fs/fuse/file.c

   1 /*
   2   FUSE: Filesystem in Userspace
   3   Copyright (C) 2001-2008  Miklos Szeredi <[email protected]>
   4
   5   This program can be distributed under the terms of the GNU GPL.
   6   See the file COPYING.
   7 */
   8
   9 #include "fuse_i.h"
  10
  11 #include <linux/pagemap.h>
  12 #include <linux/slab.h>
  13 #include <linux/kernel.h>
  14 #include <linux/sched.h>
  15 #include <linux/sched/signal.h>
  16 #include <linux/module.h>
  17 #include <linux/swap.h>
  18 #include <linux/falloc.h>
  19 #include <linux/uio.h>
  20 #include <linux/fs.h>
  21 #include <linux/filelock.h>
  22 #include <linux/splice.h>
  23 #include <linux/task_io_accounting_ops.h>
  24
  25 static int fuse_send_open(struct fuse_mount *fm, u64 nodeid,
  26                           unsigned int open_flags, int opcode,
  27                           struct fuse_open_out *outargp)
  28 {
  29         struct fuse_open_in inarg;
  30         FUSE_ARGS(args);
  31
  32         memset(&inarg, 0, sizeof(inarg));
  33         inarg.flags = open_flags & ~(O_CREAT | O_EXCL | O_NOCTTY);
  34         if (!fm->fc->atomic_o_trunc)
  35                 inarg.flags &= ~O_TRUNC;
  36
  37         if (fm->fc->handle_killpriv_v2 &&
  38             (inarg.flags & O_TRUNC) && !capable(CAP_FSETID)) {
  39                 inarg.open_flags |= FUSE_OPEN_KILL_SUIDGID;
  40         }
  41
  42         args.opcode = opcode;
  43         args.nodeid = nodeid;
  44         args.in_numargs = 1;
  45         args.in_args[0].size = sizeof(inarg);
  46         args.in_args[0].value = &inarg;
  47         args.out_numargs = 1;
  48         args.out_args[0].size = sizeof(*outargp);
  49         args.out_args[0].value = outargp;
  50
  51         return fuse_simple_request(fm, &args);
  52 }
  53
  54 struct fuse_file *fuse_file_alloc(struct fuse_mount *fm, bool release)
  55 {
  56         struct fuse_file *ff;
  57
  58         ff = kzalloc(sizeof(struct fuse_file), GFP_KERNEL_ACCOUNT);
  59         if (unlikely(!ff))
  60                 return NULL;
  61
  62         ff->fm = fm;
  63         if (release) {
  64                 ff->args = kzalloc(sizeof(*ff->args), GFP_KERNEL_ACCOUNT);
  65                 if (!ff->args) {
  66                         kfree(ff);
  67                         return NULL;
  68                 }
  69         }
  70
  71         INIT_LIST_HEAD(&ff->write_entry);
  72         refcount_set(&ff->count, 1);
  73         RB_CLEAR_NODE(&ff->polled_node);
  74         init_waitqueue_head(&ff->poll_wait);
  75
  76         ff->kh = atomic64_inc_return(&fm->fc->khctr);
  77
  78         return ff;
  79 }
  80
  81 void fuse_file_free(struct fuse_file *ff)
  82 {
  83         kfree(ff->args);
  84         kfree(ff);
  85 }
  86
  87 static struct fuse_file *fuse_file_get(struct fuse_file *ff)
  88 {
  89         refcount_inc(&ff->count);
  90         return ff;
  91 }
  92
  93 static void fuse_release_end(struct fuse_mount *fm, struct fuse_args *args,
  94                              int error)
  95 {
  96         struct fuse_release_args *ra = container_of(args, typeof(*ra), args);
  97
  98         iput(ra->inode);
  99         kfree(ra);
 100 }
 101
 102 static void fuse_file_put(struct fuse_file *ff, bool sync)
 103 {
 104         if (refcount_dec_and_test(&ff->count)) {
 105                 struct fuse_release_args *ra = &ff->args->release_args;
 106                 struct fuse_args *args = (ra ? &ra->args : NULL);
 107
 108                 if (ra && ra->inode)
 109                         fuse_file_io_release(ff, ra->inode);
 110
 111                 if (!args) {
 112                         /* Do nothing when server does not implement 'open' */
 113                 } else if (sync) {
 114                         fuse_simple_request(ff->fm, args);
 115                         fuse_release_end(ff->fm, args, 0);
 116                 } else {
 117                         args->end = fuse_release_end;
 118                         if (fuse_simple_background(ff->fm, args,
 119                                                    GFP_KERNEL | __GFP_NOFAIL))
 120                                 fuse_release_end(ff->fm, args, -ENOTCONN);
 121                 }
 122                 kfree(ff);
 123         }
 124 }
 125
 126 struct fuse_file *fuse_file_open(struct fuse_mount *fm, u64 nodeid,
 127                                  unsigned int open_flags, bool isdir)
 128 {
 129         struct fuse_conn *fc = fm->fc;
 130         struct fuse_file *ff;
 131         int opcode = isdir ? FUSE_OPENDIR : FUSE_OPEN;
 132         bool open = isdir ? !fc->no_opendir : !fc->no_open;
 133
 134         ff = fuse_file_alloc(fm, open);
 135         if (!ff)
 136                 return ERR_PTR(-ENOMEM);
 137
 138         ff->fh = 0;
 139         /* Default for no-open */
 140         ff->open_flags = FOPEN_KEEP_CACHE | (isdir ? FOPEN_CACHE_DIR : 0);
 141         if (open) {
 142                 /* Store outarg for fuse_finish_open() */
 143                 struct fuse_open_out *outargp = &ff->args->open_outarg;
 144                 int err;
 145
 146                 err = fuse_send_open(fm, nodeid, open_flags, opcode, outargp);
 147                 if (!err) {
 148                         ff->fh = outargp->fh;
 149                         ff->open_flags = outargp->open_flags;
 150                 } else if (err != -ENOSYS) {
 151                         fuse_file_free(ff);
 152                         return ERR_PTR(err);
 153                 } else {
 154                         /* No release needed */
 155                         kfree(ff->args);
 156                         ff->args = NULL;
 157                         if (isdir)
 158                                 fc->no_opendir = 1;
 159                         else
 160                                 fc->no_open = 1;
 161                 }
 162         }
 163
 164         if (isdir)
 165                 ff->open_flags &= ~FOPEN_DIRECT_IO;
 166
 167         ff->nodeid = nodeid;
 168
 169         return ff;
 170 }
 171
 172 int fuse_do_open(struct fuse_mount *fm, u64 nodeid, struct file *file,
 173                  bool isdir)
 174 {
 175         struct fuse_file *ff = fuse_file_open(fm, nodeid, file->f_flags, isdir);
 176
 177         if (!IS_ERR(ff))
 178                 file->private_data = ff;
 179
 180         return PTR_ERR_OR_ZERO(ff);
 181 }
 182 EXPORT_SYMBOL_GPL(fuse_do_open);
 183
 184 static void fuse_link_write_file(struct file *file)
 185 {
 186         struct inode *inode = file_inode(file);
 187         struct fuse_inode *fi = get_fuse_inode(inode);
 188         struct fuse_file *ff = file->private_data;
 189         /*
 190          * file may be written through mmap, so chain it onto the
 191          * inodes's write_file list
 192          */
 193         spin_lock(&fi->lock);
 194         if (list_empty(&ff->write_entry))
 195                 list_add(&ff->write_entry, &fi->write_files);
 196         spin_unlock(&fi->lock);
 197 }
 198
 199 int fuse_finish_open(struct inode *inode, struct file *file)
 200 {
 201         struct fuse_file *ff = file->private_data;
 202         struct fuse_conn *fc = get_fuse_conn(inode);
 203         int err;
 204
 205         err = fuse_file_io_open(file, inode);
 206         if (err)
 207                 return err;
 208
 209         if (ff->open_flags & FOPEN_STREAM)
 210                 stream_open(inode, file);
 211         else if (ff->open_flags & FOPEN_NONSEEKABLE)
 212                 nonseekable_open(inode, file);
 213
 214         if ((file->f_mode & FMODE_WRITE) && fc->writeback_cache)
 215                 fuse_link_write_file(file);
 216
 217         return 0;
 218 }
 219
 220 static void fuse_truncate_update_attr(struct inode *inode, struct file *file)
 221 {
 222         struct fuse_conn *fc = get_fuse_conn(inode);
 223         struct fuse_inode *fi = get_fuse_inode(inode);
 224
 225         spin_lock(&fi->lock);
 226         fi->attr_version = atomic64_inc_return(&fc->attr_version);
 227         i_size_write(inode, 0);
 228         spin_unlock(&fi->lock);
 229         file_update_time(file);
 230         fuse_invalidate_attr_mask(inode, FUSE_STATX_MODSIZE);
 231 }
 232
 233 static int fuse_open(struct inode *inode, struct file *file)
 234 {
 235         struct fuse_mount *fm = get_fuse_mount(inode);
 236         struct fuse_inode *fi = get_fuse_inode(inode);
 237         struct fuse_conn *fc = fm->fc;
 238         struct fuse_file *ff;
 239         int err;
 240         bool is_truncate = (file->f_flags & O_TRUNC) && fc->atomic_o_trunc;
 241         bool is_wb_truncate = is_truncate && fc->writeback_cache;
 242         bool dax_truncate = is_truncate && FUSE_IS_DAX(inode);
 243
 244         if (fuse_is_bad(inode))
 245                 return -EIO;
 246
 247         err = generic_file_open(inode, file);
 248         if (err)
 249                 return err;
 250
 251         if (is_wb_truncate || dax_truncate)
 252                 inode_lock(inode);
 253
 254         if (dax_truncate) {
 255                 filemap_invalidate_lock(inode->i_mapping);
 256                 err = fuse_dax_break_layouts(inode, 0, 0);
 257                 if (err)
 258                         goto out_inode_unlock;
 259         }
 260
 261         if (is_wb_truncate || dax_truncate)
 262                 fuse_set_nowrite(inode);
 263
 264         err = fuse_do_open(fm, get_node_id(inode), file, false);
 265         if (!err) {
 266                 ff = file->private_data;
 267                 err = fuse_finish_open(inode, file);
 268                 if (err)
 269                         fuse_sync_release(fi, ff, file->f_flags);
 270                 else if (is_truncate)
 271                         fuse_truncate_update_attr(inode, file);
 272         }
 273
 274         if (is_wb_truncate || dax_truncate)
 275                 fuse_release_nowrite(inode);
 276         if (!err) {
 277                 if (is_truncate)
 278                         truncate_pagecache(inode, 0);
 279                 else if (!(ff->open_flags & FOPEN_KEEP_CACHE))
 280                         invalidate_inode_pages2(inode->i_mapping);
 281         }
 282         if (dax_truncate)
 283                 filemap_invalidate_unlock(inode->i_mapping);
 284 out_inode_unlock:
 285         if (is_wb_truncate || dax_truncate)
 286                 inode_unlock(inode);
 287
 288         return err;
 289 }
 290
 291 static void fuse_prepare_release(struct fuse_inode *fi, struct fuse_file *ff,
 292                                  unsigned int flags, int opcode, bool sync)
 293 {
 294         struct fuse_conn *fc = ff->fm->fc;
 295         struct fuse_release_args *ra = &ff->args->release_args;
 296
 297         if (fuse_file_passthrough(ff))
 298                 fuse_passthrough_release(ff, fuse_inode_backing(fi));
 299
 300         /* Inode is NULL on error path of fuse_create_open() */
 301         if (likely(fi)) {
 302                 spin_lock(&fi->lock);
 303                 list_del(&ff->write_entry);
 304                 spin_unlock(&fi->lock);
 305         }
 306         spin_lock(&fc->lock);
 307         if (!RB_EMPTY_NODE(&ff->polled_node))
 308                 rb_erase(&ff->polled_node, &fc->polled_files);
 309         spin_unlock(&fc->lock);
 310
 311         wake_up_interruptible_all(&ff->poll_wait);
 312
 313         if (!ra)
 314                 return;
 315
 316         /* ff->args was used for open outarg */
 317         memset(ff->args, 0, sizeof(*ff->args));
 318         ra->inarg.fh = ff->fh;
 319         ra->inarg.flags = flags;
 320         ra->args.in_numargs = 1;
 321         ra->args.in_args[0].size = sizeof(struct fuse_release_in);
 322         ra->args.in_args[0].value = &ra->inarg;
 323         ra->args.opcode = opcode;
 324         ra->args.nodeid = ff->nodeid;
 325         ra->args.force = true;
 326         ra->args.nocreds = true;
 327
 328         /*
 329          * Hold inode until release is finished.
 330          * From fuse_sync_release() the refcount is 1 and everything's
 331          * synchronous, so we are fine with not doing igrab() here.
 332          */
 333         ra->inode = sync ? NULL : igrab(&fi->inode);
 334 }
 335
 336 void fuse_file_release(struct inode *inode, struct fuse_file *ff,
 337                        unsigned int open_flags, fl_owner_t id, bool isdir)
 338 {
 339         struct fuse_inode *fi = get_fuse_inode(inode);
 340         struct fuse_release_args *ra = &ff->args->release_args;
 341         int opcode = isdir ? FUSE_RELEASEDIR : FUSE_RELEASE;
 342
 343         fuse_prepare_release(fi, ff, open_flags, opcode, false);
 344
 345         if (ra && ff->flock) {
 346                 ra->inarg.release_flags |= FUSE_RELEASE_FLOCK_UNLOCK;
 347                 ra->inarg.lock_owner = fuse_lock_owner_id(ff->fm->fc, id);
 348         }
 349
 350         /*
 351          * Normally this will send the RELEASE request, however if
 352          * some asynchronous READ or WRITE requests are outstanding,
 353          * the sending will be delayed.
 354          *
 355          * Make the release synchronous if this is a fuseblk mount,
 356          * synchronous RELEASE is allowed (and desirable) in this case
 357          * because the server can be trusted not to screw up.
 358          */
 359         fuse_file_put(ff, ff->fm->fc->destroy);
 360 }
 361
 362 void fuse_release_common(struct file *file, bool isdir)
 363 {
 364         fuse_file_release(file_inode(file), file->private_data, file->f_flags,
 365                           (fl_owner_t) file, isdir);
 366 }
 367
 368 static int fuse_release(struct inode *inode, struct file *file)
 369 {
 370         struct fuse_conn *fc = get_fuse_conn(inode);
 371
 372         /*
 373          * Dirty pages might remain despite write_inode_now() call from
 374          * fuse_flush() due to writes racing with the close.
 375          */
 376         if (fc->writeback_cache)
 377                 write_inode_now(inode, 1);
 378
 379         fuse_release_common(file, false);
 380
 381         /* return value is ignored by VFS */
 382         return 0;
 383 }
 384
 385 void fuse_sync_release(struct fuse_inode *fi, struct fuse_file *ff,
 386                        unsigned int flags)
 387 {
 388         WARN_ON(refcount_read(&ff->count) > 1);
 389         fuse_prepare_release(fi, ff, flags, FUSE_RELEASE, true);
 390         fuse_file_put(ff, true);
 391 }
 392 EXPORT_SYMBOL_GPL(fuse_sync_release);
 393
 394 /*
 395  * Scramble the ID space with XTEA, so that the value of the files_struct
 396  * pointer is not exposed to userspace.
 397  */
 398 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id)
 399 {
 400         u32 *k = fc->scramble_key;
 401         u64 v = (unsigned long) id;
 402         u32 v0 = v;
 403         u32 v1 = v >> 32;
 404         u32 sum = 0;
 405         int i;
 406
 407         for (i = 0; i < 32; i++) {
 408                 v0 += ((v1 << 4 ^ v1 >> 5) + v1) ^ (sum + k[sum & 3]);
 409                 sum += 0x9E3779B9;
 410                 v1 += ((v0 << 4 ^ v0 >> 5) + v0) ^ (sum + k[sum>>11 & 3]);
 411         }
 412
 413         return (u64) v0 + ((u64) v1 << 32);
 414 }
 415
 416 struct fuse_writepage_args {
 417         struct fuse_io_args ia;
 418         struct rb_node writepages_entry;
 419         struct list_head queue_entry;
 420         struct fuse_writepage_args *next;
 421         struct inode *inode;
 422         struct fuse_sync_bucket *bucket;
 423 };
 424
 425 static struct fuse_writepage_args *fuse_find_writeback(struct fuse_inode *fi,
 426                                             pgoff_t idx_from, pgoff_t idx_to)
 427 {
 428         struct rb_node *n;
 429
 430         n = fi->writepages.rb_node;
 431
 432         while (n) {
 433                 struct fuse_writepage_args *wpa;
 434                 pgoff_t curr_index;
 435
 436                 wpa = rb_entry(n, struct fuse_writepage_args, writepages_entry);
 437                 WARN_ON(get_fuse_inode(wpa->inode) != fi);
 438                 curr_index = wpa->ia.write.in.offset >> PAGE_SHIFT;
 439                 if (idx_from >= curr_index + wpa->ia.ap.num_pages)
 440                         n = n->rb_right;
 441                 else if (idx_to < curr_index)
 442                         n = n->rb_left;
 443                 else
 444                         return wpa;
 445         }
 446         return NULL;
 447 }
 448
 449 /*
 450  * Check if any page in a range is under writeback
 451  *
 452  * This is currently done by walking the list of writepage requests
 453  * for the inode, which can be pretty inefficient.
 454  */
 455 static bool fuse_range_is_writeback(struct inode *inode, pgoff_t idx_from,
 456                                    pgoff_t idx_to)
 457 {
 458         struct fuse_inode *fi = get_fuse_inode(inode);
 459         bool found;
 460
 461         spin_lock(&fi->lock);
 462         found = fuse_find_writeback(fi, idx_from, idx_to);
 463         spin_unlock(&fi->lock);
 464
 465         return found;
 466 }
 467
 468 static inline bool fuse_page_is_writeback(struct inode *inode, pgoff_t index)
 469 {
 470         return fuse_range_is_writeback(inode, index, index);
 471 }
 472
 473 /*
 474  * Wait for page writeback to be completed.
 475  *
 476  * Since fuse doesn't rely on the VM writeback tracking, this has to
 477  * use some other means.
 478  */
 479 static void fuse_wait_on_page_writeback(struct inode *inode, pgoff_t index)
 480 {
 481         struct fuse_inode *fi = get_fuse_inode(inode);
 482
 483         wait_event(fi->page_waitq, !fuse_page_is_writeback(inode, index));
 484 }
 485
 486 /*
 487  * Wait for all pending writepages on the inode to finish.
 488  *
 489  * This is currently done by blocking further writes with FUSE_NOWRITE
 490  * and waiting for all sent writes to complete.
 491  *
 492  * This must be called under i_mutex, otherwise the FUSE_NOWRITE usage
 493  * could conflict with truncation.
 494  */
 495 static void fuse_sync_writes(struct inode *inode)
 496 {
 497         fuse_set_nowrite(inode);
 498         fuse_release_nowrite(inode);
 499 }
 500
 501 static int fuse_flush(struct file *file, fl_owner_t id)
 502 {
 503         struct inode *inode = file_inode(file);
 504         struct fuse_mount *fm = get_fuse_mount(inode);
 505         struct fuse_file *ff = file->private_data;
 506         struct fuse_flush_in inarg;
 507         FUSE_ARGS(args);
 508         int err;
 509
 510         if (fuse_is_bad(inode))
 511                 return -EIO;
 512
 513         if (ff->open_flags & FOPEN_NOFLUSH && !fm->fc->writeback_cache)
 514                 return 0;
 515
 516         err = write_inode_now(inode, 1);
 517         if (err)
 518                 return err;
 519
 520         inode_lock(inode);
 521         fuse_sync_writes(inode);
 522         inode_unlock(inode);
 523
 524         err = filemap_check_errors(file->f_mapping);
 525         if (err)
 526                 return err;
 527
 528         err = 0;
 529         if (fm->fc->no_flush)
 530                 goto inval_attr_out;
 531
 532         memset(&inarg, 0, sizeof(inarg));
 533         inarg.fh = ff->fh;
 534         inarg.lock_owner = fuse_lock_owner_id(fm->fc, id);
 535         args.opcode = FUSE_FLUSH;
 536         args.nodeid = get_node_id(inode);
 537         args.in_numargs = 1;
 538         args.in_args[0].size = sizeof(inarg);
 539         args.in_args[0].value = &inarg;
 540         args.force = true;
 541
 542         err = fuse_simple_request(fm, &args);
 543         if (err == -ENOSYS) {
 544                 fm->fc->no_flush = 1;
 545                 err = 0;
 546         }
 547
 548 inval_attr_out:
 549         /*
 550          * In memory i_blocks is not maintained by fuse, if writeback cache is
 551          * enabled, i_blocks from cached attr may not be accurate.
 552          */
 553         if (!err && fm->fc->writeback_cache)
 554                 fuse_invalidate_attr_mask(inode, STATX_BLOCKS);
 555         return err;
 556 }
 557
 558 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
 559                       int datasync, int opcode)
 560 {
 561         struct inode *inode = file->f_mapping->host;
 562         struct fuse_mount *fm = get_fuse_mount(inode);
 563         struct fuse_file *ff = file->private_data;
 564         FUSE_ARGS(args);
 565         struct fuse_fsync_in inarg;
 566
 567         memset(&inarg, 0, sizeof(inarg));
 568         inarg.fh = ff->fh;
 569         inarg.fsync_flags = datasync ? FUSE_FSYNC_FDATASYNC : 0;
 570         args.opcode = opcode;
 571         args.nodeid = get_node_id(inode);
 572         args.in_numargs = 1;
 573         args.in_args[0].size = sizeof(inarg);
 574         args.in_args[0].value = &inarg;
 575         return fuse_simple_request(fm, &args);
 576 }
 577
 578 static int fuse_fsync(struct file *file, loff_t start, loff_t end,
 579                       int datasync)
 580 {
 581         struct inode *inode = file->f_mapping->host;
 582         struct fuse_conn *fc = get_fuse_conn(inode);
 583         int err;
 584
 585         if (fuse_is_bad(inode))
 586                 return -EIO;
 587
 588         inode_lock(inode);
 589
 590         /*
 591          * Start writeback against all dirty pages of the inode, then
 592          * wait for all outstanding writes, before sending the FSYNC
 593          * request.
 594          */
 595         err = file_write_and_wait_range(file, start, end);
 596         if (err)
 597                 goto out;
 598
 599         fuse_sync_writes(inode);
 600
 601         /*
 602          * Due to implementation of fuse writeback
 603          * file_write_and_wait_range() does not catch errors.
 604          * We have to do this directly after fuse_sync_writes()
 605          */
 606         err = file_check_and_advance_wb_err(file);
 607         if (err)
 608                 goto out;
 609
 610         err = sync_inode_metadata(inode, 1);
 611         if (err)
 612                 goto out;
 613
 614         if (fc->no_fsync)
 615                 goto out;
 616
 617         err = fuse_fsync_common(file, start, end, datasync, FUSE_FSYNC);
 618         if (err == -ENOSYS) {
 619                 fc->no_fsync = 1;
 620                 err = 0;
 621         }
 622 out:
 623         inode_unlock(inode);
 624
 625         return err;
 626 }
 627
 628 void fuse_read_args_fill(struct fuse_io_args *ia, struct file *file, loff_t pos,
 629                          size_t count, int opcode)
 630 {
 631         struct fuse_file *ff = file->private_data;
 632         struct fuse_args *args = &ia->ap.args;
 633
 634         ia->read.in.fh = ff->fh;
 635         ia->read.in.offset = pos;
 636         ia->read.in.size = count;
 637         ia->read.in.flags = file->f_flags;
 638         args->opcode = opcode;
 639         args->nodeid = ff->nodeid;
 640         args->in_numargs = 1;
 641         args->in_args[0].size = sizeof(ia->read.in);
 642         args->in_args[0].value = &ia->read.in;
 643         args->out_argvar = true;
 644         args->out_numargs = 1;
 645         args->out_args[0].size = count;
 646 }
 647
 648 static void fuse_release_user_pages(struct fuse_args_pages *ap,
 649                                     bool should_dirty)
 650 {
 651         unsigned int i;
 652
 653         for (i = 0; i < ap->num_pages; i++) {
 654                 if (should_dirty)
 655                         set_page_dirty_lock(ap->pages[i]);
 656                 if (ap->args.is_pinned)
 657                         unpin_user_page(ap->pages[i]);
 658         }
 659 }
 660
 661 static void fuse_io_release(struct kref *kref)
 662 {
 663         kfree(container_of(kref, struct fuse_io_priv, refcnt));
 664 }
 665
 666 static ssize_t fuse_get_res_by_io(struct fuse_io_priv *io)
 667 {
 668         if (io->err)
 669                 return io->err;
 670
 671         if (io->bytes >= 0 && io->write)
 672                 return -EIO;
 673
 674         return io->bytes < 0 ? io->size : io->bytes;
 675 }
 676
 677 /*
 678  * In case of short read, the caller sets 'pos' to the position of
 679  * actual end of fuse request in IO request. Otherwise, if bytes_requested
 680  * == bytes_transferred or rw == WRITE, the caller sets 'pos' to -1.
 681  *
 682  * An example:
 683  * User requested DIO read of 64K. It was split into two 32K fuse requests,
 684  * both submitted asynchronously. The first of them was ACKed by userspace as
 685  * fully completed (req->out.args[0].size == 32K) resulting in pos == -1. The
 686  * second request was ACKed as short, e.g. only 1K was read, resulting in
 687  * pos == 33K.
 688  *
 689  * Thus, when all fuse requests are completed, the minimal non-negative 'pos'
 690  * will be equal to the length of the longest contiguous fragment of
 691  * transferred data starting from the beginning of IO request.
 692  */
 693 static void fuse_aio_complete(struct fuse_io_priv *io, int err, ssize_t pos)
 694 {
 695         int left;
 696
 697         spin_lock(&io->lock);
 698         if (err)
 699                 io->err = io->err ? : err;
 700         else if (pos >= 0 && (io->bytes < 0 || pos < io->bytes))
 701                 io->bytes = pos;
 702
 703         left = --io->reqs;
 704         if (!left && io->blocking)
 705                 complete(io->done);
 706         spin_unlock(&io->lock);
 707
 708         if (!left && !io->blocking) {
 709                 ssize_t res = fuse_get_res_by_io(io);
 710
 711                 if (res >= 0) {
 712                         struct inode *inode = file_inode(io->iocb->ki_filp);
 713                         struct fuse_conn *fc = get_fuse_conn(inode);
 714                         struct fuse_inode *fi = get_fuse_inode(inode);
 715
 716                         spin_lock(&fi->lock);
 717                         fi->attr_version = atomic64_inc_return(&fc->attr_version);
 718                         spin_unlock(&fi->lock);
 719                 }
 720
 721                 io->iocb->ki_complete(io->iocb, res);
 722         }
 723
 724         kref_put(&io->refcnt, fuse_io_release);
 725 }
 726
 727 static struct fuse_io_args *fuse_io_alloc(struct fuse_io_priv *io,
 728                                           unsigned int npages)
 729 {
 730         struct fuse_io_args *ia;
 731
 732         ia = kzalloc(sizeof(*ia), GFP_KERNEL);
 733         if (ia) {
 734                 ia->io = io;
 735                 ia->ap.pages = fuse_pages_alloc(npages, GFP_KERNEL,
 736                                                 &ia->ap.descs);
 737                 if (!ia->ap.pages) {
 738                         kfree(ia);
 739                         ia = NULL;
 740                 }
 741         }
 742         return ia;
 743 }
 744
 745 static void fuse_io_free(struct fuse_io_args *ia)
 746 {
 747         kfree(ia->ap.pages);
 748         kfree(ia);
 749 }
 750
 751 static void fuse_aio_complete_req(struct fuse_mount *fm, struct fuse_args *args,
 752                                   int err)
 753 {
 754         struct fuse_io_args *ia = container_of(args, typeof(*ia), ap.args);
 755         struct fuse_io_priv *io = ia->io;
 756         ssize_t pos = -1;
 757
 758         fuse_release_user_pages(&ia->ap, io->should_dirty);
 759
 760         if (err) {
 761                 /* Nothing */
 762         } else if (io->write) {
 763                 if (ia->write.out.size > ia->write.in.size) {
 764                         err = -EIO;
 765                 } else if (ia->write.in.size != ia->write.out.size) {
 766                         pos = ia->write.in.offset - io->offset +
 767                                 ia->write.out.size;
 768                 }
 769         } else {
 770                 u32 outsize = args->out_args[0].size;
 771
 772                 if (ia->read.in.size != outsize)
 773                         pos = ia->read.in.offset - io->offset + outsize;
 774         }
 775
 776         fuse_aio_complete(io, err, pos);
 777         fuse_io_free(ia);
 778 }
 779
 780 static ssize_t fuse_async_req_send(struct fuse_mount *fm,
 781                                    struct fuse_io_args *ia, size_t num_bytes)
 782 {
 783         ssize_t err;
 784         struct fuse_io_priv *io = ia->io;
 785
 786         spin_lock(&io->lock);
 787         kref_get(&io->refcnt);
 788         io->size += num_bytes;
 789         io->reqs++;
 790         spin_unlock(&io->lock);
 791
 792         ia->ap.args.end = fuse_aio_complete_req;
 793         ia->ap.args.may_block = io->should_dirty;
 794         err = fuse_simple_background(fm, &ia->ap.args, GFP_KERNEL);
 795         if (err)
 796                 fuse_aio_complete_req(fm, &ia->ap.args, err);
 797
 798         return num_bytes;
 799 }
 800
 801 static ssize_t fuse_send_read(struct fuse_io_args *ia, loff_t pos, size_t count,
 802                               fl_owner_t owner)
 803 {
 804         struct file *file = ia->io->iocb->ki_filp;
 805         struct fuse_file *ff = file->private_data;
 806         struct fuse_mount *fm = ff->fm;
 807
 808         fuse_read_args_fill(ia, file, pos, count, FUSE_READ);
 809         if (owner != NULL) {
 810                 ia->read.in.read_flags |= FUSE_READ_LOCKOWNER;
 811                 ia->read.in.lock_owner = fuse_lock_owner_id(fm->fc, owner);
 812         }
 813
 814         if (ia->io->async)
 815                 return fuse_async_req_send(fm, ia, count);
 816
 817         return fuse_simple_request(fm, &ia->ap.args);
 818 }
 819
 820 static void fuse_read_update_size(struct inode *inode, loff_t size,
 821                                   u64 attr_ver)
 822 {
 823         struct fuse_conn *fc = get_fuse_conn(inode);
 824         struct fuse_inode *fi = get_fuse_inode(inode);
 825
 826         spin_lock(&fi->lock);
 827         if (attr_ver >= fi->attr_version && size < inode->i_size &&
 828             !test_bit(FUSE_I_SIZE_UNSTABLE, &fi->state)) {
 829                 fi->attr_version = atomic64_inc_return(&fc->attr_version);
 830                 i_size_write(inode, size);
 831         }
 832         spin_unlock(&fi->lock);
 833 }
 834
 835 static void fuse_short_read(struct inode *inode, u64 attr_ver, size_t num_read,
 836                             struct fuse_args_pages *ap)
 837 {
 838         struct fuse_conn *fc = get_fuse_conn(inode);
 839
 840         /*
 841          * If writeback_cache is enabled, a short read means there's a hole in
 842          * the file.  Some data after the hole is in page cache, but has not
 843          * reached the client fs yet.  So the hole is not present there.
 844          */
 845         if (!fc->writeback_cache) {
 846                 loff_t pos = page_offset(ap->pages[0]) + num_read;
 847                 fuse_read_update_size(inode, pos, attr_ver);
 848         }
 849 }
 850
 851 static int fuse_do_readpage(struct file *file, struct page *page)
 852 {
 853         struct inode *inode = page->mapping->host;
 854         struct fuse_mount *fm = get_fuse_mount(inode);
 855         loff_t pos = page_offset(page);
 856         struct fuse_page_desc desc = { .length = PAGE_SIZE };
 857         struct fuse_io_args ia = {
 858                 .ap.args.page_zeroing = true,
 859                 .ap.args.out_pages = true,
 860                 .ap.num_pages = 1,
 861                 .ap.pages = &page,
 862                 .ap.descs = &desc,
 863         };
 864         ssize_t res;
 865         u64 attr_ver;
 866
 867         /*
 868          * Page writeback can extend beyond the lifetime of the
 869          * page-cache page, so make sure we read a properly synced
 870          * page.
 871          */
 872         fuse_wait_on_page_writeback(inode, page->index);
 873
 874         attr_ver = fuse_get_attr_version(fm->fc);
 875
 876         /* Don't overflow end offset */
 877         if (pos + (desc.length - 1) == LLONG_MAX)
 878                 desc.length--;
 879
 880         fuse_read_args_fill(&ia, file, pos, desc.length, FUSE_READ);
 881         res = fuse_simple_request(fm, &ia.ap.args);
 882         if (res < 0)
 883                 return res;
 884         /*
 885          * Short read means EOF.  If file size is larger, truncate it
 886          */
 887         if (res < desc.length)
 888                 fuse_short_read(inode, attr_ver, res, &ia.ap);
 889
 890         SetPageUptodate(page);
 891
 892         return 0;
 893 }
 894
 895 static int fuse_read_folio(struct file *file, struct folio *folio)
 896 {
 897         struct page *page = &folio->page;
 898         struct inode *inode = page->mapping->host;
 899         int err;
 900
 901         err = -EIO;
 902         if (fuse_is_bad(inode))
 903                 goto out;
 904
 905         err = fuse_do_readpage(file, page);
 906         fuse_invalidate_atime(inode);
 907  out:
 908         unlock_page(page);
 909         return err;
 910 }
 911
 912 static void fuse_readpages_end(struct fuse_mount *fm, struct fuse_args *args,
 913                                int err)
 914 {
 915         int i;
 916         struct fuse_io_args *ia = container_of(args, typeof(*ia), ap.args);
 917         struct fuse_args_pages *ap = &ia->ap;
 918         size_t count = ia->read.in.size;
 919         size_t num_read = args->out_args[0].size;
 920         struct address_space *mapping = NULL;
 921
 922         for (i = 0; mapping == NULL && i < ap->num_pages; i++)
 923                 mapping = ap->pages[i]->mapping;
 924
 925         if (mapping) {
 926                 struct inode *inode = mapping->host;
 927
 928                 /*
 929                  * Short read means EOF. If file size is larger, truncate it
 930                  */
 931                 if (!err && num_read < count)
 932                         fuse_short_read(inode, ia->read.attr_ver, num_read, ap);
 933
 934                 fuse_invalidate_atime(inode);
 935         }
 936
 937         for (i = 0; i < ap->num_pages; i++) {
 938                 struct folio *folio = page_folio(ap->pages[i]);
 939
 940                 folio_end_read(folio, !err);
 941                 folio_put(folio);
 942         }
 943         if (ia->ff)
 944                 fuse_file_put(ia->ff, false);
 945
 946         fuse_io_free(ia);
 947 }
 948
 949 static void fuse_send_readpages(struct fuse_io_args *ia, struct file *file)
 950 {
 951         struct fuse_file *ff = file->private_data;
 952         struct fuse_mount *fm = ff->fm;
 953         struct fuse_args_pages *ap = &ia->ap;
 954         loff_t pos = page_offset(ap->pages[0]);
 955         size_t count = ap->num_pages << PAGE_SHIFT;
 956         ssize_t res;
 957         int err;
 958
 959         ap->args.out_pages = true;
 960         ap->args.page_zeroing = true;
 961         ap->args.page_replace = true;
 962
 963         /* Don't overflow end offset */
 964         if (pos + (count - 1) == LLONG_MAX) {
 965                 count--;
 966                 ap->descs[ap->num_pages - 1].length--;
 967         }
 968         WARN_ON((loff_t) (pos + count) < 0);
 969
 970         fuse_read_args_fill(ia, file, pos, count, FUSE_READ);
 971         ia->read.attr_ver = fuse_get_attr_version(fm->fc);
 972         if (fm->fc->async_read) {
 973                 ia->ff = fuse_file_get(ff);
 974                 ap->args.end = fuse_readpages_end;
 975                 err = fuse_simple_background(fm, &ap->args, GFP_KERNEL);
 976                 if (!err)
 977                         return;
 978         } else {
 979                 res = fuse_simple_request(fm, &ap->args);
 980                 err = res < 0 ? res : 0;
 981         }
 982         fuse_readpages_end(fm, &ap->args, err);
 983 }
 984
 985 static void fuse_readahead(struct readahead_control *rac)
 986 {
 987         struct inode *inode = rac->mapping->host;
 988         struct fuse_conn *fc = get_fuse_conn(inode);
 989         unsigned int i, max_pages, nr_pages = 0;
 990
 991         if (fuse_is_bad(inode))
 992                 return;
 993
 994         max_pages = min_t(unsigned int, fc->max_pages,
 995                         fc->max_read / PAGE_SIZE);
 996
 997         for (;;) {
 998                 struct fuse_io_args *ia;
 999                 struct fuse_args_pages *ap;
1000
1001                 if (fc->num_background >= fc->congestion_threshold &&
1002                     rac->ra->async_size >= readahead_count(rac))
1003                         /*
1004                          * Congested and only async pages left, so skip the
1005                          * rest.
1006                          */
1007                         break;
1008
1009                 nr_pages = readahead_count(rac) - nr_pages;
1010                 if (nr_pages > max_pages)
1011                         nr_pages = max_pages;
1012                 if (nr_pages == 0)
1013                         break;
1014                 ia = fuse_io_alloc(NULL, nr_pages);
1015                 if (!ia)
1016                         return;
1017                 ap = &ia->ap;
1018                 nr_pages = __readahead_batch(rac, ap->pages, nr_pages);
1019                 for (i = 0; i < nr_pages; i++) {
1020                         fuse_wait_on_page_writeback(inode,
1021                                                     readahead_index(rac) + i);
1022                         ap->descs[i].length = PAGE_SIZE;
1023                 }
1024                 ap->num_pages = nr_pages;
1025                 fuse_send_readpages(ia, rac->file);
1026         }
1027 }
1028
1029 static ssize_t fuse_cache_read_iter(struct kiocb *iocb, struct iov_iter *to)
1030 {
1031         struct inode *inode = iocb->ki_filp->f_mapping->host;
1032         struct fuse_conn *fc = get_fuse_conn(inode);
1033
1034         /*
1035          * In auto invalidate mode, always update attributes on read.
1036          * Otherwise, only update if we attempt to read past EOF (to ensure
1037          * i_size is up to date).
1038          */
1039         if (fc->auto_inval_data ||
1040             (iocb->ki_pos + iov_iter_count(to) > i_size_read(inode))) {
1041                 int err;
1042                 err = fuse_update_attributes(inode, iocb->ki_filp, STATX_SIZE);
1043                 if (err)
1044                         return err;
1045         }
1046
1047         return generic_file_read_iter(iocb, to);
1048 }
1049
1050 static void fuse_write_args_fill(struct fuse_io_args *ia, struct fuse_file *ff,
1051                                  loff_t pos, size_t count)
1052 {
1053         struct fuse_args *args = &ia->ap.args;
1054
1055         ia->write.in.fh = ff->fh;
1056         ia->write.in.offset = pos;
1057         ia->write.in.size = count;
1058         args->opcode = FUSE_WRITE;
1059         args->nodeid = ff->nodeid;
1060         args->in_numargs = 2;
1061         if (ff->fm->fc->minor < 9)
1062                 args->in_args[0].size = FUSE_COMPAT_WRITE_IN_SIZE;
1063         else
1064                 args->in_args[0].size = sizeof(ia->write.in);
1065         args->in_args[0].value = &ia->write.in;
1066         args->in_args[1].size = count;
1067         args->out_numargs = 1;
1068         args->out_args[0].size = sizeof(ia->write.out);
1069         args->out_args[0].value = &ia->write.out;
1070 }
1071
1072 static unsigned int fuse_write_flags(struct kiocb *iocb)
1073 {
1074         unsigned int flags = iocb->ki_filp->f_flags;
1075
1076         if (iocb_is_dsync(iocb))
1077                 flags |= O_DSYNC;
1078         if (iocb->ki_flags & IOCB_SYNC)
1079                 flags |= O_SYNC;
1080
1081         return flags;
1082 }
1083
1084 static ssize_t fuse_send_write(struct fuse_io_args *ia, loff_t pos,
1085                                size_t count, fl_owner_t owner)
1086 {
1087         struct kiocb *iocb = ia->io->iocb;
1088         struct file *file = iocb->ki_filp;
1089         struct fuse_file *ff = file->private_data;
1090         struct fuse_mount *fm = ff->fm;
1091         struct fuse_write_in *inarg = &ia->write.in;
1092         ssize_t err;
1093
1094         fuse_write_args_fill(ia, ff, pos, count);
1095         inarg->flags = fuse_write_flags(iocb);
1096         if (owner != NULL) {
1097                 inarg->write_flags |= FUSE_WRITE_LOCKOWNER;
1098                 inarg->lock_owner = fuse_lock_owner_id(fm->fc, owner);
1099         }
1100
1101         if (ia->io->async)
1102                 return fuse_async_req_send(fm, ia, count);
1103
1104         err = fuse_simple_request(fm, &ia->ap.args);
1105         if (!err && ia->write.out.size > count)
1106                 err = -EIO;
1107
1108         return err ?: ia->write.out.size;
1109 }
1110
1111 bool fuse_write_update_attr(struct inode *inode, loff_t pos, ssize_t written)
1112 {
1113         struct fuse_conn *fc = get_fuse_conn(inode);
1114         struct fuse_inode *fi = get_fuse_inode(inode);
1115         bool ret = false;
1116
1117         spin_lock(&fi->lock);
1118         fi->attr_version = atomic64_inc_return(&fc->attr_version);
1119         if (written > 0 && pos > inode->i_size) {
1120                 i_size_write(inode, pos);
1121                 ret = true;
1122         }
1123         spin_unlock(&fi->lock);
1124
1125         fuse_invalidate_attr_mask(inode, FUSE_STATX_MODSIZE);
1126
1127         return ret;
1128 }
1129
1130 static ssize_t fuse_send_write_pages(struct fuse_io_args *ia,
1131                                      struct kiocb *iocb, struct inode *inode,
1132                                      loff_t pos, size_t count)
1133 {
1134         struct fuse_args_pages *ap = &ia->ap;
1135         struct file *file = iocb->ki_filp;
1136         struct fuse_file *ff = file->private_data;
1137         struct fuse_mount *fm = ff->fm;
1138         unsigned int offset, i;
1139         bool short_write;
1140         int err;
1141
1142         for (i = 0; i < ap->num_pages; i++)
1143                 fuse_wait_on_page_writeback(inode, ap->pages[i]->index);
1144
1145         fuse_write_args_fill(ia, ff, pos, count);
1146         ia->write.in.flags = fuse_write_flags(iocb);
1147         if (fm->fc->handle_killpriv_v2 && !capable(CAP_FSETID))
1148                 ia->write.in.write_flags |= FUSE_WRITE_KILL_SUIDGID;
1149
1150         err = fuse_simple_request(fm, &ap->args);
1151         if (!err && ia->write.out.size > count)
1152                 err = -EIO;
1153
1154         short_write = ia->write.out.size < count;
1155         offset = ap->descs[0].offset;
1156         count = ia->write.out.size;
1157         for (i = 0; i < ap->num_pages; i++) {
1158                 struct page *page = ap->pages[i];
1159
1160                 if (err) {
1161                         ClearPageUptodate(page);
1162                 } else {
1163                         if (count >= PAGE_SIZE - offset)
1164                                 count -= PAGE_SIZE - offset;
1165                         else {
1166                                 if (short_write)
1167                                         ClearPageUptodate(page);
1168                                 count = 0;
1169                         }
1170                         offset = 0;
1171                 }
1172                 if (ia->write.page_locked && (i == ap->num_pages - 1))
1173                         unlock_page(page);
1174                 put_page(page);
1175         }
1176
1177         return err;
1178 }
1179
1180 static ssize_t fuse_fill_write_pages(struct fuse_io_args *ia,
1181                                      struct address_space *mapping,
1182                                      struct iov_iter *ii, loff_t pos,
1183                                      unsigned int max_pages)
1184 {
1185         struct fuse_args_pages *ap = &ia->ap;
1186         struct fuse_conn *fc = get_fuse_conn(mapping->host);
1187         unsigned offset = pos & (PAGE_SIZE - 1);
1188         size_t count = 0;
1189         int err;
1190
1191         ap->args.in_pages = true;
1192         ap->descs[0].offset = offset;
1193
1194         do {
1195                 size_t tmp;
1196                 struct page *page;
1197                 pgoff_t index = pos >> PAGE_SHIFT;
1198                 size_t bytes = min_t(size_t, PAGE_SIZE - offset,
1199                                      iov_iter_count(ii));
1200
1201                 bytes = min_t(size_t, bytes, fc->max_write - count);
1202
1203  again:
1204                 err = -EFAULT;
1205                 if (fault_in_iov_iter_readable(ii, bytes))
1206                         break;
1207
1208                 err = -ENOMEM;
1209                 page = grab_cache_page_write_begin(mapping, index);
1210                 if (!page)
1211                         break;
1212
1213                 if (mapping_writably_mapped(mapping))
1214                         flush_dcache_page(page);
1215
1216                 tmp = copy_page_from_iter_atomic(page, offset, bytes, ii);
1217                 flush_dcache_page(page);
1218
1219                 if (!tmp) {
1220                         unlock_page(page);
1221                         put_page(page);
1222                         goto again;
1223                 }
1224
1225                 err = 0;
1226                 ap->pages[ap->num_pages] = page;
1227                 ap->descs[ap->num_pages].length = tmp;
1228                 ap->num_pages++;
1229
1230                 count += tmp;
1231                 pos += tmp;
1232                 offset += tmp;
1233                 if (offset == PAGE_SIZE)
1234                         offset = 0;
1235
1236                 /* If we copied full page, mark it uptodate */
1237                 if (tmp == PAGE_SIZE)
1238                         SetPageUptodate(page);
1239
1240                 if (PageUptodate(page)) {
1241                         unlock_page(page);
1242                 } else {
1243                         ia->write.page_locked = true;
1244                         break;
1245                 }
1246                 if (!fc->big_writes)
1247                         break;
1248         } while (iov_iter_count(ii) && count < fc->max_write &&
1249                  ap->num_pages < max_pages && offset == 0);
1250
1251         return count > 0 ? count : err;
1252 }
1253
1254 static inline unsigned int fuse_wr_pages(loff_t pos, size_t len,
1255                                      unsigned int max_pages)
1256 {
1257         return min_t(unsigned int,
1258                      ((pos + len - 1) >> PAGE_SHIFT) -
1259                      (pos >> PAGE_SHIFT) + 1,
1260                      max_pages);
1261 }
1262
1263 static ssize_t fuse_perform_write(struct kiocb *iocb, struct iov_iter *ii)
1264 {
1265         struct address_space *mapping = iocb->ki_filp->f_mapping;
1266         struct inode *inode = mapping->host;
1267         struct fuse_conn *fc = get_fuse_conn(inode);
1268         struct fuse_inode *fi = get_fuse_inode(inode);
1269         loff_t pos = iocb->ki_pos;
1270         int err = 0;
1271         ssize_t res = 0;
1272
1273         if (inode->i_size < pos + iov_iter_count(ii))
1274                 set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
1275
1276         do {
1277                 ssize_t count;
1278                 struct fuse_io_args ia = {};
1279                 struct fuse_args_pages *ap = &ia.ap;
1280                 unsigned int nr_pages = fuse_wr_pages(pos, iov_iter_count(ii),
1281                                                       fc->max_pages);
1282
1283                 ap->pages = fuse_pages_alloc(nr_pages, GFP_KERNEL, &ap->descs);
1284                 if (!ap->pages) {
1285                         err = -ENOMEM;
1286                         break;
1287                 }
1288
1289                 count = fuse_fill_write_pages(&ia, mapping, ii, pos, nr_pages);
1290                 if (count <= 0) {
1291                         err = count;
1292                 } else {
1293                         err = fuse_send_write_pages(&ia, iocb, inode,
1294                                                     pos, count);
1295                         if (!err) {
1296                                 size_t num_written = ia.write.out.size;
1297
1298                                 res += num_written;
1299                                 pos += num_written;
1300
1301                                 /* break out of the loop on short write */
1302                                 if (num_written != count)
1303                                         err = -EIO;
1304                         }
1305                 }
1306                 kfree(ap->pages);
1307         } while (!err && iov_iter_count(ii));
1308
1309         fuse_write_update_attr(inode, pos, res);
1310         clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
1311
1312         if (!res)
1313                 return err;
1314         iocb->ki_pos += res;
1315         return res;
1316 }
1317
1318 static bool fuse_io_past_eof(struct kiocb *iocb, struct iov_iter *iter)
1319 {
1320         struct inode *inode = file_inode(iocb->ki_filp);
1321
1322         return iocb->ki_pos + iov_iter_count(iter) > i_size_read(inode);
1323 }
1324
1325 /*
1326  * @return true if an exclusive lock for direct IO writes is needed
1327  */
1328 static bool fuse_dio_wr_exclusive_lock(struct kiocb *iocb, struct iov_iter *from)
1329 {
1330         struct file *file = iocb->ki_filp;
1331         struct fuse_file *ff = file->private_data;
1332         struct inode *inode = file_inode(iocb->ki_filp);
1333         struct fuse_inode *fi = get_fuse_inode(inode);
1334
1335         /* Server side has to advise that it supports parallel dio writes. */
1336         if (!(ff->open_flags & FOPEN_PARALLEL_DIRECT_WRITES))
1337                 return true;
1338
1339         /*
1340          * Append will need to know the eventual EOF - always needs an
1341          * exclusive lock.
1342          */
1343         if (iocb->ki_flags & IOCB_APPEND)
1344                 return true;
1345
1346         /* shared locks are not allowed with parallel page cache IO */
1347         if (test_bit(FUSE_I_CACHE_IO_MODE, &fi->state))
1348                 return false;
1349
1350         /* Parallel dio beyond EOF is not supported, at least for now. */
1351         if (fuse_io_past_eof(iocb, from))
1352                 return true;
1353
1354         return false;
1355 }
1356
1357 static void fuse_dio_lock(struct kiocb *iocb, struct iov_iter *from,
1358                           bool *exclusive)
1359 {
1360         struct inode *inode = file_inode(iocb->ki_filp);
1361         struct fuse_inode *fi = get_fuse_inode(inode);
1362
1363         *exclusive = fuse_dio_wr_exclusive_lock(iocb, from);
1364         if (*exclusive) {
1365                 inode_lock(inode);
1366         } else {
1367                 inode_lock_shared(inode);
1368                 /*
1369                  * New parallal dio allowed only if inode is not in caching
1370                  * mode and denies new opens in caching mode. This check
1371                  * should be performed only after taking shared inode lock.
1372                  * Previous past eof check was without inode lock and might
1373                  * have raced, so check it again.
1374                  */
1375                 if (fuse_io_past_eof(iocb, from) ||
1376                     fuse_inode_uncached_io_start(fi, NULL) != 0) {
1377                         inode_unlock_shared(inode);
1378                         inode_lock(inode);
1379                         *exclusive = true;
1380                 }
1381         }
1382 }
1383
1384 static void fuse_dio_unlock(struct kiocb *iocb, bool exclusive)
1385 {
1386         struct inode *inode = file_inode(iocb->ki_filp);
1387         struct fuse_inode *fi = get_fuse_inode(inode);
1388
1389         if (exclusive) {
1390                 inode_unlock(inode);
1391         } else {
1392                 /* Allow opens in caching mode after last parallel dio end */
1393                 fuse_inode_uncached_io_end(fi);
1394                 inode_unlock_shared(inode);
1395         }
1396 }
1397
1398 static ssize_t fuse_cache_write_iter(struct kiocb *iocb, struct iov_iter *from)
1399 {
1400         struct file *file = iocb->ki_filp;
1401         struct address_space *mapping = file->f_mapping;
1402         ssize_t written = 0;
1403         struct inode *inode = mapping->host;
1404         ssize_t err, count;
1405         struct fuse_conn *fc = get_fuse_conn(inode);
1406
1407         if (fc->writeback_cache) {
1408                 /* Update size (EOF optimization) and mode (SUID clearing) */
1409                 err = fuse_update_attributes(mapping->host, file,
1410                                              STATX_SIZE | STATX_MODE);
1411                 if (err)
1412                         return err;
1413
1414                 if (fc->handle_killpriv_v2 &&
1415                     setattr_should_drop_suidgid(&nop_mnt_idmap,
1416                                                 file_inode(file))) {
1417                         goto writethrough;
1418                 }
1419
1420                 return generic_file_write_iter(iocb, from);
1421         }
1422
1423 writethrough:
1424         inode_lock(inode);
1425
1426         err = count = generic_write_checks(iocb, from);
1427         if (err <= 0)
1428                 goto out;
1429
1430         task_io_account_write(count);
1431
1432         err = file_remove_privs(file);
1433         if (err)
1434                 goto out;
1435
1436         err = file_update_time(file);
1437         if (err)
1438                 goto out;
1439
1440         if (iocb->ki_flags & IOCB_DIRECT) {
1441                 written = generic_file_direct_write(iocb, from);
1442                 if (written < 0 || !iov_iter_count(from))
1443                         goto out;
1444                 written = direct_write_fallback(iocb, from, written,
1445                                 fuse_perform_write(iocb, from));
1446         } else {
1447                 written = fuse_perform_write(iocb, from);
1448         }
1449 out:
1450         inode_unlock(inode);
1451         if (written > 0)
1452                 written = generic_write_sync(iocb, written);
1453
1454         return written ? written : err;
1455 }
1456
1457 static inline unsigned long fuse_get_user_addr(const struct iov_iter *ii)
1458 {
1459         return (unsigned long)iter_iov(ii)->iov_base + ii->iov_offset;
1460 }
1461
1462 static inline size_t fuse_get_frag_size(const struct iov_iter *ii,
1463                                         size_t max_size)
1464 {
1465         return min(iov_iter_single_seg_count(ii), max_size);
1466 }
1467
1468 static int fuse_get_user_pages(struct fuse_args_pages *ap, struct iov_iter *ii,
1469                                size_t *nbytesp, int write,
1470                                unsigned int max_pages)
1471 {
1472         size_t nbytes = 0;  /* # bytes already packed in req */
1473         ssize_t ret = 0;
1474
1475         /* Special case for kernel I/O: can copy directly into the buffer */
1476         if (iov_iter_is_kvec(ii)) {
1477                 unsigned long user_addr = fuse_get_user_addr(ii);
1478                 size_t frag_size = fuse_get_frag_size(ii, *nbytesp);
1479
1480                 if (write)
1481                         ap->args.in_args[1].value = (void *) user_addr;
1482                 else
1483                         ap->args.out_args[0].value = (void *) user_addr;
1484
1485                 iov_iter_advance(ii, frag_size);
1486                 *nbytesp = frag_size;
1487                 return 0;
1488         }
1489
1490         while (nbytes < *nbytesp && ap->num_pages < max_pages) {
1491                 unsigned npages;
1492                 size_t start;
1493                 struct page **pt_pages;
1494
1495                 pt_pages = &ap->pages[ap->num_pages];
1496                 ret = iov_iter_extract_pages(ii, &pt_pages,
1497                                              *nbytesp - nbytes,
1498                                              max_pages - ap->num_pages,
1499                                              0, &start);
1500                 if (ret < 0)
1501                         break;
1502
1503                 nbytes += ret;
1504
1505                 ret += start;
1506                 npages = DIV_ROUND_UP(ret, PAGE_SIZE);
1507
1508                 ap->descs[ap->num_pages].offset = start;
1509                 fuse_page_descs_length_init(ap->descs, ap->num_pages, npages);
1510
1511                 ap->num_pages += npages;
1512                 ap->descs[ap->num_pages - 1].length -=
1513                         (PAGE_SIZE - ret) & (PAGE_SIZE - 1);
1514         }
1515
1516         ap->args.is_pinned = iov_iter_extract_will_pin(ii);
1517         ap->args.user_pages = true;
1518         if (write)
1519                 ap->args.in_pages = true;
1520         else
1521                 ap->args.out_pages = true;
1522
1523         *nbytesp = nbytes;
1524
1525         return ret < 0 ? ret : 0;
1526 }
1527
1528 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
1529                        loff_t *ppos, int flags)
1530 {
1531         int write = flags & FUSE_DIO_WRITE;
1532         int cuse = flags & FUSE_DIO_CUSE;
1533         struct file *file = io->iocb->ki_filp;
1534         struct address_space *mapping = file->f_mapping;
1535         struct inode *inode = mapping->host;
1536         struct fuse_file *ff = file->private_data;
1537         struct fuse_conn *fc = ff->fm->fc;
1538         size_t nmax = write ? fc->max_write : fc->max_read;
1539         loff_t pos = *ppos;
1540         size_t count = iov_iter_count(iter);
1541         pgoff_t idx_from = pos >> PAGE_SHIFT;
1542         pgoff_t idx_to = (pos + count - 1) >> PAGE_SHIFT;
1543         ssize_t res = 0;
1544         int err = 0;
1545         struct fuse_io_args *ia;
1546         unsigned int max_pages;
1547         bool fopen_direct_io = ff->open_flags & FOPEN_DIRECT_IO;
1548
1549         max_pages = iov_iter_npages(iter, fc->max_pages);
1550         ia = fuse_io_alloc(io, max_pages);
1551         if (!ia)
1552                 return -ENOMEM;
1553
1554         if (fopen_direct_io && fc->direct_io_allow_mmap) {
1555                 res = filemap_write_and_wait_range(mapping, pos, pos + count - 1);
1556                 if (res) {
1557                         fuse_io_free(ia);
1558                         return res;
1559                 }
1560         }
1561         if (!cuse && fuse_range_is_writeback(inode, idx_from, idx_to)) {
1562                 if (!write)
1563                         inode_lock(inode);
1564                 fuse_sync_writes(inode);
1565                 if (!write)
1566                         inode_unlock(inode);
1567         }
1568
1569         if (fopen_direct_io && write) {
1570                 res = invalidate_inode_pages2_range(mapping, idx_from, idx_to);
1571                 if (res) {
1572                         fuse_io_free(ia);
1573                         return res;
1574                 }
1575         }
1576
1577         io->should_dirty = !write && user_backed_iter(iter);
1578         while (count) {
1579                 ssize_t nres;
1580                 fl_owner_t owner = current->files;
1581                 size_t nbytes = min(count, nmax);
1582
1583                 err = fuse_get_user_pages(&ia->ap, iter, &nbytes, write,
1584                                           max_pages);
1585                 if (err && !nbytes)
1586                         break;
1587
1588                 if (write) {
1589                         if (!capable(CAP_FSETID))
1590                                 ia->write.in.write_flags |= FUSE_WRITE_KILL_SUIDGID;
1591
1592                         nres = fuse_send_write(ia, pos, nbytes, owner);
1593                 } else {
1594                         nres = fuse_send_read(ia, pos, nbytes, owner);
1595                 }
1596
1597                 if (!io->async || nres < 0) {
1598                         fuse_release_user_pages(&ia->ap, io->should_dirty);
1599                         fuse_io_free(ia);
1600                 }
1601                 ia = NULL;
1602                 if (nres < 0) {
1603                         iov_iter_revert(iter, nbytes);
1604                         err = nres;
1605                         break;
1606                 }
1607                 WARN_ON(nres > nbytes);
1608
1609                 count -= nres;
1610                 res += nres;
1611                 pos += nres;
1612                 if (nres != nbytes) {
1613                         iov_iter_revert(iter, nbytes - nres);
1614                         break;
1615                 }
1616                 if (count) {
1617                         max_pages = iov_iter_npages(iter, fc->max_pages);
1618                         ia = fuse_io_alloc(io, max_pages);
1619                         if (!ia)
1620                                 break;
1621                 }
1622         }
1623         if (ia)
1624                 fuse_io_free(ia);
1625         if (res > 0)
1626                 *ppos = pos;
1627
1628         return res > 0 ? res : err;
1629 }
1630 EXPORT_SYMBOL_GPL(fuse_direct_io);
1631
1632 static ssize_t __fuse_direct_read(struct fuse_io_priv *io,
1633                                   struct iov_iter *iter,
1634                                   loff_t *ppos)
1635 {
1636         ssize_t res;
1637         struct inode *inode = file_inode(io->iocb->ki_filp);
1638
1639         res = fuse_direct_io(io, iter, ppos, 0);
1640
1641         fuse_invalidate_atime(inode);
1642
1643         return res;
1644 }
1645
1646 static ssize_t fuse_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
1647
1648 static ssize_t fuse_direct_read_iter(struct kiocb *iocb, struct iov_iter *to)
1649 {
1650         ssize_t res;
1651
1652         if (!is_sync_kiocb(iocb) && iocb->ki_flags & IOCB_DIRECT) {
1653                 res = fuse_direct_IO(iocb, to);
1654         } else {
1655                 struct fuse_io_priv io = FUSE_IO_PRIV_SYNC(iocb);
1656
1657                 res = __fuse_direct_read(&io, to, &iocb->ki_pos);
1658         }
1659
1660         return res;
1661 }
1662
1663 static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from)
1664 {
1665         struct inode *inode = file_inode(iocb->ki_filp);
1666         struct fuse_io_priv io = FUSE_IO_PRIV_SYNC(iocb);
1667         ssize_t res;
1668         bool exclusive;
1669
1670         fuse_dio_lock(iocb, from, &exclusive);
1671         res = generic_write_checks(iocb, from);
1672         if (res > 0) {
1673                 task_io_account_write(res);
1674                 if (!is_sync_kiocb(iocb) && iocb->ki_flags & IOCB_DIRECT) {
1675                         res = fuse_direct_IO(iocb, from);
1676                 } else {
1677                         res = fuse_direct_io(&io, from, &iocb->ki_pos,
1678                                              FUSE_DIO_WRITE);
1679                         fuse_write_update_attr(inode, iocb->ki_pos, res);
1680                 }
1681         }
1682         fuse_dio_unlock(iocb, exclusive);
1683
1684         return res;
1685 }
1686
1687 static ssize_t fuse_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
1688 {
1689         struct file *file = iocb->ki_filp;
1690         struct fuse_file *ff = file->private_data;
1691         struct inode *inode = file_inode(file);
1692
1693         if (fuse_is_bad(inode))
1694                 return -EIO;
1695
1696         if (FUSE_IS_DAX(inode))
1697                 return fuse_dax_read_iter(iocb, to);
1698
1699         /* FOPEN_DIRECT_IO overrides FOPEN_PASSTHROUGH */
1700         if (ff->open_flags & FOPEN_DIRECT_IO)
1701                 return fuse_direct_read_iter(iocb, to);
1702         else if (fuse_file_passthrough(ff))
1703                 return fuse_passthrough_read_iter(iocb, to);
1704         else
1705                 return fuse_cache_read_iter(iocb, to);
1706 }
1707
1708 static ssize_t fuse_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
1709 {
1710         struct file *file = iocb->ki_filp;
1711         struct fuse_file *ff = file->private_data;
1712         struct inode *inode = file_inode(file);
1713
1714         if (fuse_is_bad(inode))
1715                 return -EIO;
1716
1717         if (FUSE_IS_DAX(inode))
1718                 return fuse_dax_write_iter(iocb, from);
1719
1720         /* FOPEN_DIRECT_IO overrides FOPEN_PASSTHROUGH */
1721         if (ff->open_flags & FOPEN_DIRECT_IO)
1722                 return fuse_direct_write_iter(iocb, from);
1723         else if (fuse_file_passthrough(ff))
1724                 return fuse_passthrough_write_iter(iocb, from);
1725         else
1726                 return fuse_cache_write_iter(iocb, from);
1727 }
1728
1729 static ssize_t fuse_splice_read(struct file *in, loff_t *ppos,
1730                                 struct pipe_inode_info *pipe, size_t len,
1731                                 unsigned int flags)
1732 {
1733         struct fuse_file *ff = in->private_data;
1734
1735         /* FOPEN_DIRECT_IO overrides FOPEN_PASSTHROUGH */
1736         if (fuse_file_passthrough(ff) && !(ff->open_flags & FOPEN_DIRECT_IO))
1737                 return fuse_passthrough_splice_read(in, ppos, pipe, len, flags);
1738         else
1739                 return filemap_splice_read(in, ppos, pipe, len, flags);
1740 }
1741
1742 static ssize_t fuse_splice_write(struct pipe_inode_info *pipe, struct file *out,
1743                                  loff_t *ppos, size_t len, unsigned int flags)
1744 {
1745         struct fuse_file *ff = out->private_data;
1746
1747         /* FOPEN_DIRECT_IO overrides FOPEN_PASSTHROUGH */
1748         if (fuse_file_passthrough(ff) && !(ff->open_flags & FOPEN_DIRECT_IO))
1749                 return fuse_passthrough_splice_write(pipe, out, ppos, len, flags);
1750         else
1751                 return iter_file_splice_write(pipe, out, ppos, len, flags);
1752 }
1753
1754 static void fuse_writepage_free(struct fuse_writepage_args *wpa)
1755 {
1756         struct fuse_args_pages *ap = &wpa->ia.ap;
1757         int i;
1758
1759         if (wpa->bucket)
1760                 fuse_sync_bucket_dec(wpa->bucket);
1761
1762         for (i = 0; i < ap->num_pages; i++)
1763                 __free_page(ap->pages[i]);
1764
1765         if (wpa->ia.ff)
1766                 fuse_file_put(wpa->ia.ff, false);
1767
1768         kfree(ap->pages);
1769         kfree(wpa);
1770 }
1771
1772 static void fuse_writepage_finish(struct fuse_mount *fm,
1773                                   struct fuse_writepage_args *wpa)
1774 {
1775         struct fuse_args_pages *ap = &wpa->ia.ap;
1776         struct inode *inode = wpa->inode;
1777         struct fuse_inode *fi = get_fuse_inode(inode);
1778         struct backing_dev_info *bdi = inode_to_bdi(inode);
1779         int i;
1780
1781         for (i = 0; i < ap->num_pages; i++) {
1782                 dec_wb_stat(&bdi->wb, WB_WRITEBACK);
1783                 dec_node_page_state(ap->pages[i], NR_WRITEBACK_TEMP);
1784                 wb_writeout_inc(&bdi->wb);
1785         }
1786         wake_up(&fi->page_waitq);
1787 }
1788
1789 /* Called under fi->lock, may release and reacquire it */
1790 static void fuse_send_writepage(struct fuse_mount *fm,
1791                                 struct fuse_writepage_args *wpa, loff_t size)
1792 __releases(fi->lock)
1793 __acquires(fi->lock)
1794 {
1795         struct fuse_writepage_args *aux, *next;
1796         struct fuse_inode *fi = get_fuse_inode(wpa->inode);
1797         struct fuse_write_in *inarg = &wpa->ia.write.in;
1798         struct fuse_args *args = &wpa->ia.ap.args;
1799         __u64 data_size = wpa->ia.ap.num_pages * PAGE_SIZE;
1800         int err;
1801
1802         fi->writectr++;
1803         if (inarg->offset + data_size <= size) {
1804                 inarg->size = data_size;
1805         } else if (inarg->offset < size) {
1806                 inarg->size = size - inarg->offset;
1807         } else {
1808                 /* Got truncated off completely */
1809                 goto out_free;
1810         }
1811
1812         args->in_args[1].size = inarg->size;
1813         args->force = true;
1814         args->nocreds = true;
1815
1816         err = fuse_simple_background(fm, args, GFP_ATOMIC);
1817         if (err == -ENOMEM) {
1818                 spin_unlock(&fi->lock);
1819                 err = fuse_simple_background(fm, args, GFP_NOFS | __GFP_NOFAIL);
1820                 spin_lock(&fi->lock);
1821         }
1822
1823         /* Fails on broken connection only */
1824         if (unlikely(err))
1825                 goto out_free;
1826
1827         return;
1828
1829  out_free:
1830         fi->writectr--;
1831         rb_erase(&wpa->writepages_entry, &fi->writepages);
1832         fuse_writepage_finish(fm, wpa);
1833         spin_unlock(&fi->lock);
1834
1835         /* After fuse_writepage_finish() aux request list is private */
1836         for (aux = wpa->next; aux; aux = next) {
1837                 next = aux->next;
1838                 aux->next = NULL;
1839                 fuse_writepage_free(aux);
1840         }
1841
1842         fuse_writepage_free(wpa);
1843         spin_lock(&fi->lock);
1844 }
1845
1846 /*
1847  * If fi->writectr is positive (no truncate or fsync going on) send
1848  * all queued writepage requests.
1849  *
1850  * Called with fi->lock
1851  */
1852 void fuse_flush_writepages(struct inode *inode)
1853 __releases(fi->lock)
1854 __acquires(fi->lock)
1855 {
1856         struct fuse_mount *fm = get_fuse_mount(inode);
1857         struct fuse_inode *fi = get_fuse_inode(inode);
1858         loff_t crop = i_size_read(inode);
1859         struct fuse_writepage_args *wpa;
1860
1861         while (fi->writectr >= 0 && !list_empty(&fi->queued_writes)) {
1862                 wpa = list_entry(fi->queued_writes.next,
1863                                  struct fuse_writepage_args, queue_entry);
1864                 list_del_init(&wpa->queue_entry);
1865                 fuse_send_writepage(fm, wpa, crop);
1866         }
1867 }
1868
1869 static struct fuse_writepage_args *fuse_insert_writeback(struct rb_root *root,
1870                                                 struct fuse_writepage_args *wpa)
1871 {
1872         pgoff_t idx_from = wpa->ia.write.in.offset >> PAGE_SHIFT;
1873         pgoff_t idx_to = idx_from + wpa->ia.ap.num_pages - 1;
1874         struct rb_node **p = &root->rb_node;
1875         struct rb_node  *parent = NULL;
1876
1877         WARN_ON(!wpa->ia.ap.num_pages);
1878         while (*p) {
1879                 struct fuse_writepage_args *curr;
1880                 pgoff_t curr_index;
1881
1882                 parent = *p;
1883                 curr = rb_entry(parent, struct fuse_writepage_args,
1884                                 writepages_entry);
1885                 WARN_ON(curr->inode != wpa->inode);
1886                 curr_index = curr->ia.write.in.offset >> PAGE_SHIFT;
1887
1888                 if (idx_from >= curr_index + curr->ia.ap.num_pages)
1889                         p = &(*p)->rb_right;
1890                 else if (idx_to < curr_index)
1891                         p = &(*p)->rb_left;
1892                 else
1893                         return curr;
1894         }
1895
1896         rb_link_node(&wpa->writepages_entry, parent, p);
1897         rb_insert_color(&wpa->writepages_entry, root);
1898         return NULL;
1899 }
1900
1901 static void tree_insert(struct rb_root *root, struct fuse_writepage_args *wpa)
1902 {
1903         WARN_ON(fuse_insert_writeback(root, wpa));
1904 }
1905
1906 static void fuse_writepage_end(struct fuse_mount *fm, struct fuse_args *args,
1907                                int error)
1908 {
1909         struct fuse_writepage_args *wpa =
1910                 container_of(args, typeof(*wpa), ia.ap.args);
1911         struct inode *inode = wpa->inode;
1912         struct fuse_inode *fi = get_fuse_inode(inode);
1913         struct fuse_conn *fc = get_fuse_conn(inode);
1914
1915         mapping_set_error(inode->i_mapping, error);
1916         /*
1917          * A writeback finished and this might have updated mtime/ctime on
1918          * server making local mtime/ctime stale.  Hence invalidate attrs.
1919          * Do this only if writeback_cache is not enabled.  If writeback_cache
1920          * is enabled, we trust local ctime/mtime.
1921          */
1922         if (!fc->writeback_cache)
1923                 fuse_invalidate_attr_mask(inode, FUSE_STATX_MODIFY);
1924         spin_lock(&fi->lock);
1925         rb_erase(&wpa->writepages_entry, &fi->writepages);
1926         while (wpa->next) {
1927                 struct fuse_mount *fm = get_fuse_mount(inode);
1928                 struct fuse_write_in *inarg = &wpa->ia.write.in;
1929                 struct fuse_writepage_args *next = wpa->next;
1930
1931                 wpa->next = next->next;
1932                 next->next = NULL;
1933                 next->ia.ff = fuse_file_get(wpa->ia.ff);
1934                 tree_insert(&fi->writepages, next);
1935
1936                 /*
1937                  * Skip fuse_flush_writepages() to make it easy to crop requests
1938                  * based on primary request size.
1939                  *
1940                  * 1st case (trivial): there are no concurrent activities using
1941                  * fuse_set/release_nowrite.  Then we're on safe side because
1942                  * fuse_flush_writepages() would call fuse_send_writepage()
1943                  * anyway.
1944                  *
1945                  * 2nd case: someone called fuse_set_nowrite and it is waiting
1946                  * now for completion of all in-flight requests.  This happens
1947                  * rarely and no more than once per page, so this should be
1948                  * okay.
1949                  *
1950                  * 3rd case: someone (e.g. fuse_do_setattr()) is in the middle
1951                  * of fuse_set_nowrite..fuse_release_nowrite section.  The fact
1952                  * that fuse_set_nowrite returned implies that all in-flight
1953                  * requests were completed along with all of their secondary
1954                  * requests.  Further primary requests are blocked by negative
1955                  * writectr.  Hence there cannot be any in-flight requests and
1956                  * no invocations of fuse_writepage_end() while we're in
1957                  * fuse_set_nowrite..fuse_release_nowrite section.
1958                  */
1959                 fuse_send_writepage(fm, next, inarg->offset + inarg->size);
1960         }
1961         fi->writectr--;
1962         fuse_writepage_finish(fm, wpa);
1963         spin_unlock(&fi->lock);
1964         fuse_writepage_free(wpa);
1965 }
1966
1967 static struct fuse_file *__fuse_write_file_get(struct fuse_inode *fi)
1968 {
1969         struct fuse_file *ff;
1970
1971         spin_lock(&fi->lock);
1972         ff = list_first_entry_or_null(&fi->write_files, struct fuse_file,
1973                                       write_entry);
1974         if (ff)
1975                 fuse_file_get(ff);
1976         spin_unlock(&fi->lock);
1977
1978         return ff;
1979 }
1980
1981 static struct fuse_file *fuse_write_file_get(struct fuse_inode *fi)
1982 {
1983         struct fuse_file *ff = __fuse_write_file_get(fi);
1984         WARN_ON(!ff);
1985         return ff;
1986 }
1987
1988 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc)
1989 {
1990         struct fuse_inode *fi = get_fuse_inode(inode);
1991         struct fuse_file *ff;
1992         int err;
1993
1994         /*
1995          * Inode is always written before the last reference is dropped and
1996          * hence this should not be reached from reclaim.
1997          *
1998          * Writing back the inode from reclaim can deadlock if the request
1999          * processing itself needs an allocation.  Allocations triggering
2000          * reclaim while serving a request can't be prevented, because it can
2001          * involve any number of unrelated userspace processes.
2002          */
2003         WARN_ON(wbc->for_reclaim);
2004
2005         ff = __fuse_write_file_get(fi);
2006         err = fuse_flush_times(inode, ff);
2007         if (ff)
2008                 fuse_file_put(ff, false);
2009
2010         return err;
2011 }
2012
2013 static struct fuse_writepage_args *fuse_writepage_args_alloc(void)
2014 {
2015         struct fuse_writepage_args *wpa;
2016         struct fuse_args_pages *ap;
2017
2018         wpa = kzalloc(sizeof(*wpa), GFP_NOFS);
2019         if (wpa) {
2020                 ap = &wpa->ia.ap;
2021                 ap->num_pages = 0;
2022                 ap->pages = fuse_pages_alloc(1, GFP_NOFS, &ap->descs);
2023                 if (!ap->pages) {
2024                         kfree(wpa);
2025                         wpa = NULL;
2026                 }
2027         }
2028         return wpa;
2029
2030 }
2031
2032 static void fuse_writepage_add_to_bucket(struct fuse_conn *fc,
2033                                          struct fuse_writepage_args *wpa)
2034 {
2035         if (!fc->sync_fs)
2036                 return;
2037
2038         rcu_read_lock();
2039         /* Prevent resurrection of dead bucket in unlikely race with syncfs */
2040         do {
2041                 wpa->bucket = rcu_dereference(fc->curr_bucket);
2042         } while (unlikely(!atomic_inc_not_zero(&wpa->bucket->count)));
2043         rcu_read_unlock();
2044 }
2045
2046 static int fuse_writepage_locked(struct folio *folio)
2047 {
2048         struct address_space *mapping = folio->mapping;
2049         struct inode *inode = mapping->host;
2050         struct fuse_conn *fc = get_fuse_conn(inode);
2051         struct fuse_inode *fi = get_fuse_inode(inode);
2052         struct fuse_writepage_args *wpa;
2053         struct fuse_args_pages *ap;
2054         struct folio *tmp_folio;
2055         int error = -ENOMEM;
2056
2057         folio_start_writeback(folio);
2058
2059         wpa = fuse_writepage_args_alloc();
2060         if (!wpa)
2061                 goto err;
2062         ap = &wpa->ia.ap;
2063
2064         tmp_folio = folio_alloc(GFP_NOFS | __GFP_HIGHMEM, 0);
2065         if (!tmp_folio)
2066                 goto err_free;
2067
2068         error = -EIO;
2069         wpa->ia.ff = fuse_write_file_get(fi);
2070         if (!wpa->ia.ff)
2071                 goto err_nofile;
2072
2073         fuse_writepage_add_to_bucket(fc, wpa);
2074         fuse_write_args_fill(&wpa->ia, wpa->ia.ff, folio_pos(folio), 0);
2075
2076         folio_copy(tmp_folio, folio);
2077         wpa->ia.write.in.write_flags |= FUSE_WRITE_CACHE;
2078         wpa->next = NULL;
2079         ap->args.in_pages = true;
2080         ap->num_pages = 1;
2081         ap->pages[0] = &tmp_folio->page;
2082         ap->descs[0].offset = 0;
2083         ap->descs[0].length = PAGE_SIZE;
2084         ap->args.end = fuse_writepage_end;
2085         wpa->inode = inode;
2086
2087         inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
2088         node_stat_add_folio(tmp_folio, NR_WRITEBACK_TEMP);
2089
2090         spin_lock(&fi->lock);
2091         tree_insert(&fi->writepages, wpa);
2092         list_add_tail(&wpa->queue_entry, &fi->queued_writes);
2093         fuse_flush_writepages(inode);
2094         spin_unlock(&fi->lock);
2095
2096         folio_end_writeback(folio);
2097
2098         return 0;
2099
2100 err_nofile:
2101         folio_put(tmp_folio);
2102 err_free:
2103         kfree(wpa);
2104 err:
2105         mapping_set_error(folio->mapping, error);
2106         folio_end_writeback(folio);
2107         return error;
2108 }
2109
2110 struct fuse_fill_wb_data {
2111         struct fuse_writepage_args *wpa;
2112         struct fuse_file *ff;
2113         struct inode *inode;
2114         struct page **orig_pages;
2115         unsigned int max_pages;
2116 };
2117
2118 static bool fuse_pages_realloc(struct fuse_fill_wb_data *data)
2119 {
2120         struct fuse_args_pages *ap = &data->wpa->ia.ap;
2121         struct fuse_conn *fc = get_fuse_conn(data->inode);
2122         struct page **pages;
2123         struct fuse_page_desc *descs;
2124         unsigned int npages = min_t(unsigned int,
2125                                     max_t(unsigned int, data->max_pages * 2,
2126                                           FUSE_DEFAULT_MAX_PAGES_PER_REQ),
2127                                     fc->max_pages);
2128         WARN_ON(npages <= data->max_pages);
2129
2130         pages = fuse_pages_alloc(npages, GFP_NOFS, &descs);
2131         if (!pages)
2132                 return false;
2133
2134         memcpy(pages, ap->pages, sizeof(struct page *) * ap->num_pages);
2135         memcpy(descs, ap->descs, sizeof(struct fuse_page_desc) * ap->num_pages);
2136         kfree(ap->pages);
2137         ap->pages = pages;
2138         ap->descs = descs;
2139         data->max_pages = npages;
2140
2141         return true;
2142 }
2143
2144 static void fuse_writepages_send(struct fuse_fill_wb_data *data)
2145 {
2146         struct fuse_writepage_args *wpa = data->wpa;
2147         struct inode *inode = data->inode;
2148         struct fuse_inode *fi = get_fuse_inode(inode);
2149         int num_pages = wpa->ia.ap.num_pages;
2150         int i;
2151
2152         wpa->ia.ff = fuse_file_get(data->ff);
2153         spin_lock(&fi->lock);
2154         list_add_tail(&wpa->queue_entry, &fi->queued_writes);
2155         fuse_flush_writepages(inode);
2156         spin_unlock(&fi->lock);
2157
2158         for (i = 0; i < num_pages; i++)
2159                 end_page_writeback(data->orig_pages[i]);
2160 }
2161
2162 /*
2163  * Check under fi->lock if the page is under writeback, and insert it onto the
2164  * rb_tree if not. Otherwise iterate auxiliary write requests, to see if there's
2165  * one already added for a page at this offset.  If there's none, then insert
2166  * this new request onto the auxiliary list, otherwise reuse the existing one by
2167  * swapping the new temp page with the old one.
2168  */
2169 static bool fuse_writepage_add(struct fuse_writepage_args *new_wpa,
2170                                struct page *page)
2171 {
2172         struct fuse_inode *fi = get_fuse_inode(new_wpa->inode);
2173         struct fuse_writepage_args *tmp;
2174         struct fuse_writepage_args *old_wpa;
2175         struct fuse_args_pages *new_ap = &new_wpa->ia.ap;
2176
2177         WARN_ON(new_ap->num_pages != 0);
2178         new_ap->num_pages = 1;
2179
2180         spin_lock(&fi->lock);
2181         old_wpa = fuse_insert_writeback(&fi->writepages, new_wpa);
2182         if (!old_wpa) {
2183                 spin_unlock(&fi->lock);
2184                 return true;
2185         }
2186
2187         for (tmp = old_wpa->next; tmp; tmp = tmp->next) {
2188                 pgoff_t curr_index;
2189
2190                 WARN_ON(tmp->inode != new_wpa->inode);
2191                 curr_index = tmp->ia.write.in.offset >> PAGE_SHIFT;
2192                 if (curr_index == page->index) {
2193                         WARN_ON(tmp->ia.ap.num_pages != 1);
2194                         swap(tmp->ia.ap.pages[0], new_ap->pages[0]);
2195                         break;
2196                 }
2197         }
2198
2199         if (!tmp) {
2200                 new_wpa->next = old_wpa->next;
2201                 old_wpa->next = new_wpa;
2202         }
2203
2204         spin_unlock(&fi->lock);
2205
2206         if (tmp) {
2207                 struct backing_dev_info *bdi = inode_to_bdi(new_wpa->inode);
2208
2209                 dec_wb_stat(&bdi->wb, WB_WRITEBACK);
2210                 dec_node_page_state(new_ap->pages[0], NR_WRITEBACK_TEMP);
2211                 wb_writeout_inc(&bdi->wb);
2212                 fuse_writepage_free(new_wpa);
2213         }
2214
2215         return false;
2216 }
2217
2218 static bool fuse_writepage_need_send(struct fuse_conn *fc, struct page *page,
2219                                      struct fuse_args_pages *ap,
2220                                      struct fuse_fill_wb_data *data)
2221 {
2222         WARN_ON(!ap->num_pages);
2223
2224         /*
2225          * Being under writeback is unlikely but possible.  For example direct
2226          * read to an mmaped fuse file will set the page dirty twice; once when
2227          * the pages are faulted with get_user_pages(), and then after the read
2228          * completed.
2229          */
2230         if (fuse_page_is_writeback(data->inode, page->index))
2231                 return true;
2232
2233         /* Reached max pages */
2234         if (ap->num_pages == fc->max_pages)
2235                 return true;
2236
2237         /* Reached max write bytes */
2238         if ((ap->num_pages + 1) * PAGE_SIZE > fc->max_write)
2239                 return true;
2240
2241         /* Discontinuity */
2242         if (data->orig_pages[ap->num_pages - 1]->index + 1 != page->index)
2243                 return true;
2244
2245         /* Need to grow the pages array?  If so, did the expansion fail? */
2246         if (ap->num_pages == data->max_pages && !fuse_pages_realloc(data))
2247                 return true;
2248
2249         return false;
2250 }
2251
2252 static int fuse_writepages_fill(struct folio *folio,
2253                 struct writeback_control *wbc, void *_data)
2254 {
2255         struct fuse_fill_wb_data *data = _data;
2256         struct fuse_writepage_args *wpa = data->wpa;
2257         struct fuse_args_pages *ap = &wpa->ia.ap;
2258         struct inode *inode = data->inode;
2259         struct fuse_inode *fi = get_fuse_inode(inode);
2260         struct fuse_conn *fc = get_fuse_conn(inode);
2261         struct page *tmp_page;
2262         int err;
2263
2264         if (!data->ff) {
2265                 err = -EIO;
2266                 data->ff = fuse_write_file_get(fi);
2267                 if (!data->ff)
2268                         goto out_unlock;
2269         }
2270
2271         if (wpa && fuse_writepage_need_send(fc, &folio->page, ap, data)) {
2272                 fuse_writepages_send(data);
2273                 data->wpa = NULL;
2274         }
2275
2276         err = -ENOMEM;
2277         tmp_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
2278         if (!tmp_page)
2279                 goto out_unlock;
2280
2281         /*
2282          * The page must not be redirtied until the writeout is completed
2283          * (i.e. userspace has sent a reply to the write request).  Otherwise
2284          * there could be more than one temporary page instance for each real
2285          * page.
2286          *
2287          * This is ensured by holding the page lock in page_mkwrite() while
2288          * checking fuse_page_is_writeback().  We already hold the page lock
2289          * since clear_page_dirty_for_io() and keep it held until we add the
2290          * request to the fi->writepages list and increment ap->num_pages.
2291          * After this fuse_page_is_writeback() will indicate that the page is
2292          * under writeback, so we can release the page lock.
2293          */
2294         if (data->wpa == NULL) {
2295                 err = -ENOMEM;
2296                 wpa = fuse_writepage_args_alloc();
2297                 if (!wpa) {
2298                         __free_page(tmp_page);
2299                         goto out_unlock;
2300                 }
2301                 fuse_writepage_add_to_bucket(fc, wpa);
2302
2303                 data->max_pages = 1;
2304
2305                 ap = &wpa->ia.ap;
2306                 fuse_write_args_fill(&wpa->ia, data->ff, folio_pos(folio), 0);
2307                 wpa->ia.write.in.write_flags |= FUSE_WRITE_CACHE;
2308                 wpa->next = NULL;
2309                 ap->args.in_pages = true;
2310                 ap->args.end = fuse_writepage_end;
2311                 ap->num_pages = 0;
2312                 wpa->inode = inode;
2313         }
2314         folio_start_writeback(folio);
2315
2316         copy_highpage(tmp_page, &folio->page);
2317         ap->pages[ap->num_pages] = tmp_page;
2318         ap->descs[ap->num_pages].offset = 0;
2319         ap->descs[ap->num_pages].length = PAGE_SIZE;
2320         data->orig_pages[ap->num_pages] = &folio->page;
2321
2322         inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
2323         inc_node_page_state(tmp_page, NR_WRITEBACK_TEMP);
2324
2325         err = 0;
2326         if (data->wpa) {
2327                 /*
2328                  * Protected by fi->lock against concurrent access by
2329                  * fuse_page_is_writeback().
2330                  */
2331                 spin_lock(&fi->lock);
2332                 ap->num_pages++;
2333                 spin_unlock(&fi->lock);
2334         } else if (fuse_writepage_add(wpa, &folio->page)) {
2335                 data->wpa = wpa;
2336         } else {
2337                 folio_end_writeback(folio);
2338         }
2339 out_unlock:
2340         folio_unlock(folio);
2341
2342         return err;
2343 }
2344
2345 static int fuse_writepages(struct address_space *mapping,
2346                            struct writeback_control *wbc)
2347 {
2348         struct inode *inode = mapping->host;
2349         struct fuse_conn *fc = get_fuse_conn(inode);
2350         struct fuse_fill_wb_data data;
2351         int err;
2352
2353         err = -EIO;
2354         if (fuse_is_bad(inode))
2355                 goto out;
2356
2357         if (wbc->sync_mode == WB_SYNC_NONE &&
2358             fc->num_background >= fc->congestion_threshold)
2359                 return 0;
2360
2361         data.inode = inode;
2362         data.wpa = NULL;
2363         data.ff = NULL;
2364
2365         err = -ENOMEM;
2366         data.orig_pages = kcalloc(fc->max_pages,
2367                                   sizeof(struct page *),
2368                                   GFP_NOFS);
2369         if (!data.orig_pages)
2370                 goto out;
2371
2372         err = write_cache_pages(mapping, wbc, fuse_writepages_fill, &data);
2373         if (data.wpa) {
2374                 WARN_ON(!data.wpa->ia.ap.num_pages);
2375                 fuse_writepages_send(&data);
2376         }
2377         if (data.ff)
2378                 fuse_file_put(data.ff, false);
2379
2380         kfree(data.orig_pages);
2381 out:
2382         return err;
2383 }
2384
2385 /*
2386  * It's worthy to make sure that space is reserved on disk for the write,
2387  * but how to implement it without killing performance need more thinking.
2388  */
2389 static int fuse_write_begin(struct file *file, struct address_space *mapping,
2390                 loff_t pos, unsigned len, struct page **pagep, void **fsdata)
2391 {
2392         pgoff_t index = pos >> PAGE_SHIFT;
2393         struct fuse_conn *fc = get_fuse_conn(file_inode(file));
2394         struct page *page;
2395         loff_t fsize;
2396         int err = -ENOMEM;
2397
2398         WARN_ON(!fc->writeback_cache);
2399
2400         page = grab_cache_page_write_begin(mapping, index);
2401         if (!page)
2402                 goto error;
2403
2404         fuse_wait_on_page_writeback(mapping->host, page->index);
2405
2406         if (PageUptodate(page) || len == PAGE_SIZE)
2407                 goto success;
2408         /*
2409          * Check if the start this page comes after the end of file, in which
2410          * case the readpage can be optimized away.
2411          */
2412         fsize = i_size_read(mapping->host);
2413         if (fsize <= (pos & PAGE_MASK)) {
2414                 size_t off = pos & ~PAGE_MASK;
2415                 if (off)
2416                         zero_user_segment(page, 0, off);
2417                 goto success;
2418         }
2419         err = fuse_do_readpage(file, page);
2420         if (err)
2421                 goto cleanup;
2422 success:
2423         *pagep = page;
2424         return 0;
2425
2426 cleanup:
2427         unlock_page(page);
2428         put_page(page);
2429 error:
2430         return err;
2431 }
2432
2433 static int fuse_write_end(struct file *file, struct address_space *mapping,
2434                 loff_t pos, unsigned len, unsigned copied,
2435                 struct page *page, void *fsdata)
2436 {
2437         struct inode *inode = page->mapping->host;
2438
2439         /* Haven't copied anything?  Skip zeroing, size extending, dirtying. */
2440         if (!copied)
2441                 goto unlock;
2442
2443         pos += copied;
2444         if (!PageUptodate(page)) {
2445                 /* Zero any unwritten bytes at the end of the page */
2446                 size_t endoff = pos & ~PAGE_MASK;
2447                 if (endoff)
2448                         zero_user_segment(page, endoff, PAGE_SIZE);
2449                 SetPageUptodate(page);
2450         }
2451
2452         if (pos > inode->i_size)
2453                 i_size_write(inode, pos);
2454
2455         set_page_dirty(page);
2456
2457 unlock:
2458         unlock_page(page);
2459         put_page(page);
2460
2461         return copied;
2462 }
2463
2464 static int fuse_launder_folio(struct folio *folio)
2465 {
2466         int err = 0;
2467         if (folio_clear_dirty_for_io(folio)) {
2468                 struct inode *inode = folio->mapping->host;
2469
2470                 /* Serialize with pending writeback for the same page */
2471                 fuse_wait_on_page_writeback(inode, folio->index);
2472                 err = fuse_writepage_locked(folio);
2473                 if (!err)
2474                         fuse_wait_on_page_writeback(inode, folio->index);
2475         }
2476         return err;
2477 }
2478
2479 /*
2480  * Write back dirty data/metadata now (there may not be any suitable
2481  * open files later for data)
2482  */
2483 static void fuse_vma_close(struct vm_area_struct *vma)
2484 {
2485         int err;
2486
2487         err = write_inode_now(vma->vm_file->f_mapping->host, 1);
2488         mapping_set_error(vma->vm_file->f_mapping, err);
2489 }
2490
2491 /*
2492  * Wait for writeback against this page to complete before allowing it
2493  * to be marked dirty again, and hence written back again, possibly
2494  * before the previous writepage completed.
2495  *
2496  * Block here, instead of in ->writepage(), so that the userspace fs
2497  * can only block processes actually operating on the filesystem.
2498  *
2499  * Otherwise unprivileged userspace fs would be able to block
2500  * unrelated:
2501  *
2502  * - page migration
2503  * - sync(2)
2504  * - try_to_free_pages() with order > PAGE_ALLOC_COSTLY_ORDER
2505  */
2506 static vm_fault_t fuse_page_mkwrite(struct vm_fault *vmf)
2507 {
2508         struct page *page = vmf->page;
2509         struct inode *inode = file_inode(vmf->vma->vm_file);
2510
2511         file_update_time(vmf->vma->vm_file);
2512         lock_page(page);
2513         if (page->mapping != inode->i_mapping) {
2514                 unlock_page(page);
2515                 return VM_FAULT_NOPAGE;
2516         }
2517
2518         fuse_wait_on_page_writeback(inode, page->index);
2519         return VM_FAULT_LOCKED;
2520 }
2521
2522 static const struct vm_operations_struct fuse_file_vm_ops = {
2523         .close          = fuse_vma_close,
2524         .fault          = filemap_fault,
2525         .map_pages      = filemap_map_pages,
2526         .page_mkwrite   = fuse_page_mkwrite,
2527 };
2528
2529 static int fuse_file_mmap(struct file *file, struct vm_area_struct *vma)
2530 {
2531         struct fuse_file *ff = file->private_data;
2532         struct fuse_conn *fc = ff->fm->fc;
2533         struct inode *inode = file_inode(file);
2534         int rc;
2535
2536         /* DAX mmap is superior to direct_io mmap */
2537         if (FUSE_IS_DAX(inode))
2538                 return fuse_dax_mmap(file, vma);
2539
2540         /*
2541          * If inode is in passthrough io mode, because it has some file open
2542          * in passthrough mode, either mmap to backing file or fail mmap,
2543          * because mixing cached mmap and passthrough io mode is not allowed.
2544          */
2545         if (fuse_file_passthrough(ff))
2546                 return fuse_passthrough_mmap(file, vma);
2547         else if (fuse_inode_backing(get_fuse_inode(inode)))
2548                 return -ENODEV;
2549
2550         /*
2551          * FOPEN_DIRECT_IO handling is special compared to O_DIRECT,
2552          * as does not allow MAP_SHARED mmap without FUSE_DIRECT_IO_ALLOW_MMAP.
2553          */
2554         if (ff->open_flags & FOPEN_DIRECT_IO) {
2555                 /*
2556                  * Can't provide the coherency needed for MAP_SHARED
2557                  * if FUSE_DIRECT_IO_ALLOW_MMAP isn't set.
2558                  */
2559                 if ((vma->vm_flags & VM_MAYSHARE) && !fc->direct_io_allow_mmap)
2560                         return -ENODEV;
2561
2562                 invalidate_inode_pages2(file->f_mapping);
2563
2564                 if (!(vma->vm_flags & VM_MAYSHARE)) {
2565                         /* MAP_PRIVATE */
2566                         return generic_file_mmap(file, vma);
2567                 }
2568
2569                 /*
2570                  * First mmap of direct_io file enters caching inode io mode.
2571                  * Also waits for parallel dio writers to go into serial mode
2572                  * (exclusive instead of shared lock).
2573                  * After first mmap, the inode stays in caching io mode until
2574                  * the direct_io file release.
2575                  */
2576                 rc = fuse_file_cached_io_open(inode, ff);
2577                 if (rc)
2578                         return rc;
2579         }
2580
2581         if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE))
2582                 fuse_link_write_file(file);
2583
2584         file_accessed(file);
2585         vma->vm_ops = &fuse_file_vm_ops;
2586         return 0;
2587 }
2588
2589 static int convert_fuse_file_lock(struct fuse_conn *fc,
2590                                   const struct fuse_file_lock *ffl,
2591                                   struct file_lock *fl)
2592 {
2593         switch (ffl->type) {
2594         case F_UNLCK:
2595                 break;
2596
2597         case F_RDLCK:
2598         case F_WRLCK:
2599                 if (ffl->start > OFFSET_MAX || ffl->end > OFFSET_MAX ||
2600                     ffl->end < ffl->start)
2601                         return -EIO;
2602
2603                 fl->fl_start = ffl->start;
2604                 fl->fl_end = ffl->end;
2605
2606                 /*
2607                  * Convert pid into init's pid namespace.  The locks API will
2608                  * translate it into the caller's pid namespace.
2609                  */
2610                 rcu_read_lock();
2611                 fl->c.flc_pid = pid_nr_ns(find_pid_ns(ffl->pid, fc->pid_ns), &init_pid_ns);
2612                 rcu_read_unlock();
2613                 break;
2614
2615         default:
2616                 return -EIO;
2617         }
2618         fl->c.flc_type = ffl->type;
2619         return 0;
2620 }
2621
2622 static void fuse_lk_fill(struct fuse_args *args, struct file *file,
2623                          const struct file_lock *fl, int opcode, pid_t pid,
2624                          int flock, struct fuse_lk_in *inarg)
2625 {
2626         struct inode *inode = file_inode(file);
2627         struct fuse_conn *fc = get_fuse_conn(inode);
2628         struct fuse_file *ff = file->private_data;
2629
2630         memset(inarg, 0, sizeof(*inarg));
2631         inarg->fh = ff->fh;
2632         inarg->owner = fuse_lock_owner_id(fc, fl->c.flc_owner);
2633         inarg->lk.start = fl->fl_start;
2634         inarg->lk.end = fl->fl_end;
2635         inarg->lk.type = fl->c.flc_type;
2636         inarg->lk.pid = pid;
2637         if (flock)
2638                 inarg->lk_flags |= FUSE_LK_FLOCK;
2639         args->opcode = opcode;
2640         args->nodeid = get_node_id(inode);
2641         args->in_numargs = 1;
2642         args->in_args[0].size = sizeof(*inarg);
2643         args->in_args[0].value = inarg;
2644 }
2645
2646 static int fuse_getlk(struct file *file, struct file_lock *fl)
2647 {
2648         struct inode *inode = file_inode(file);
2649         struct fuse_mount *fm = get_fuse_mount(inode);
2650         FUSE_ARGS(args);
2651         struct fuse_lk_in inarg;
2652         struct fuse_lk_out outarg;
2653         int err;
2654
2655         fuse_lk_fill(&args, file, fl, FUSE_GETLK, 0, 0, &inarg);
2656         args.out_numargs = 1;
2657         args.out_args[0].size = sizeof(outarg);
2658         args.out_args[0].value = &outarg;
2659         err = fuse_simple_request(fm, &args);
2660         if (!err)
2661                 err = convert_fuse_file_lock(fm->fc, &outarg.lk, fl);
2662
2663         return err;
2664 }
2665
2666 static int fuse_setlk(struct file *file, struct file_lock *fl, int flock)
2667 {
2668         struct inode *inode = file_inode(file);
2669         struct fuse_mount *fm = get_fuse_mount(inode);
2670         FUSE_ARGS(args);
2671         struct fuse_lk_in inarg;
2672         int opcode = (fl->c.flc_flags & FL_SLEEP) ? FUSE_SETLKW : FUSE_SETLK;
2673         struct pid *pid = fl->c.flc_type != F_UNLCK ? task_tgid(current) : NULL;
2674         pid_t pid_nr = pid_nr_ns(pid, fm->fc->pid_ns);
2675         int err;
2676
2677         if (fl->fl_lmops && fl->fl_lmops->lm_grant) {
2678                 /* NLM needs asynchronous locks, which we don't support yet */
2679                 return -ENOLCK;
2680         }
2681
2682         fuse_lk_fill(&args, file, fl, opcode, pid_nr, flock, &inarg);
2683         err = fuse_simple_request(fm, &args);
2684
2685         /* locking is restartable */
2686         if (err == -EINTR)
2687                 err = -ERESTARTSYS;
2688
2689         return err;
2690 }
2691
2692 static int fuse_file_lock(struct file *file, int cmd, struct file_lock *fl)
2693 {
2694         struct inode *inode = file_inode(file);
2695         struct fuse_conn *fc = get_fuse_conn(inode);
2696         int err;
2697
2698         if (cmd == F_CANCELLK) {
2699                 err = 0;
2700         } else if (cmd == F_GETLK) {
2701                 if (fc->no_lock) {
2702                         posix_test_lock(file, fl);
2703                         err = 0;
2704                 } else
2705                         err = fuse_getlk(file, fl);
2706         } else {
2707                 if (fc->no_lock)
2708                         err = posix_lock_file(file, fl, NULL);
2709                 else
2710                         err = fuse_setlk(file, fl, 0);
2711         }
2712         return err;
2713 }
2714
2715 static int fuse_file_flock(struct file *file, int cmd, struct file_lock *fl)
2716 {
2717         struct inode *inode = file_inode(file);
2718         struct fuse_conn *fc = get_fuse_conn(inode);
2719         int err;
2720
2721         if (fc->no_flock) {
2722                 err = locks_lock_file_wait(file, fl);
2723         } else {
2724                 struct fuse_file *ff = file->private_data;
2725
2726                 /* emulate flock with POSIX locks */
2727                 ff->flock = true;
2728                 err = fuse_setlk(file, fl, 1);
2729         }
2730
2731         return err;
2732 }
2733
2734 static sector_t fuse_bmap(struct address_space *mapping, sector_t block)
2735 {
2736         struct inode *inode = mapping->host;
2737         struct fuse_mount *fm = get_fuse_mount(inode);
2738         FUSE_ARGS(args);
2739         struct fuse_bmap_in inarg;
2740         struct fuse_bmap_out outarg;
2741         int err;
2742
2743         if (!inode->i_sb->s_bdev || fm->fc->no_bmap)
2744                 return 0;
2745
2746         memset(&inarg, 0, sizeof(inarg));
2747         inarg.block = block;
2748         inarg.blocksize = inode->i_sb->s_blocksize;
2749         args.opcode = FUSE_BMAP;
2750         args.nodeid = get_node_id(inode);
2751         args.in_numargs = 1;
2752         args.in_args[0].size = sizeof(inarg);
2753         args.in_args[0].value = &inarg;
2754         args.out_numargs = 1;
2755         args.out_args[0].size = sizeof(outarg);
2756         args.out_args[0].value = &outarg;
2757         err = fuse_simple_request(fm, &args);
2758         if (err == -ENOSYS)
2759                 fm->fc->no_bmap = 1;
2760
2761         return err ? 0 : outarg.block;
2762 }
2763
2764 static loff_t fuse_lseek(struct file *file, loff_t offset, int whence)
2765 {
2766         struct inode *inode = file->f_mapping->host;
2767         struct fuse_mount *fm = get_fuse_mount(inode);
2768         struct fuse_file *ff = file->private_data;
2769         FUSE_ARGS(args);
2770         struct fuse_lseek_in inarg = {
2771                 .fh = ff->fh,
2772                 .offset = offset,
2773                 .whence = whence
2774         };
2775         struct fuse_lseek_out outarg;
2776         int err;
2777
2778         if (fm->fc->no_lseek)
2779                 goto fallback;
2780
2781         args.opcode = FUSE_LSEEK;
2782         args.nodeid = ff->nodeid;
2783         args.in_numargs = 1;
2784         args.in_args[0].size = sizeof(inarg);
2785         args.in_args[0].value = &inarg;
2786         args.out_numargs = 1;
2787         args.out_args[0].size = sizeof(outarg);
2788         args.out_args[0].value = &outarg;
2789         err = fuse_simple_request(fm, &args);
2790         if (err) {
2791                 if (err == -ENOSYS) {
2792                         fm->fc->no_lseek = 1;
2793                         goto fallback;
2794                 }
2795                 return err;
2796         }
2797
2798         return vfs_setpos(file, outarg.offset, inode->i_sb->s_maxbytes);
2799
2800 fallback:
2801         err = fuse_update_attributes(inode, file, STATX_SIZE);
2802         if (!err)
2803                 return generic_file_llseek(file, offset, whence);
2804         else
2805                 return err;
2806 }
2807
2808 static loff_t fuse_file_llseek(struct file *file, loff_t offset, int whence)
2809 {
2810         loff_t retval;
2811         struct inode *inode = file_inode(file);
2812
2813         switch (whence) {
2814         case SEEK_SET:
2815         case SEEK_CUR:
2816                  /* No i_mutex protection necessary for SEEK_CUR and SEEK_SET */
2817                 retval = generic_file_llseek(file, offset, whence);
2818                 break;
2819         case SEEK_END:
2820                 inode_lock(inode);
2821                 retval = fuse_update_attributes(inode, file, STATX_SIZE);
2822                 if (!retval)
2823                         retval = generic_file_llseek(file, offset, whence);
2824                 inode_unlock(inode);
2825                 break;
2826         case SEEK_HOLE:
2827         case SEEK_DATA:
2828                 inode_lock(inode);
2829                 retval = fuse_lseek(file, offset, whence);
2830                 inode_unlock(inode);
2831                 break;
2832         default:
2833                 retval = -EINVAL;
2834         }
2835
2836         return retval;
2837 }
2838
2839 /*
2840  * All files which have been polled are linked to RB tree
2841  * fuse_conn->polled_files which is indexed by kh.  Walk the tree and
2842  * find the matching one.
2843  */
2844 static struct rb_node **fuse_find_polled_node(struct fuse_conn *fc, u64 kh,
2845                                               struct rb_node **parent_out)
2846 {
2847         struct rb_node **link = &fc->polled_files.rb_node;
2848         struct rb_node *last = NULL;
2849
2850         while (*link) {
2851                 struct fuse_file *ff;
2852
2853                 last = *link;
2854                 ff = rb_entry(last, struct fuse_file, polled_node);
2855
2856                 if (kh < ff->kh)
2857                         link = &last->rb_left;
2858                 else if (kh > ff->kh)
2859                         link = &last->rb_right;
2860                 else
2861                         return link;
2862         }
2863
2864         if (parent_out)
2865                 *parent_out = last;
2866         return link;
2867 }
2868
2869 /*
2870  * The file is about to be polled.  Make sure it's on the polled_files
2871  * RB tree.  Note that files once added to the polled_files tree are
2872  * not removed before the file is released.  This is because a file
2873  * polled once is likely to be polled again.
2874  */
2875 static void fuse_register_polled_file(struct fuse_conn *fc,
2876                                       struct fuse_file *ff)
2877 {
2878         spin_lock(&fc->lock);
2879         if (RB_EMPTY_NODE(&ff->polled_node)) {
2880                 struct rb_node **link, *parent;
2881
2882                 link = fuse_find_polled_node(fc, ff->kh, &parent);
2883                 BUG_ON(*link);
2884                 rb_link_node(&ff->polled_node, parent, link);
2885                 rb_insert_color(&ff->polled_node, &fc->polled_files);
2886         }
2887         spin_unlock(&fc->lock);
2888 }
2889
2890 __poll_t fuse_file_poll(struct file *file, poll_table *wait)
2891 {
2892         struct fuse_file *ff = file->private_data;
2893         struct fuse_mount *fm = ff->fm;
2894         struct fuse_poll_in inarg = { .fh = ff->fh, .kh = ff->kh };
2895         struct fuse_poll_out outarg;
2896         FUSE_ARGS(args);
2897         int err;
2898
2899         if (fm->fc->no_poll)
2900                 return DEFAULT_POLLMASK;
2901
2902         poll_wait(file, &ff->poll_wait, wait);
2903         inarg.events = mangle_poll(poll_requested_events(wait));
2904
2905         /*
2906          * Ask for notification iff there's someone waiting for it.
2907          * The client may ignore the flag and always notify.
2908          */
2909         if (waitqueue_active(&ff->poll_wait)) {
2910                 inarg.flags |= FUSE_POLL_SCHEDULE_NOTIFY;
2911                 fuse_register_polled_file(fm->fc, ff);
2912         }
2913
2914         args.opcode = FUSE_POLL;
2915         args.nodeid = ff->nodeid;
2916         args.in_numargs = 1;
2917         args.in_args[0].size = sizeof(inarg);
2918         args.in_args[0].value = &inarg;
2919         args.out_numargs = 1;
2920         args.out_args[0].size = sizeof(outarg);
2921         args.out_args[0].value = &outarg;
2922         err = fuse_simple_request(fm, &args);
2923
2924         if (!err)
2925                 return demangle_poll(outarg.revents);
2926         if (err == -ENOSYS) {
2927                 fm->fc->no_poll = 1;
2928                 return DEFAULT_POLLMASK;
2929         }
2930         return EPOLLERR;
2931 }
2932 EXPORT_SYMBOL_GPL(fuse_file_poll);
2933
2934 /*
2935  * This is called from fuse_handle_notify() on FUSE_NOTIFY_POLL and
2936  * wakes up the poll waiters.
2937  */
2938 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
2939                             struct fuse_notify_poll_wakeup_out *outarg)
2940 {
2941         u64 kh = outarg->kh;
2942         struct rb_node **link;
2943
2944         spin_lock(&fc->lock);
2945
2946         link = fuse_find_polled_node(fc, kh, NULL);
2947         if (*link) {
2948                 struct fuse_file *ff;
2949
2950                 ff = rb_entry(*link, struct fuse_file, polled_node);
2951                 wake_up_interruptible_sync(&ff->poll_wait);
2952         }
2953
2954         spin_unlock(&fc->lock);
2955         return 0;
2956 }
2957
2958 static void fuse_do_truncate(struct file *file)
2959 {
2960         struct inode *inode = file->f_mapping->host;
2961         struct iattr attr;
2962
2963         attr.ia_valid = ATTR_SIZE;
2964         attr.ia_size = i_size_read(inode);
2965
2966         attr.ia_file = file;
2967         attr.ia_valid |= ATTR_FILE;
2968
2969         fuse_do_setattr(file_dentry(file), &attr, file);
2970 }
2971
2972 static inline loff_t fuse_round_up(struct fuse_conn *fc, loff_t off)
2973 {
2974         return round_up(off, fc->max_pages << PAGE_SHIFT);
2975 }
2976
2977 static ssize_t
2978 fuse_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
2979 {
2980         DECLARE_COMPLETION_ONSTACK(wait);
2981         ssize_t ret = 0;
2982         struct file *file = iocb->ki_filp;
2983         struct fuse_file *ff = file->private_data;
2984         loff_t pos = 0;
2985         struct inode *inode;
2986         loff_t i_size;
2987         size_t count = iov_iter_count(iter), shortened = 0;
2988         loff_t offset = iocb->ki_pos;
2989         struct fuse_io_priv *io;
2990
2991         pos = offset;
2992         inode = file->f_mapping->host;
2993         i_size = i_size_read(inode);
2994
2995         if ((iov_iter_rw(iter) == READ) && (offset >= i_size))
2996                 return 0;
2997
2998         io = kmalloc(sizeof(struct fuse_io_priv), GFP_KERNEL);
2999         if (!io)
3000                 return -ENOMEM;
3001         spin_lock_init(&io->lock);
3002         kref_init(&io->refcnt);
3003         io->reqs = 1;
3004         io->bytes = -1;
3005         io->size = 0;
3006         io->offset = offset;
3007         io->write = (iov_iter_rw(iter) == WRITE);
3008         io->err = 0;
3009         /*
3010          * By default, we want to optimize all I/Os with async request
3011          * submission to the client filesystem if supported.
3012          */
3013         io->async = ff->fm->fc->async_dio;
3014         io->iocb = iocb;
3015         io->blocking = is_sync_kiocb(iocb);
3016
3017         /* optimization for short read */
3018         if (io->async && !io->write && offset + count > i_size) {
3019                 iov_iter_truncate(iter, fuse_round_up(ff->fm->fc, i_size - offset));
3020                 shortened = count - iov_iter_count(iter);
3021                 count -= shortened;
3022         }
3023
3024         /*
3025          * We cannot asynchronously extend the size of a file.
3026          * In such case the aio will behave exactly like sync io.
3027          */
3028         if ((offset + count > i_size) && io->write)
3029                 io->blocking = true;
3030
3031         if (io->async && io->blocking) {
3032                 /*
3033                  * Additional reference to keep io around after
3034                  * calling fuse_aio_complete()
3035                  */
3036                 kref_get(&io->refcnt);
3037                 io->done = &wait;
3038         }
3039
3040         if (iov_iter_rw(iter) == WRITE) {
3041                 ret = fuse_direct_io(io, iter, &pos, FUSE_DIO_WRITE);
3042                 fuse_invalidate_attr_mask(inode, FUSE_STATX_MODSIZE);
3043         } else {
3044                 ret = __fuse_direct_read(io, iter, &pos);
3045         }
3046         iov_iter_reexpand(iter, iov_iter_count(iter) + shortened);
3047
3048         if (io->async) {
3049                 bool blocking = io->blocking;
3050
3051                 fuse_aio_complete(io, ret < 0 ? ret : 0, -1);
3052
3053                 /* we have a non-extending, async request, so return */
3054                 if (!blocking)
3055                         return -EIOCBQUEUED;
3056
3057                 wait_for_completion(&wait);
3058                 ret = fuse_get_res_by_io(io);
3059         }
3060
3061         kref_put(&io->refcnt, fuse_io_release);
3062
3063         if (iov_iter_rw(iter) == WRITE) {
3064                 fuse_write_update_attr(inode, pos, ret);
3065                 /* For extending writes we already hold exclusive lock */
3066                 if (ret < 0 && offset + count > i_size)
3067                         fuse_do_truncate(file);
3068         }
3069
3070         return ret;
3071 }
3072
3073 static int fuse_writeback_range(struct inode *inode, loff_t start, loff_t end)
3074 {
3075         int err = filemap_write_and_wait_range(inode->i_mapping, start, LLONG_MAX);
3076
3077         if (!err)
3078                 fuse_sync_writes(inode);
3079
3080         return err;
3081 }
3082
3083 static long fuse_file_fallocate(struct file *file, int mode, loff_t offset,
3084                                 loff_t length)
3085 {
3086         struct fuse_file *ff = file->private_data;
3087         struct inode *inode = file_inode(file);
3088         struct fuse_inode *fi = get_fuse_inode(inode);
3089         struct fuse_mount *fm = ff->fm;
3090         FUSE_ARGS(args);
3091         struct fuse_fallocate_in inarg = {
3092                 .fh = ff->fh,
3093                 .offset = offset,
3094                 .length = length,
3095                 .mode = mode
3096         };
3097         int err;
3098         bool block_faults = FUSE_IS_DAX(inode) &&
3099                 (!(mode & FALLOC_FL_KEEP_SIZE) ||
3100                  (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)));
3101
3102         if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
3103                      FALLOC_FL_ZERO_RANGE))
3104                 return -EOPNOTSUPP;
3105
3106         if (fm->fc->no_fallocate)
3107                 return -EOPNOTSUPP;
3108
3109         inode_lock(inode);
3110         if (block_faults) {
3111                 filemap_invalidate_lock(inode->i_mapping);
3112                 err = fuse_dax_break_layouts(inode, 0, 0);
3113                 if (err)
3114                         goto out;
3115         }
3116
3117         if (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) {
3118                 loff_t endbyte = offset + length - 1;
3119
3120                 err = fuse_writeback_range(inode, offset, endbyte);
3121                 if (err)
3122                         goto out;
3123         }
3124
3125         if (!(mode & FALLOC_FL_KEEP_SIZE) &&
3126             offset + length > i_size_read(inode)) {
3127                 err = inode_newsize_ok(inode, offset + length);
3128                 if (err)
3129                         goto out;
3130         }
3131
3132         err = file_modified(file);
3133         if (err)
3134                 goto out;
3135
3136         if (!(mode & FALLOC_FL_KEEP_SIZE))
3137                 set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
3138
3139         args.opcode = FUSE_FALLOCATE;
3140         args.nodeid = ff->nodeid;
3141         args.in_numargs = 1;
3142         args.in_args[0].size = sizeof(inarg);
3143         args.in_args[0].value = &inarg;
3144         err = fuse_simple_request(fm, &args);
3145         if (err == -ENOSYS) {
3146                 fm->fc->no_fallocate = 1;
3147                 err = -EOPNOTSUPP;
3148         }
3149         if (err)
3150                 goto out;
3151
3152         /* we could have extended the file */
3153         if (!(mode & FALLOC_FL_KEEP_SIZE)) {
3154                 if (fuse_write_update_attr(inode, offset + length, length))
3155                         file_update_time(file);
3156         }
3157
3158         if (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
3159                 truncate_pagecache_range(inode, offset, offset + length - 1);
3160
3161         fuse_invalidate_attr_mask(inode, FUSE_STATX_MODSIZE);
3162
3163 out:
3164         if (!(mode & FALLOC_FL_KEEP_SIZE))
3165                 clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
3166
3167         if (block_faults)
3168                 filemap_invalidate_unlock(inode->i_mapping);
3169
3170         inode_unlock(inode);
3171
3172         fuse_flush_time_update(inode);
3173
3174         return err;
3175 }
3176
3177 static ssize_t __fuse_copy_file_range(struct file *file_in, loff_t pos_in,
3178                                       struct file *file_out, loff_t pos_out,
3179                                       size_t len, unsigned int flags)
3180 {
3181         struct fuse_file *ff_in = file_in->private_data;
3182         struct fuse_file *ff_out = file_out->private_data;
3183         struct inode *inode_in = file_inode(file_in);
3184         struct inode *inode_out = file_inode(file_out);
3185         struct fuse_inode *fi_out = get_fuse_inode(inode_out);
3186         struct fuse_mount *fm = ff_in->fm;
3187         struct fuse_conn *fc = fm->fc;
3188         FUSE_ARGS(args);
3189         struct fuse_copy_file_range_in inarg = {
3190                 .fh_in = ff_in->fh,
3191                 .off_in = pos_in,
3192                 .nodeid_out = ff_out->nodeid,
3193                 .fh_out = ff_out->fh,
3194                 .off_out = pos_out,
3195                 .len = len,
3196                 .flags = flags
3197         };
3198         struct fuse_write_out outarg;
3199         ssize_t err;
3200         /* mark unstable when write-back is not used, and file_out gets
3201          * extended */
3202         bool is_unstable = (!fc->writeback_cache) &&
3203                            ((pos_out + len) > inode_out->i_size);
3204
3205         if (fc->no_copy_file_range)
3206                 return -EOPNOTSUPP;
3207
3208         if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb)
3209                 return -EXDEV;
3210
3211         inode_lock(inode_in);
3212         err = fuse_writeback_range(inode_in, pos_in, pos_in + len - 1);
3213         inode_unlock(inode_in);
3214         if (err)
3215                 return err;
3216
3217         inode_lock(inode_out);
3218
3219         err = file_modified(file_out);
3220         if (err)
3221                 goto out;
3222
3223         /*
3224          * Write out dirty pages in the destination file before sending the COPY
3225          * request to userspace.  After the request is completed, truncate off
3226          * pages (including partial ones) from the cache that have been copied,
3227          * since these contain stale data at that point.
3228          *
3229          * This should be mostly correct, but if the COPY writes to partial
3230          * pages (at the start or end) and the parts not covered by the COPY are
3231          * written through a memory map after calling fuse_writeback_range(),
3232          * then these partial page modifications will be lost on truncation.
3233          *
3234          * It is unlikely that someone would rely on such mixed style
3235          * modifications.  Yet this does give less guarantees than if the
3236          * copying was performed with write(2).
3237          *
3238          * To fix this a mapping->invalidate_lock could be used to prevent new
3239          * faults while the copy is ongoing.
3240          */
3241         err = fuse_writeback_range(inode_out, pos_out, pos_out + len - 1);
3242         if (err)
3243                 goto out;
3244
3245         if (is_unstable)
3246                 set_bit(FUSE_I_SIZE_UNSTABLE, &fi_out->state);
3247
3248         args.opcode = FUSE_COPY_FILE_RANGE;
3249         args.nodeid = ff_in->nodeid;
3250         args.in_numargs = 1;
3251         args.in_args[0].size = sizeof(inarg);
3252         args.in_args[0].value = &inarg;
3253         args.out_numargs = 1;
3254         args.out_args[0].size = sizeof(outarg);
3255         args.out_args[0].value = &outarg;
3256         err = fuse_simple_request(fm, &args);
3257         if (err == -ENOSYS) {
3258                 fc->no_copy_file_range = 1;
3259                 err = -EOPNOTSUPP;
3260         }
3261         if (err)
3262                 goto out;
3263
3264         truncate_inode_pages_range(inode_out->i_mapping,
3265                                    ALIGN_DOWN(pos_out, PAGE_SIZE),
3266                                    ALIGN(pos_out + outarg.size, PAGE_SIZE) - 1);
3267
3268         file_update_time(file_out);
3269         fuse_write_update_attr(inode_out, pos_out + outarg.size, outarg.size);
3270
3271         err = outarg.size;
3272 out:
3273         if (is_unstable)
3274                 clear_bit(FUSE_I_SIZE_UNSTABLE, &fi_out->state);
3275
3276         inode_unlock(inode_out);
3277         file_accessed(file_in);
3278
3279         fuse_flush_time_update(inode_out);
3280
3281         return err;
3282 }
3283
3284 static ssize_t fuse_copy_file_range(struct file *src_file, loff_t src_off,
3285                                     struct file *dst_file, loff_t dst_off,
3286                                     size_t len, unsigned int flags)
3287 {
3288         ssize_t ret;
3289
3290         ret = __fuse_copy_file_range(src_file, src_off, dst_file, dst_off,
3291                                      len, flags);
3292
3293         if (ret == -EOPNOTSUPP || ret == -EXDEV)
3294                 ret = splice_copy_file_range(src_file, src_off, dst_file,
3295                                              dst_off, len);
3296         return ret;
3297 }
3298
3299 static const struct file_operations fuse_file_operations = {
3300         .llseek         = fuse_file_llseek,
3301         .read_iter      = fuse_file_read_iter,
3302         .write_iter     = fuse_file_write_iter,
3303         .mmap           = fuse_file_mmap,
3304         .open           = fuse_open,
3305         .flush          = fuse_flush,
3306         .release        = fuse_release,
3307         .fsync          = fuse_fsync,
3308         .lock           = fuse_file_lock,
3309         .get_unmapped_area = thp_get_unmapped_area,
3310         .flock          = fuse_file_flock,
3311         .splice_read    = fuse_splice_read,
3312         .splice_write   = fuse_splice_write,
3313         .unlocked_ioctl = fuse_file_ioctl,
3314         .compat_ioctl   = fuse_file_compat_ioctl,
3315         .poll           = fuse_file_poll,
3316         .fallocate      = fuse_file_fallocate,
3317         .copy_file_range = fuse_copy_file_range,
3318 };
3319
3320 static const struct address_space_operations fuse_file_aops  = {
3321         .read_folio     = fuse_read_folio,
3322         .readahead      = fuse_readahead,
3323         .writepages     = fuse_writepages,
3324         .launder_folio  = fuse_launder_folio,
3325         .dirty_folio    = filemap_dirty_folio,
3326         .migrate_folio  = filemap_migrate_folio,
3327         .bmap           = fuse_bmap,
3328         .direct_IO      = fuse_direct_IO,
3329         .write_begin    = fuse_write_begin,
3330         .write_end      = fuse_write_end,
3331 };
3332
3333 void fuse_init_file_inode(struct inode *inode, unsigned int flags)
3334 {
3335         struct fuse_inode *fi = get_fuse_inode(inode);
3336
3337         inode->i_fop = &fuse_file_operations;
3338         inode->i_data.a_ops = &fuse_file_aops;
3339
3340         INIT_LIST_HEAD(&fi->write_files);
3341         INIT_LIST_HEAD(&fi->queued_writes);
3342         fi->writectr = 0;
3343         fi->iocachectr = 0;
3344         init_waitqueue_head(&fi->page_waitq);
3345         init_waitqueue_head(&fi->direct_io_waitq);
3346         fi->writepages = RB_ROOT;
3347
3348         if (IS_ENABLED(CONFIG_FUSE_DAX))
3349                 fuse_dax_inode_init(inode, flags);
3350 }