fs/ceph/file.c

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