_open result=%d\n", err); 426 return err; 427 } 428 429 int ceph_release(struct inode *inode, struct file *file) 430 { 431 struct ceph_inode_info *ci = ceph_inode(inode); 432 struct ceph_file_info *cf = file->private_data; 433 434 dout("release inode %p file %p\n", inode, file); 435 ceph_put_fmode(ci, cf->fmode); 436 if (cf->last_readdir) 437 ceph_mdsc_put_request(cf->last_readdir); 438 kfree(cf->last_name); 439 kfree(cf->dir_info); 440 kmem_cache_free(ceph_file_cachep, cf); 441 442 /* wake up anyone waiting for caps on this inode */ 443 wake_up_all(&ci->i_cap_wq); 444 return 0; 445 } 446 447 enum { 448 HAVE_RETRIED = 1, 449 CHECK_EOF = 2, 450 READ_INLINE = 3, 451 }; 452 453 /* 454 * Read a range of bytes striped over one or more objects. Iterate over 455 * objects we stripe over. (That's not atomic, but good enough for now.) 456 * 457 * If we get a short result from the OSD, check against i_size; we need to 458 * only return a short read to the caller if we hit EOF. 459 */ 460 static int striped_read(struct inode *inode, 461 u64 off, u64 len, 462 struct page **pages, int num_pages, 463 int *checkeof) 464 { 465 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 466 struct ceph_inode_info *ci = ceph_inode(inode); 467 u64 pos, this_len, left; 468 loff_t i_size; 469 int page_align, pages_left; 470 int read, ret; 471 struct page **page_pos; 472 bool hit_stripe, was_short; 473 474 /* 475 * we may need to do multiple reads. not atomic, unfortunately. 476 */ 477 pos = off; 478 left = len; 479 page_pos = pages; 480 pages_left = num_pages; 481 read = 0; 482 483 more: 484 page_align = pos & ~PAGE_MASK; 485 this_len = left; 486 ret = ceph_osdc_readpages(&fsc->client->osdc, ceph_vino(inode), 487 &ci->i_layout, pos, &this_len, 488 ci->i_truncate_seq, 489 ci->i_truncate_size, 490 page_pos, pages_left, page_align); 491 if (ret == -ENOENT) 492 ret = 0; 493 hit_stripe = this_len < left; 494 was_short = ret >= 0 && ret < this_len; 495 dout("striped_read %llu~%llu (read %u) got %d%s%s\n", pos, left, read, 496 ret, hit_stripe ? " HITSTRIPE" : "", was_short ? " SHORT" : ""); 497 498 i_size = i_size_read(inode); 499 if (ret >= 0) { 500 int didpages; 501 if (was_short && (pos + ret < i_size)) { 502 int zlen = min(this_len - ret, i_size - pos - ret); 503 int zoff = (off & ~PAGE_MASK) + read + ret; 504 dout(" zero gap %llu to %llu\n", 505 pos + ret, pos + ret + zlen); 506 ceph_zero_page_vector_range(zoff, zlen, pages); 507 ret += zlen; 508 } 509 510 didpages = (page_align + ret) >> PAGE_SHIFT; 511 pos += ret; 512 read = pos - off; 513 left -= ret; 514 page_pos += didpages; 515 pages_left -= didpages; 516 517 /* hit stripe and need continue*/ 518 if (left && hit_stripe && pos < i_size) 519 goto more; 520 } 521 522 if (read > 0) { 523 ret = read; 524 /* did we bounce off eof? */ 525 if (pos + left > i_size) 526 *checkeof = CHECK_EOF; 527 } 528 529 dout("striped_read returns %d\n", ret); 530 return ret; 531 } 532 533 /* 534 * Completely synchronous read and write methods. Direct from __user 535 * buffer to osd, or directly to user pages (if O_DIRECT). 536 * 537 * If the read spans object boundary, just do multiple reads. 538 */ 539 static ssize_t ceph_sync_read(struct kiocb *iocb, struct iov_iter *i, 540 int *checkeof) 541 { 542 struct file *file = iocb->ki_filp; 543 struct inode *inode = file_inode(file); 544 struct page **pages; 545 u64 off = iocb->ki_pos; 546 int num_pages, ret; 547 size_t len = iov_iter_count(i); 548 549 dout("sync_read on file %p %llu~%u %s\n", file, off, 550 (unsigned)len, 551 (file->f_flags & O_DIRECT) ? "O_DIRECT" : ""); 552 553 if (!len) 554 return 0; 555 /* 556 * flush any page cache pages in this range. this 557 * will make concurrent normal and sync io slow, 558 * but it will at least behave sensibly when they are 559 * in sequence. 560 */ 561 ret = filemap_write_and_wait_range(inode->i_mapping, off, 562 off + len); 563 if (ret < 0) 564 return ret; 565 566 num_pages = calc_pages_for(off, len); 567 pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 568 if (IS_ERR(pages)) 569 return PTR_ERR(pages); 570 ret = striped_read(inode, off, len, pages, 571 num_pages, checkeof); 572 if (ret > 0) { 573 int l, k = 0; 574 size_t left = ret; 575 576 while (left) { 577 size_t page_off = off & ~PAGE_MASK; 578 size_t copy = min_t(size_t, left, 579 PAGE_SIZE - page_off); 580 l = copy_page_to_iter(pages[k++], page_off, copy, i); 581 off += l; 582 left -= l; 583 if (l < copy) 584 break; 585 } 586 } 587 ceph_release_page_vector(pages, num_pages); 588 589 if (off > iocb->ki_pos) { 590 ret = off - iocb->ki_pos; 591 iocb->ki_pos = off; 592 } 593 594 dout("sync_read result %d\n", ret); 595 return ret; 596 } 597 598 struct ceph_aio_request { 599 struct kiocb *iocb; 600 size_t total_len; 601 int write; 602 int error; 603 struct list_head osd_reqs; 604 unsigned num_reqs; 605 atomic_t pending_reqs; 606 struct timespec mtime; 607 struct ceph_cap_flush *prealloc_cf; 608 }; 609 610 struct ceph_aio_work { 611 struct work_struct work; 612 struct ceph_osd_request *req; 613 }; 614 615 static void ceph_aio_retry_work(struct work_struct *work); 616 617 static void ceph_aio_complete(struct inode *inode, 618 struct ceph_aio_request *aio_req) 619 { 620 struct ceph_inode_info *ci = ceph_inode(inode); 621 int ret; 622 623 if (!atomic_dec_and_test(&aio_req->pending_reqs)) 624 return; 625 626 ret = aio_req->error; 627 if (!ret) 628 ret = aio_req->total_len; 629 630 dout("ceph_aio_complete %p rc %d\n", inode, ret); 631 632 if (ret >= 0 && aio_req->write) { 633 int dirty; 634 635 loff_t endoff = aio_req->iocb->ki_pos + aio_req->total_len; 636 if (endoff > i_size_read(inode)) { 637 if (ceph_inode_set_size(inode, endoff)) 638 ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL); 639 } 640 641 spin_lock(&ci->i_ceph_lock); 642 ci->i_inline_version = CEPH_INLINE_NONE; 643 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, 644 &aio_req->prealloc_cf); 645 spin_unlock(&ci->i_ceph_lock); 646 if (dirty) 647 __mark_inode_dirty(inode, dirty); 648 649 } 650 651 ceph_put_cap_refs(ci, (aio_req->write ? CEPH_CAP_FILE_WR : 652 CEPH_CAP_FILE_RD)); 653 654 aio_req->iocb->ki_complete(aio_req->iocb, ret, 0); 655 656 ceph_free_cap_flush(aio_req->prealloc_cf); 657 kfree(aio_req); 658 } 659 660 static void ceph_aio_complete_req(struct ceph_osd_request *req) 661 { 662 int rc = req->r_result; 663 struct inode *inode = req->r_inode; 664 struct ceph_aio_request *aio_req = req->r_priv; 665 struct ceph_osd_data *osd_data = osd_req_op_extent_osd_data(req, 0); 666 int num_pages = calc_pages_for((u64)osd_data->alignment, 667 osd_data->length); 668 669 dout("ceph_aio_complete_req %p rc %d bytes %llu\n", 670 inode, rc, osd_data->length); 671 672 if (rc == -EOLDSNAPC) { 673 struct ceph_aio_work *aio_work; 674 BUG_ON(!aio_req->write); 675 676 aio_work = kmalloc(sizeof(*aio_work), GFP_NOFS); 677 if (aio_work) { 678 INIT_WORK(&aio_work->work, ceph_aio_retry_work); 679 aio_work->req = req; 680 queue_work(ceph_inode_to_client(inode)->wb_wq, 681 &aio_work->work); 682 return; 683 } 684 rc = -ENOMEM; 685 } else if (!aio_req->write) { 686 if (rc == -ENOENT) 687 rc = 0; 688 if (rc >= 0 && osd_data->length > rc) { 689 int zoff = osd_data->alignment + rc; 690 int zlen = osd_data->length - rc; 691 /* 692 * If read is satisfied by single OSD request, 693 * it can pass EOF. Otherwise read is within 694 * i_size. 695 */ 696 if (aio_req->num_reqs == 1) { 697 loff_t i_size = i_size_read(inode); 698 loff_t endoff = aio_req->iocb->ki_pos + rc; 699 if (endoff < i_size) 700 zlen = min_t(size_t, zlen, 701 i_size - endoff); 702 aio_req->total_len = rc + zlen; 703 } 704 705 if (zlen > 0) 706 ceph_zero_page_vector_range(zoff, zlen, 707 osd_data->pages); 708 } 709 } 710 711 ceph_put_page_vector(osd_data->pages, num_pages, !aio_req->write); 712 ceph_osdc_put_request(req); 713 714 if (rc < 0) 715 cmpxchg(&aio_req->error, 0, rc); 716 717 ceph_aio_complete(inode, aio_req); 718 return; 719 } 720 721 static void ceph_aio_retry_work(struct work_struct *work) 722 { 723 struct ceph_aio_work *aio_work = 724 container_of(work, struct ceph_aio_work, work); 725 struct ceph_osd_request *orig_req = aio_work->req; 726 struct ceph_aio_request *aio_req = orig_req->r_priv; 727 struct inode *inode = orig_req->r_inode; 728 struct ceph_inode_info *ci = ceph_inode(inode); 729 struct ceph_snap_context *snapc; 730 struct ceph_osd_request *req; 731 int ret; 732 733 spin_lock(&ci->i_ceph_lock); 734 if (__ceph_have_pending_cap_snap(ci)) { 735 struct ceph_cap_snap *capsnap = 736 list_last_entry(&ci->i_cap_snaps, 737 struct ceph_cap_snap, 738 ci_item); 739 snapc = ceph_get_snap_context(capsnap->context); 740 } else { 741 BUG_ON(!ci->i_head_snapc); 742 snapc = ceph_get_snap_context(ci->i_head_snapc); 743 } 744 spin_unlock(&ci->i_ceph_lock); 745 746 req = ceph_osdc_alloc_request(orig_req->r_osdc, snapc, 2, 747 false, GFP_NOFS); 748 if (!req) { 749 ret = -ENOMEM; 750 req = orig_req; 751 goto out; 752 } 753 754 req->r_flags = CEPH_OSD_FLAG_ORDERSNAP | 755 CEPH_OSD_FLAG_ONDISK | 756 CEPH_OSD_FLAG_WRITE; 757 ceph_oloc_copy(&req->r_base_oloc, &orig_req->r_base_oloc); 758 ceph_oid_copy(&req->r_base_oid, &orig_req->r_base_oid); 759 760 ret = ceph_osdc_alloc_messages(req, GFP_NOFS); 761 if (ret) { 762 ceph_osdc_put_request(req); 763 req = orig_req; 764 goto out; 765 } 766 767 req->r_ops[0] = orig_req->r_ops[0]; 768 osd_req_op_init(req, 1, CEPH_OSD_OP_STARTSYNC, 0); 769 770 req->r_mtime = aio_req->mtime; 771 req->r_data_offset = req->r_ops[0].extent.offset; 772 773 ceph_osdc_put_request(orig_req); 774 775 req->r_callback = ceph_aio_complete_req; 776 req->r_inode = inode; 777 req->r_priv = aio_req; 778 779 ret = ceph_osdc_start_request(req->r_osdc, req, false); 780 out: 781 if (ret < 0) { 782 req->r_result = ret; 783 ceph_aio_complete_req(req); 784 } 785 786 ceph_put_snap_context(snapc); 787 kfree(aio_work); 788 } 789 790 /* 791 * Write commit request unsafe callback, called to tell us when a 792 * request is unsafe (that is, in flight--has been handed to the 793 * messenger to send to its target osd). It is called again when 794 * we've received a response message indicating the request is 795 * "safe" (its CEPH_OSD_FLAG_ONDISK flag is set), or when a request 796 * is completed early (and unsuccessfully) due to a timeout or 797 * interrupt. 798 * 799 * This is used if we requested both an ACK and ONDISK commit reply 800 * from the OSD. 801 */ 802 static void ceph_sync_write_unsafe(struct ceph_osd_request *req, bool unsafe) 803 { 804 struct ceph_inode_info *ci = ceph_inode(req->r_inode); 805 806 dout("%s %p tid %llu %ssafe\n", __func__, req, req->r_tid, 807 unsafe ? "un" : ""); 808 if (unsafe) { 809 ceph_get_cap_refs(ci, CEPH_CAP_FILE_WR); 810 spin_lock(&ci->i_unsafe_lock); 811 list_add_tail(&req->r_unsafe_item, 812 &ci->i_unsafe_writes); 813 spin_unlock(&ci->i_unsafe_lock); 814 815 complete_all(&req->r_completion); 816 } else { 817 spin_lock(&ci->i_unsafe_lock); 818 list_del_init(&req->r_unsafe_item); 819 spin_unlock(&ci->i_unsafe_lock); 820 ceph_put_cap_refs(ci, CEPH_CAP_FILE_WR); 821 } 822 } 823 824 /* 825 * Wait on any unsafe replies for the given inode. First wait on the 826 * newest request, and make that the upper bound. Then, if there are 827 * more requests, keep waiting on the oldest as long as it is still older 828 * than the original request. 829 */ 830 void ceph_sync_write_wait(struct inode *inode) 831 { 832 struct ceph_inode_info *ci = ceph_inode(inode); 833 struct list_head *head = &ci->i_unsafe_writes; 834 struct ceph_osd_request *req; 835 u64 last_tid; 836 837 if (!S_ISREG(inode->i_mode)) 838 return; 839 840 spin_lock(&ci->i_unsafe_lock); 841 if (list_empty(head)) 842 goto out; 843 844 /* set upper bound as _last_ entry in chain */ 845 846 req = list_last_entry(head, struct ceph_osd_request, 847 r_unsafe_item); 848 last_tid = req->r_tid; 849 850 do { 851 ceph_osdc_get_request(req); 852 spin_unlock(&ci->i_unsafe_lock); 853 854 dout("sync_write_wait on tid %llu (until %llu)\n", 855 req->r_tid, last_tid); 856 wait_for_completion(&req->r_safe_completion); 857 ceph_osdc_put_request(req); 858 859 spin_lock(&ci->i_unsafe_lock); 860 /* 861 * from here on look at first entry in chain, since we 862 * only want to wait for anything older than last_tid 863 */ 864 if (list_empty(head)) 865 break; 866 req = list_first_entry(head, struct ceph_osd_request, 867 r_unsafe_item); 868 } while (req->r_tid < last_tid); 869 out: 870 spin_unlock(&ci->i_unsafe_lock); 871 } 872 873 static ssize_t 874 ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter, 875 struct ceph_snap_context *snapc, 876 struct ceph_cap_flush **pcf) 877 { 878 struct file *file = iocb->ki_filp; 879 struct inode *inode = file_inode(file); 880 struct ceph_inode_info *ci = ceph_inode(inode); 881 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 882 struct ceph_vino vino; 883 struct ceph_osd_request *req; 884 struct page **pages; 885 struct ceph_aio_request *aio_req = NULL; 886 int num_pages = 0; 887 int flags; 888 int ret; 889 struct timespec mtime = current_time(inode); 890 size_t count = iov_iter_count(iter); 891 loff_t pos = iocb->ki_pos; 892 bool write = iov_iter_rw(iter) == WRITE; 893 894 if (write && ceph_snap(file_inode(file)) != CEPH_NOSNAP) 895 return -EROFS; 896 897 dout("sync_direct_read_write (%s) on file %p %lld~%u\n", 898 (write ? "write" : "read"), file, pos, (unsigned)count); 899 900 ret = filemap_write_and_wait_range(inode->i_mapping, pos, pos + count); 901 if (ret < 0) 902 return ret; 903 904 if (write) { 905 int ret2 = invalidate_inode_pages2_range(inode->i_mapping, 906 pos >> PAGE_SHIFT, 907 (pos + count) >> PAGE_SHIFT); 908 if (ret2 < 0) 909 dout("invalidate_inode_pages2_range returned %d\n", ret); 910 911 flags = CEPH_OSD_FLAG_ORDERSNAP | 912 CEPH_OSD_FLAG_ONDISK | 913 CEPH_OSD_FLAG_WRITE; 914 } else { 915 flags = CEPH_OSD_FLAG_READ; 916 } 917 918 while (iov_iter_count(iter) > 0) { 919 u64 size = dio_get_pagev_size(iter); 920 size_t start = 0; 921 ssize_t len; 922 923 vino = ceph_vino(inode); 924 req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, 925 vino, pos, &size, 0, 926 /*include a 'startsync' command*/ 927 write ? 2 : 1, 928 write ? CEPH_OSD_OP_WRITE : 929 CEPH_OSD_OP_READ, 930 flags, snapc, 931 ci->i_truncate_seq, 932 ci->i_truncate_size, 933 false); 934 if (IS_ERR(req)) { 935 ret = PTR_ERR(req); 936 break; 937 } 938 939 len = size; 940 pages = dio_get_pages_alloc(iter, len, &start, &num_pages); 941 if (IS_ERR(pages)) { 942 ceph_osdc_put_request(req); 943 ret = PTR_ERR(pages); 944 break; 945 } 946 947 /* 948 * To simplify error handling, allow AIO when IO within i_size 949 * or IO can be satisfied by single OSD request. 950 */ 951 if (pos == iocb->ki_pos && !is_sync_kiocb(iocb) && 952 (len == count || pos + count <= i_size_read(inode))) { 953 aio_req = kzalloc(sizeof(*aio_req), GFP_KERNEL); 954 if (aio_req) { 955 aio_req->iocb = iocb; 956 aio_req->write = write; 957 INIT_LIST_HEAD(&aio_req->osd_reqs); 958 if (write) { 959 aio_req->mtime = mtime; 960 swap(aio_req->prealloc_cf, *pcf); 961 } 962 } 963 /* ignore error */ 964 } 965 966 if (write) { 967 /* 968 * throw out any page cache pages in this range. this 969 * may block. 970 */ 971 truncate_inode_pages_range(inode->i_mapping, pos, 972 (pos+len) | (PAGE_SIZE - 1)); 973 974 osd_req_op_init(req, 1, CEPH_OSD_OP_STARTSYNC, 0); 975 req->r_mtime = mtime; 976 } 977 978 osd_req_op_extent_osd_data_pages(req, 0, pages, len, start, 979 false, false); 980 981 if (aio_req) { 982 aio_req->total_len += len; 983 aio_req->num_reqs++; 984 atomic_inc(&aio_req->pending_reqs); 985 986 req->r_callback = ceph_aio_complete_req; 987 req->r_inode = inode; 988 req->r_priv = aio_req; 989 list_add_tail(&req->r_unsafe_item, &aio_req->osd_reqs); 990 991 pos += len; 992 iov_iter_advance(iter, len); 993 continue; 994 } 995 996 ret = ceph_osdc_start_request(req->r_osdc, req, false); 997 if (!ret) 998 ret = ceph_osdc_wait_request(&fsc->client->osdc, req); 999 1000 size = i_size_read(inode); 1001 if (!write) { 1002 if (ret == -ENOENT) 1003 ret = 0; 1004 if (ret >= 0 && ret < len && pos + ret < size) { 1005 int zlen = min_t(size_t, len - ret, 1006 size - pos - ret); 1007 ceph_zero_page_vector_range(start + ret, zlen, 1008 pages); 1009 ret += zlen; 1010 } 1011 if (ret >= 0) 1012 len = ret; 1013 } 1014 1015 ceph_put_page_vector(pages, num_pages, !write); 1016 1017 ceph_osdc_put_request(req); 1018 if (ret < 0) 1019 break; 1020 1021 pos += len; 1022 iov_iter_advance(iter, len); 1023 1024 if (!write && pos >= size) 1025 break; 1026 1027 if (write && pos > size) { 1028 if (ceph_inode_set_size(inode, pos)) 1029 ceph_check_caps(ceph_inode(inode), 1030 CHECK_CAPS_AUTHONLY, 1031 NULL); 1032 } 1033 } 1034 1035 if (aio_req) { 1036 LIST_HEAD(osd_reqs); 1037 1038 if (aio_req->num_reqs == 0) { 1039 kfree(aio_req); 1040 return ret; 1041 } 1042 1043 ceph_get_cap_refs(ci, write ? CEPH_CAP_FILE_WR : 1044 CEPH_CAP_FILE_RD); 1045 1046 list_splice(&aio_req->osd_reqs, &osd_reqs); 1047 while (!list_empty(&osd_reqs)) { 1048 req = list_first_entry(&osd_reqs, 1049 struct ceph_osd_request, 1050 r_unsafe_item); 1051 list_del_init(&req->r_unsafe_item); 1052 if (ret >= 0) 1053 ret = ceph_osdc_start_request(req->r_osdc, 1054 req, false); 1055 if (ret < 0) { 1056 req->r_result = ret; 1057 ceph_aio_complete_req(req); 1058 } 1059 } 1060 return -EIOCBQUEUED; 1061 } 1062 1063 if (ret != -EOLDSNAPC && pos > iocb->ki_pos) { 1064 ret = pos - iocb->ki_pos; 1065 iocb->ki_pos = pos; 1066 } 1067 return ret; 1068 } 1069 1070 /* 1071 * Synchronous write, straight from __user pointer or user pages. 1072 * 1073 * If write spans object boundary, just do multiple writes. (For a 1074 * correct atomic write, we should e.g. take write locks on all 1075 * objects, rollback on failure, etc.) 1076 */ 1077 static ssize_t 1078 ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos, 1079 struct ceph_snap_context *snapc) 1080 { 1081 struct file *file = iocb->ki_filp; 1082 struct inode *inode = file_inode(file); 1083 struct ceph_inode_info *ci = ceph_inode(inode); 1084 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1085 struct ceph_vino vino; 1086 struct ceph_osd_request *req; 1087 struct page **pages; 1088 u64 len; 1089 int num_pages; 1090 int written = 0; 1091 int flags; 1092 int check_caps = 0; 1093 int ret; 1094 struct timespec mtime = current_time(inode); 1095 size_t count = iov_iter_count(from); 1096 1097 if (ceph_snap(file_inode(file)) != CEPH_NOSNAP) 1098 return -EROFS; 1099 1100 dout("sync_write on file %p %lld~%u\n", file, pos, (unsigned)count); 1101 1102 ret = filemap_write_and_wait_range(inode->i_mapping, pos, pos + count); 1103 if (ret < 0) 1104 return ret; 1105 1106 ret = invalidate_inode_pages2_range(inode->i_mapping, 1107 pos >> PAGE_SHIFT, 1108 (pos + count) >> PAGE_SHIFT); 1109 if (ret < 0) 1110 dout("invalidate_inode_pages2_range returned %d\n", ret); 1111 1112 flags = CEPH_OSD_FLAG_ORDERSNAP | 1113 CEPH_OSD_FLAG_ONDISK | 1114 CEPH_OSD_FLAG_WRITE | 1115 CEPH_OSD_FLAG_ACK; 1116 1117 while ((len = iov_iter_count(from)) > 0) { 1118 size_t left; 1119 int n; 1120 1121 vino = ceph_vino(inode); 1122 req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, 1123 vino, pos, &len, 0, 1, 1124 CEPH_OSD_OP_WRITE, flags, snapc, 1125 ci->i_truncate_seq, 1126 ci->i_truncate_size, 1127 false); 1128 if (IS_ERR(req)) { 1129 ret = PTR_ERR(req); 1130 break; 1131 } 1132 1133 /* 1134 * write from beginning of first page, 1135 * regardless of io alignment 1136 */ 1137 num_pages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT; 1138 1139 pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 1140 if (IS_ERR(pages)) { 1141 ret = PTR_ERR(pages); 1142 goto out; 1143 } 1144 1145 left = len; 1146 for (n = 0; n < num_pages; n++) { 1147 size_t plen = min_t(size_t, left, PAGE_SIZE); 1148 ret = copy_page_from_iter(pages[n], 0, plen, from); 1149 if (ret != plen) { 1150 ret = -EFAULT; 1151 break; 1152 } 1153 left -= ret; 1154 } 1155 1156 if (ret < 0) { 1157 ceph_release_page_vector(pages, num_pages); 1158 goto out; 1159 } 1160 1161 /* get a second commit callback */ 1162 req->r_unsafe_callback = ceph_sync_write_unsafe; 1163 req->r_inode = inode; 1164 1165 osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0, 1166 false, true); 1167 1168 req->r_mtime = mtime; 1169 ret = ceph_osdc_start_request(&fsc->client->osdc, req, false); 1170 if (!ret) 1171 ret = ceph_osdc_wait_request(&fsc->client->osdc, req); 1172 1173 out: 1174 ceph_osdc_put_request(req); 1175 if (ret == 0) { 1176 pos += len; 1177 written += len; 1178 1179 if (pos > i_size_read(inode)) { 1180 check_caps = ceph_inode_set_size(inode, pos); 1181 if (check_caps) 1182 ceph_check_caps(ceph_inode(inode), 1183 CHECK_CAPS_AUTHONLY, 1184 NULL); 1185 } 1186 } else 1187 break; 1188 } 1189 1190 if (ret != -EOLDSNAPC && written > 0) { 1191 ret = written; 1192 iocb->ki_pos = pos; 1193 } 1194 return ret; 1195 } 1196 1197 /* 1198 * Wrap generic_file_aio_read with checks for cap bits on the inode. 1199 * Atomically grab references, so that those bits are not released 1200 * back to the MDS mid-read. 1201 * 1202 * Hmm, the sync read case isn't actually async... should it be? 1203 */ 1204 static ssize_t ceph_read_iter(struct kiocb *iocb, struct iov_iter *to) 1205 { 1206 struct file *filp = iocb->ki_filp; 1207 struct ceph_file_info *fi = filp->private_data; 1208 size_t len = iov_iter_count(to); 1209 struct inode *inode = file_inode(filp); 1210 struct ceph_inode_info *ci = ceph_inode(inode); 1211 struct page *pinned_page = NULL; 1212 ssize_t ret; 1213 int want, got = 0; 1214 int retry_op = 0, read = 0; 1215 1216 again: 1217 dout("aio_read %p %llx.%llx %llu~%u trying to get caps on %p\n", 1218 inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len, inode); 1219 1220 if (fi->fmode & CEPH_FILE_MODE_LAZY) 1221 want = CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO; 1222 else 1223 want = CEPH_CAP_FILE_CACHE; 1224 ret = ceph_get_caps(ci, CEPH_CAP_FILE_RD, want, -1, &got, &pinned_page); 1225 if (ret < 0) 1226 return ret; 1227 1228 if ((got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0 || 1229 (iocb->ki_flags & IOCB_DIRECT) || 1230 (fi->flags & CEPH_F_SYNC)) { 1231 1232 dout("aio_sync_read %p %llx.%llx %llu~%u got cap refs on %s\n", 1233 inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len, 1234 ceph_cap_string(got)); 1235 1236 if (ci->i_inline_version == CEPH_INLINE_NONE) { 1237 if (!retry_op && (iocb->ki_flags & IOCB_DIRECT)) { 1238 ret = ceph_direct_read_write(iocb, to, 1239 NULL, NULL); 1240 if (ret >= 0 && ret < len) 1241 retry_op = CHECK_EOF; 1242 } else { 1243 ret = ceph_sync_read(iocb, to, &retry_op); 1244 } 1245 } else { 1246 retry_op = READ_INLINE; 1247 } 1248 } else { 1249 dout("aio_read %p %llx.%llx %llu~%u got cap refs on %s\n", 1250 inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len, 1251 ceph_cap_string(got)); 1252 1253 ret = generic_file_read_iter(iocb, to); 1254 } 1255 dout("aio_read %p %llx.%llx dropping cap refs on %s = %d\n", 1256 inode, ceph_vinop(inode), ceph_cap_string(got), (int)ret); 1257 if (pinned_page) { 1258 put_page(pinned_page); 1259 pinned_page = NULL; 1260 } 1261 ceph_put_cap_refs(ci, got); 1262 if (retry_op > HAVE_RETRIED && ret >= 0) { 1263 int statret; 1264 struct page *page = NULL; 1265 loff_t i_size; 1266 if (retry_op == READ_INLINE) { 1267 page = __page_cache_alloc(GFP_KERNEL); 1268 if (!page) 1269 return -ENOMEM; 1270 } 1271 1272 statret = __ceph_do_getattr(inode, page, 1273 CEPH_STAT_CAP_INLINE_DATA, !!page); 1274 if (statret < 0) { 1275 if (page) 1276 __free_page(page); 1277 if (statret == -ENODATA) { 1278 BUG_ON(retry_op != READ_INLINE); 1279 goto again; 1280 } 1281 return statret; 1282 } 1283 1284 i_size = i_size_read(inode); 1285 if (retry_op == READ_INLINE) { 1286 BUG_ON(ret > 0 || read > 0); 1287 if (iocb->ki_pos < i_size && 1288 iocb->ki_pos < PAGE_SIZE) { 1289 loff_t end = min_t(loff_t, i_size, 1290 iocb->ki_pos + len); 1291 end = min_t(loff_t, end, PAGE_SIZE); 1292 if (statret < end) 1293 zero_user_segment(page, statret, end); 1294 ret = copy_page_to_iter(page, 1295 iocb->ki_pos & ~PAGE_MASK, 1296 end - iocb->ki_pos, to); 1297 iocb->ki_pos += ret; 1298 read += ret; 1299 } 1300 if (iocb->ki_pos < i_size && read < len) { 1301 size_t zlen = min_t(size_t, len - read, 1302 i_size - iocb->ki_pos); 1303 ret = iov_iter_zero(zlen, to); 1304 iocb->ki_pos += ret; 1305 read += ret; 1306 } 1307 __free_pages(page, 0); 1308 return read; 1309 } 1310 1311 /* hit EOF or hole? */ 1312 if (retry_op == CHECK_EOF && iocb->ki_pos < i_size && 1313 ret < len) { 1314 dout("sync_read hit hole, ppos %lld < size %lld" 1315 ", reading more\n", iocb->ki_pos, i_size); 1316 1317 read += ret; 1318 len -= ret; 1319 retry_op = HAVE_RETRIED; 1320 goto again; 1321 } 1322 } 1323 1324 if (ret >= 0) 1325 ret += read; 1326 1327 return ret; 1328 } 1329 1330 /* 1331 * Take cap references to avoid releasing caps to MDS mid-write. 1332 * 1333 * If we are synchronous, and write with an old snap context, the OSD 1334 * may return EOLDSNAPC. In that case, retry the write.. _after_ 1335 * dropping our cap refs and allowing the pending snap to logically 1336 * complete _before_ this write occurs. 1337 * 1338 * If we are near ENOSPC, write synchronously. 1339 */ 1340 static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) 1341 { 1342 struct file *file = iocb->ki_filp; 1343 struct ceph_file_info *fi = file->private_data; 1344 struct inode *inode = file_inode(file); 1345 struct ceph_inode_info *ci = ceph_inode(inode); 1346 struct ceph_osd_client *osdc = 1347 &ceph_sb_to_client(inode->i_sb)->client->osdc; 1348 struct ceph_cap_flush *prealloc_cf; 1349 ssize_t count, written = 0; 1350 int err, want, got; 1351 loff_t pos; 1352 1353 if (ceph_snap(inode) != CEPH_NOSNAP) 1354 return -EROFS; 1355 1356 prealloc_cf = ceph_alloc_cap_flush(); 1357 if (!prealloc_cf) 1358 return -ENOMEM; 1359 1360 inode_lock(inode); 1361 1362 /* We can write back this queue in page reclaim */ 1363 current->backing_dev_info = inode_to_bdi(inode); 1364 1365 if (iocb->ki_flags & IOCB_APPEND) { 1366 err = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE, false); 1367 if (err < 0) 1368 goto out; 1369 } 1370 1371 err = generic_write_checks(iocb, from); 1372 if (err <= 0) 1373 goto out; 1374 1375 pos = iocb->ki_pos; 1376 count = iov_iter_count(from); 1377 err = file_remove_privs(file); 1378 if (err) 1379 goto out; 1380 1381 err = file_update_time(file); 1382 if (err) 1383 goto out; 1384 1385 if (ci->i_inline_version != CEPH_INLINE_NONE) { 1386 err = ceph_uninline_data(file, NULL); 1387 if (err < 0) 1388 goto out; 1389 } 1390 1391 retry_snap: 1392 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL)) { 1393 err = -ENOSPC; 1394 goto out; 1395 } 1396 1397 dout("aio_write %p %llx.%llx %llu~%zd getting caps. i_size %llu\n", 1398 inode, ceph_vinop(inode), pos, count, i_size_read(inode)); 1399 if (fi->fmode & CEPH_FILE_MODE_LAZY) 1400 want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO; 1401 else 1402 want = CEPH_CAP_FILE_BUFFER; 1403 got = 0; 1404 err = ceph_get_caps(ci, CEPH_CAP_FILE_WR, want, pos + count, 1405 &got, NULL); 1406 if (err < 0) 1407 goto out; 1408 1409 dout("aio_write %p %llx.%llx %llu~%zd got cap refs on %s\n", 1410 inode, ceph_vinop(inode), pos, count, ceph_cap_string(got)); 1411 1412 if ((got & (CEPH_CAP_FILE_BUFFER|CEPH_CAP_FILE_LAZYIO)) == 0 || 1413 (iocb->ki_flags & IOCB_DIRECT) || (fi->flags & CEPH_F_SYNC)) { 1414 struct ceph_snap_context *snapc; 1415 struct iov_iter data; 1416 inode_unlock(inode); 1417 1418 spin_lock(&ci->i_ceph_lock); 1419 if (__ceph_have_pending_cap_snap(ci)) { 1420 struct ceph_cap_snap *capsnap = 1421 list_last_entry(&ci->i_cap_snaps, 1422 struct ceph_cap_snap, 1423 ci_item); 1424 snapc = ceph_get_snap_context(capsnap->context); 1425 } else { 1426 BUG_ON(!ci->i_head_snapc); 1427 snapc = ceph_get_snap_context(ci->i_head_snapc); 1428 } 1429 spin_unlock(&ci->i_ceph_lock); 1430 1431 /* we might need to revert back to that point */ 1432 data = *from; 1433 if (iocb->ki_flags & IOCB_DIRECT) 1434 written = ceph_direct_read_write(iocb, &data, snapc, 1435 &prealloc_cf); 1436 else 1437 written = ceph_sync_write(iocb, &data, pos, snapc); 1438 if (written == -EOLDSNAPC) { 1439 dout("aio_write %p %llx.%llx %llu~%u" 1440 "got EOLDSNAPC, retrying\n", 1441 inode, ceph_vinop(inode), 1442 pos, (unsigned)count); 1443 inode_lock(inode); 1444 goto retry_snap; 1445 } 1446 if (written > 0) 1447 iov_iter_advance(from, written); 1448 ceph_put_snap_context(snapc); 1449 } else { 1450 /* 1451 * No need to acquire the i_truncate_mutex. Because 1452 * the MDS revokes Fwb caps before sending truncate 1453 * message to us. We can't get Fwb cap while there 1454 * are pending vmtruncate. So write and vmtruncate 1455 * can not run at the same time 1456 */ 1457 written = generic_perform_write(file, from, pos); 1458 if (likely(written >= 0)) 1459 iocb->ki_pos = pos + written; 1460 inode_unlock(inode); 1461 } 1462 1463 if (written >= 0) { 1464 int dirty; 1465 spin_lock(&ci->i_ceph_lock); 1466 ci->i_inline_version = CEPH_INLINE_NONE; 1467 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, 1468 &prealloc_cf); 1469 spin_unlock(&ci->i_ceph_lock); 1470 if (dirty) 1471 __mark_inode_dirty(inode, dirty); 1472 } 1473 1474 dout("aio_write %p %llx.%llx %llu~%u dropping cap refs on %s\n", 1475 inode, ceph_vinop(inode), pos, (unsigned)count, 1476 ceph_cap_string(got)); 1477 ceph_put_cap_refs(ci, got); 1478 1479 if (written >= 0) { 1480 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_NEARFULL)) 1481 iocb->ki_flags |= IOCB_DSYNC; 1482 1483 written = generic_write_sync(iocb, written); 1484 } 1485 1486 goto out_unlocked; 1487 1488 out: 1489 inode_unlock(inode); 1490 out_unlocked: 1491 ceph_free_cap_flush(prealloc_cf); 1492 current->backing_dev_info = NULL; 1493 return written ? written : err; 1494 } 1495 1496 /* 1497 * llseek. be sure to verify file size on SEEK_END. 1498 */ 1499 static loff_t ceph_llseek(struct file *file, loff_t offset, int whence) 1500 { 1501 struct inode *inode = file->f_mapping->host; 1502 loff_t i_size; 1503 loff_t ret; 1504 1505 inode_lock(inode); 1506 1507 if (whence == SEEK_END || whence == SEEK_DATA || whence == SEEK_HOLE) { 1508 ret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE, false); 1509 if (ret < 0) 1510 goto out; 1511 } 1512 1513 i_size = i_size_read(inode); 1514 switch (whence) { 1515 case SEEK_END: 1516 offset += i_size; 1517 break; 1518 case SEEK_CUR: 1519 /* 1520 * Here we special-case the lseek(fd, 0, SEEK_CUR) 1521 * position-querying operation. Avoid rewriting the "same" 1522 * f_pos value back to the file because a concurrent read(), 1523 * write() or lseek() might have altered it 1524 */ 1525 if (offset == 0) { 1526 ret = file->f_pos; 1527 goto out; 1528 } 1529 offset += file->f_pos; 1530 break; 1531 case SEEK_DATA: 1532 if (offset >= i_size) { 1533 ret = -ENXIO; 1534 goto out; 1535 } 1536 break; 1537 case SEEK_HOLE: 1538 if (offset >= i_size) { 1539 ret = -ENXIO; 1540 goto out; 1541 } 1542 offset = i_size; 1543 break; 1544 } 1545 1546 ret = vfs_setpos(file, offset, inode->i_sb->s_maxbytes); 1547 1548 out: 1549 inode_unlock(inode); 1550 return ret; 1551 } 1552 1553 static inline void ceph_zero_partial_page( 1554 struct inode *inode, loff_t offset, unsigned size) 1555 { 1556 struct page *page; 1557 pgoff_t index = offset >> PAGE_SHIFT; 1558 1559 page = find_lock_page(inode->i_mapping, index); 1560 if (page) { 1561 wait_on_page_writeback(page); 1562 zero_user(page, offset & (PAGE_SIZE - 1), size); 1563 unlock_page(page); 1564 put_page(page); 1565 } 1566 } 1567 1568 static void ceph_zero_pagecache_range(struct inode *inode, loff_t offset, 1569 loff_t length) 1570 { 1571 loff_t nearly = round_up(offset, PAGE_SIZE); 1572 if (offset < nearly) { 1573 loff_t size = nearly - offset; 1574 if (length < size) 1575 size = length; 1576 ceph_zero_partial_page(inode, offset, size); 1577 offset += size; 1578 length -= size; 1579 } 1580 if (length >= PAGE_SIZE) { 1581 loff_t size = round_down(length, PAGE_SIZE); 1582 truncate_pagecache_range(inode, offset, offset + size - 1); 1583 offset += size; 1584 length -= size; 1585 } 1586 if (length) 1587 ceph_zero_partial_page(inode, offset, length); 1588 } 1589 1590 static int ceph_zero_partial_object(struct inode *inode, 1591 loff_t offset, loff_t *length) 1592 { 1593 struct ceph_inode_info *ci = ceph_inode(inode); 1594 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1595 struct ceph_osd_request *req; 1596 int ret = 0; 1597 loff_t zero = 0; 1598 int op; 1599 1600 if (!length) { 1601 op = offset ? CEPH_OSD_OP_DELETE : CEPH_OSD_OP_TRUNCATE; 1602 length = &zero; 1603 } else { 1604 op = CEPH_OSD_OP_ZERO; 1605 } 1606 1607 req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, 1608 ceph_vino(inode), 1609 offset, length, 1610 0, 1, op, 1611 CEPH_OSD_FLAG_WRITE | 1612 CEPH_OSD_FLAG_ONDISK, 1613 NULL, 0, 0, false); 1614 if (IS_ERR(req)) { 1615 ret = PTR_ERR(req); 1616 goto out; 1617 } 1618 1619 req->r_mtime = inode->i_mtime; 1620 ret = ceph_osdc_start_request(&fsc->client->osdc, req, false); 1621 if (!ret) { 1622 ret = ceph_osdc_wait_request(&fsc->client->osdc, req); 1623 if (ret == -ENOENT) 1624 ret = 0; 1625 } 1626 ceph_osdc_put_request(req); 1627 1628 out: 1629 return ret; 1630 } 1631 1632 static int ceph_zero_objects(struct inode *inode, loff_t offset, loff_t length) 1633 { 1634 int ret = 0; 1635 struct ceph_inode_info *ci = ceph_inode(inode); 1636 s32 stripe_unit = ci->i_layout.stripe_unit; 1637 s32 stripe_count = ci->i_layout.stripe_count; 1638 s32 object_size = ci->i_layout.object_size; 1639 u64 object_set_size = object_size * stripe_count; 1640 u64 nearly, t; 1641 1642 /* round offset up to next period boundary */ 1643 nearly = offset + object_set_size - 1; 1644 t = nearly; 1645 nearly -= do_div(t, object_set_size); 1646 1647 while (length && offset < nearly) { 1648 loff_t size = length; 1649 ret = ceph_zero_partial_object(inode, offset, &size); 1650 if (ret < 0) 1651 return ret; 1652 offset += size; 1653 length -= size; 1654 } 1655 while (length >= object_set_size) { 1656 int i; 1657 loff_t pos = offset; 1658 for (i = 0; i < stripe_count; ++i) { 1659 ret = ceph_zero_partial_object(inode, pos, NULL); 1660 if (ret < 0) 1661 return ret; 1662 pos += stripe_unit; 1663 } 1664 offset += object_set_size; 1665 length -= object_set_size; 1666 } 1667 while (length) { 1668 loff_t size = length; 1669 ret = ceph_zero_partial_object(inode, offset, &size); 1670 if (ret < 0) 1671 return ret; 1672 offset += size; 1673 length -= size; 1674 } 1675 return ret; 1676 } 1677 1678 static long ceph_fallocate(struct file *file, int mode, 1679 loff_t offset, loff_t length) 1680 { 1681 struct ceph_file_info *fi = file->private_data; 1682 struct inode *inode = file_inode(file); 1683 struct ceph_inode_info *ci = ceph_inode(inode); 1684 struct ceph_osd_client *osdc = 1685 &ceph_inode_to_client(inode)->client->osdc; 1686 struct ceph_cap_flush *prealloc_cf; 1687 int want, got = 0; 1688 int dirty; 1689 int ret = 0; 1690 loff_t endoff = 0; 1691 loff_t size; 1692 1693 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) 1694 return -EOPNOTSUPP; 1695 1696 if (!S_ISREG(inode->i_mode)) 1697 return -EOPNOTSUPP; 1698 1699 prealloc_cf = ceph_alloc_cap_flush(); 1700 if (!prealloc_cf) 1701 return -ENOMEM; 1702 1703 inode_lock(inode); 1704 1705 if (ceph_snap(inode) != CEPH_NOSNAP) { 1706 ret = -EROFS; 1707 goto unlock; 1708 } 1709 1710 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) && 1711 !(mode & FALLOC_FL_PUNCH_HOLE)) { 1712 ret = -ENOSPC; 1713 goto unlock; 1714 } 1715 1716 if (ci->i_inline_version != CEPH_INLINE_NONE) { 1717 ret = ceph_uninline_data(file, NULL); 1718 if (ret < 0) 1719 goto unlock; 1720 } 1721 1722 size = i_size_read(inode); 1723 if (!(mode & FALLOC_FL_KEEP_SIZE)) 1724 endoff = offset + length; 1725 1726 if (fi->fmode & CEPH_FILE_MODE_LAZY) 1727 want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO; 1728 else 1729 want = CEPH_CAP_FILE_BUFFER; 1730 1731 ret = ceph_get_caps(ci, CEPH_CAP_FILE_WR, want, endoff, &got, NULL); 1732 if (ret < 0) 1733 goto unlock; 1734 1735 if (mode & FALLOC_FL_PUNCH_HOLE) { 1736 if (offset < size) 1737 ceph_zero_pagecache_range(inode, offset, length); 1738 ret = ceph_zero_objects(inode, offset, length); 1739 } else if (endoff > size) { 1740 truncate_pagecache_range(inode, size, -1); 1741 if (ceph_inode_set_size(inode, endoff)) 1742 ceph_check_caps(ceph_inode(inode), 1743 CHECK_CAPS_AUTHONLY, NULL); 1744 } 1745 1746 if (!ret) { 1747 spin_lock(&ci->i_ceph_lock); 1748 ci->i_inline_version = CEPH_INLINE_NONE; 1749 dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, 1750 &prealloc_cf); 1751 spin_unlock(&ci->i_ceph_lock); 1752 if (dirty) 1753 __mark_inode_dirty(inode, dirty); 1754 } 1755 1756 ceph_put_cap_refs(ci, got); 1757 unlock: 1758 inode_unlock(inode); 1759 ceph_free_cap_flush(prealloc_cf); 1760 return ret; 1761 } 1762 1763 const struct file_operations ceph_file_fops = { 1764 .open = ceph_open, 1765 .release = ceph_release, 1766 .llseek = ceph_llseek, 1767 .read_iter = ceph_read_iter, 1768 .write_iter = ceph_write_iter, 1769 .mmap = ceph_mmap, 1770 .fsync = ceph_fsync, 1771 .lock = ceph_lock, 1772 .flock = ceph_flock, 1773 .splice_read = generic_file_splice_read, 1774 .splice_write = iter_file_splice_write, 1775 .unlocked_ioctl = ceph_ioctl, 1776 .compat_ioctl = ceph_ioctl, 1777 .fallocate = ceph_fallocate, 1778 }; 1779 1780