file.c 75.1 KB
Newer Older
1 2
/*
  FUSE: Filesystem in Userspace
Miklos Szeredi's avatar
Miklos Szeredi committed
3
  Copyright (C) 2001-2008  Miklos Szeredi <miklos@szeredi.hu>
4 5 6 7 8 9 10 11 12 13

  This program can be distributed under the terms of the GNU GPL.
  See the file COPYING.
*/

#include "fuse_i.h"

#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/kernel.h>
Alexey Dobriyan's avatar
Alexey Dobriyan committed
14
#include <linux/sched.h>
15
#include <linux/module.h>
16
#include <linux/compat.h>
17
#include <linux/swap.h>
18
#include <linux/falloc.h>
19
#include <linux/uio.h>
20

21
static const struct file_operations fuse_direct_io_file_operations;
22

23 24
static int fuse_send_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
			  int opcode, struct fuse_open_out *outargp)
25 26
{
	struct fuse_open_in inarg;
27
	FUSE_ARGS(args);
28 29

	memset(&inarg, 0, sizeof(inarg));
30 31 32
	inarg.flags = file->f_flags & ~(O_CREAT | O_EXCL | O_NOCTTY);
	if (!fc->atomic_o_trunc)
		inarg.flags &= ~O_TRUNC;
33 34 35 36 37 38 39 40
	args.in.h.opcode = opcode;
	args.in.h.nodeid = nodeid;
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.out.numargs = 1;
	args.out.args[0].size = sizeof(*outargp);
	args.out.args[0].value = outargp;
41

42
	return fuse_simple_request(fc, &args);
43 44
}

Tejun Heo's avatar
Tejun Heo committed
45
struct fuse_file *fuse_file_alloc(struct fuse_conn *fc)
46 47
{
	struct fuse_file *ff;
Tejun Heo's avatar
Tejun Heo committed
48

49
	ff = kmalloc(sizeof(struct fuse_file), GFP_KERNEL);
Tejun Heo's avatar
Tejun Heo committed
50 51 52
	if (unlikely(!ff))
		return NULL;

53
	ff->fc = fc;
54
	ff->reserved_req = fuse_request_alloc(0);
Tejun Heo's avatar
Tejun Heo committed
55 56 57
	if (unlikely(!ff->reserved_req)) {
		kfree(ff);
		return NULL;
58
	}
Tejun Heo's avatar
Tejun Heo committed
59 60 61 62 63 64 65 66 67 68

	INIT_LIST_HEAD(&ff->write_entry);
	atomic_set(&ff->count, 0);
	RB_CLEAR_NODE(&ff->polled_node);
	init_waitqueue_head(&ff->poll_wait);

	spin_lock(&fc->lock);
	ff->kh = ++fc->khctr;
	spin_unlock(&fc->lock);

69 70 71 72 73
	return ff;
}

void fuse_file_free(struct fuse_file *ff)
{
74
	fuse_request_free(ff->reserved_req);
75 76 77
	kfree(ff);
}

78
struct fuse_file *fuse_file_get(struct fuse_file *ff)
79 80 81 82 83
{
	atomic_inc(&ff->count);
	return ff;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
84 85
static void fuse_release_end(struct fuse_conn *fc, struct fuse_req *req)
{
86
	iput(req->misc.release.inode);
Miklos Szeredi's avatar
Miklos Szeredi committed
87 88
}

89
static void fuse_file_put(struct fuse_file *ff, bool sync)
90 91 92
{
	if (atomic_dec_and_test(&ff->count)) {
		struct fuse_req *req = ff->reserved_req;
93

94 95 96 97 98
		if (ff->fc->no_open) {
			/*
			 * Drop the release request when client does not
			 * implement 'open'
			 */
Miklos Szeredi's avatar
Miklos Szeredi committed
99
			__clear_bit(FR_BACKGROUND, &req->flags);
100
			iput(req->misc.release.inode);
101 102
			fuse_put_request(ff->fc, req);
		} else if (sync) {
Miklos Szeredi's avatar
Miklos Szeredi committed
103
			__clear_bit(FR_BACKGROUND, &req->flags);
104
			fuse_request_send(ff->fc, req);
105
			iput(req->misc.release.inode);
106 107 108
			fuse_put_request(ff->fc, req);
		} else {
			req->end = fuse_release_end;
Miklos Szeredi's avatar
Miklos Szeredi committed
109
			__set_bit(FR_BACKGROUND, &req->flags);
110 111
			fuse_request_send_background(ff->fc, req);
		}
112 113 114 115
		kfree(ff);
	}
}

116 117
int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
		 bool isdir)
118 119 120 121 122 123 124 125
{
	struct fuse_file *ff;
	int opcode = isdir ? FUSE_OPENDIR : FUSE_OPEN;

	ff = fuse_file_alloc(fc);
	if (!ff)
		return -ENOMEM;

126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142
	ff->fh = 0;
	ff->open_flags = FOPEN_KEEP_CACHE; /* Default for no-open */
	if (!fc->no_open || isdir) {
		struct fuse_open_out outarg;
		int err;

		err = fuse_send_open(fc, nodeid, file, opcode, &outarg);
		if (!err) {
			ff->fh = outarg.fh;
			ff->open_flags = outarg.open_flags;

		} else if (err != -ENOSYS || isdir) {
			fuse_file_free(ff);
			return err;
		} else {
			fc->no_open = 1;
		}
143 144 145
	}

	if (isdir)
146
		ff->open_flags &= ~FOPEN_DIRECT_IO;
147 148 149 150 151 152

	ff->nodeid = nodeid;
	file->private_data = fuse_file_get(ff);

	return 0;
}
153
EXPORT_SYMBOL_GPL(fuse_do_open);
154

155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
static void fuse_link_write_file(struct file *file)
{
	struct inode *inode = file_inode(file);
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
	struct fuse_file *ff = file->private_data;
	/*
	 * file may be written through mmap, so chain it onto the
	 * inodes's write_file list
	 */
	spin_lock(&fc->lock);
	if (list_empty(&ff->write_entry))
		list_add(&ff->write_entry, &fi->write_files);
	spin_unlock(&fc->lock);
}

171
void fuse_finish_open(struct inode *inode, struct file *file)
172
{
173
	struct fuse_file *ff = file->private_data;
174
	struct fuse_conn *fc = get_fuse_conn(inode);
175 176

	if (ff->open_flags & FOPEN_DIRECT_IO)
177
		file->f_op = &fuse_direct_io_file_operations;
178
	if (!(ff->open_flags & FOPEN_KEEP_CACHE))
Miklos Szeredi's avatar
Miklos Szeredi committed
179
		invalidate_inode_pages2(inode->i_mapping);
180
	if (ff->open_flags & FOPEN_NONSEEKABLE)
Tejun Heo's avatar
Tejun Heo committed
181
		nonseekable_open(inode, file);
182 183 184 185 186 187 188 189
	if (fc->atomic_o_trunc && (file->f_flags & O_TRUNC)) {
		struct fuse_inode *fi = get_fuse_inode(inode);

		spin_lock(&fc->lock);
		fi->attr_version = ++fc->attr_version;
		i_size_write(inode, 0);
		spin_unlock(&fc->lock);
		fuse_invalidate_attr(inode);
190 191
		if (fc->writeback_cache)
			file_update_time(file);
192
	}
193 194
	if ((file->f_mode & FMODE_WRITE) && fc->writeback_cache)
		fuse_link_write_file(file);
195 196
}

197
int fuse_open_common(struct inode *inode, struct file *file, bool isdir)
198
{
Tejun Heo's avatar
Tejun Heo committed
199
	struct fuse_conn *fc = get_fuse_conn(inode);
200
	int err;
201 202 203
	bool lock_inode = (file->f_flags & O_TRUNC) &&
			  fc->atomic_o_trunc &&
			  fc->writeback_cache;
204 205 206 207 208

	err = generic_file_open(inode, file);
	if (err)
		return err;

209
	if (lock_inode)
Al Viro's avatar
Al Viro committed
210
		inode_lock(inode);
211

212
	err = fuse_do_open(fc, get_node_id(inode), file, isdir);
213

214 215
	if (!err)
		fuse_finish_open(inode, file);
216

217
	if (lock_inode)
Al Viro's avatar
Al Viro committed
218
		inode_unlock(inode);
219 220

	return err;
221 222
}

223
static void fuse_prepare_release(struct fuse_file *ff, int flags, int opcode)
224
{
225
	struct fuse_conn *fc = ff->fc;
226
	struct fuse_req *req = ff->reserved_req;
227
	struct fuse_release_in *inarg = &req->misc.release.in;
228

229 230 231 232 233 234
	spin_lock(&fc->lock);
	list_del(&ff->write_entry);
	if (!RB_EMPTY_NODE(&ff->polled_node))
		rb_erase(&ff->polled_node, &fc->polled_files);
	spin_unlock(&fc->lock);

235
	wake_up_interruptible_all(&ff->poll_wait);
236

237
	inarg->fh = ff->fh;
238
	inarg->flags = flags;
239
	req->in.h.opcode = opcode;
240
	req->in.h.nodeid = ff->nodeid;
241 242 243
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(struct fuse_release_in);
	req->in.args[0].value = inarg;
244 245
}

246
void fuse_release_common(struct file *file, int opcode)
247
{
Tejun Heo's avatar
Tejun Heo committed
248 249
	struct fuse_file *ff;
	struct fuse_req *req;
250

Tejun Heo's avatar
Tejun Heo committed
251 252
	ff = file->private_data;
	if (unlikely(!ff))
253
		return;
Tejun Heo's avatar
Tejun Heo committed
254 255

	req = ff->reserved_req;
256
	fuse_prepare_release(ff, file->f_flags, opcode);
Tejun Heo's avatar
Tejun Heo committed
257

Miklos Szeredi's avatar
Miklos Szeredi committed
258 259 260 261 262 263
	if (ff->flock) {
		struct fuse_release_in *inarg = &req->misc.release.in;
		inarg->release_flags |= FUSE_RELEASE_FLOCK_UNLOCK;
		inarg->lock_owner = fuse_lock_owner_id(ff->fc,
						       (fl_owner_t) file);
	}
264 265
	/* Hold inode until release is finished */
	req->misc.release.inode = igrab(file_inode(file));
Tejun Heo's avatar
Tejun Heo committed
266 267 268 269 270

	/*
	 * Normally this will send the RELEASE request, however if
	 * some asynchronous READ or WRITE requests are outstanding,
	 * the sending will be delayed.
271 272 273 274
	 *
	 * Make the release synchronous if this is a fuseblk mount,
	 * synchronous RELEASE is allowed (and desirable) in this case
	 * because the server can be trusted not to screw up.
Tejun Heo's avatar
Tejun Heo committed
275
	 */
276
	fuse_file_put(ff, ff->fc->destroy_req != NULL);
277 278
}

279 280
static int fuse_open(struct inode *inode, struct file *file)
{
281
	return fuse_open_common(inode, file, false);
282 283 284 285
}

static int fuse_release(struct inode *inode, struct file *file)
{
286 287 288 289
	struct fuse_conn *fc = get_fuse_conn(inode);

	/* see fuse_vma_close() for !writeback_cache case */
	if (fc->writeback_cache)
Miklos Szeredi's avatar
Miklos Szeredi committed
290
		write_inode_now(inode, 1);
291

292 293 294 295 296 297 298 299 300 301
	fuse_release_common(file, FUSE_RELEASE);

	/* return value is ignored by VFS */
	return 0;
}

void fuse_sync_release(struct fuse_file *ff, int flags)
{
	WARN_ON(atomic_read(&ff->count) > 1);
	fuse_prepare_release(ff, flags, FUSE_RELEASE);
Miklos Szeredi's avatar
Miklos Szeredi committed
302 303
	__set_bit(FR_FORCE, &ff->reserved_req->flags);
	__clear_bit(FR_BACKGROUND, &ff->reserved_req->flags);
304 305 306
	fuse_request_send(ff->fc, ff->reserved_req);
	fuse_put_request(ff->fc, ff->reserved_req);
	kfree(ff);
307
}
308
EXPORT_SYMBOL_GPL(fuse_sync_release);
309

310
/*
311 312
 * Scramble the ID space with XTEA, so that the value of the files_struct
 * pointer is not exposed to userspace.
313
 */
314
u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id)
315
{
316 317 318 319 320 321 322 323 324 325 326 327 328 329
	u32 *k = fc->scramble_key;
	u64 v = (unsigned long) id;
	u32 v0 = v;
	u32 v1 = v >> 32;
	u32 sum = 0;
	int i;

	for (i = 0; i < 32; i++) {
		v0 += ((v1 << 4 ^ v1 >> 5) + v1) ^ (sum + k[sum & 3]);
		sum += 0x9E3779B9;
		v1 += ((v0 << 4 ^ v0 >> 5) + v0) ^ (sum + k[sum>>11 & 3]);
	}

	return (u64) v0 + ((u64) v1 << 32);
330 331
}

Miklos Szeredi's avatar
Miklos Szeredi committed
332
/*
333
 * Check if any page in a range is under writeback
Miklos Szeredi's avatar
Miklos Szeredi committed
334 335 336 337
 *
 * This is currently done by walking the list of writepage requests
 * for the inode, which can be pretty inefficient.
 */
338 339
static bool fuse_range_is_writeback(struct inode *inode, pgoff_t idx_from,
				   pgoff_t idx_to)
Miklos Szeredi's avatar
Miklos Szeredi committed
340 341 342 343 344 345 346 347 348 349 350
{
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
	struct fuse_req *req;
	bool found = false;

	spin_lock(&fc->lock);
	list_for_each_entry(req, &fi->writepages, writepages_entry) {
		pgoff_t curr_index;

		BUG_ON(req->inode != inode);
351
		curr_index = req->misc.write.in.offset >> PAGE_SHIFT;
352 353
		if (idx_from < curr_index + req->num_pages &&
		    curr_index <= idx_to) {
Miklos Szeredi's avatar
Miklos Szeredi committed
354 355 356 357 358 359 360 361 362
			found = true;
			break;
		}
	}
	spin_unlock(&fc->lock);

	return found;
}

363 364 365 366 367
static inline bool fuse_page_is_writeback(struct inode *inode, pgoff_t index)
{
	return fuse_range_is_writeback(inode, index, index);
}

Miklos Szeredi's avatar
Miklos Szeredi committed
368 369 370 371 372 373 374 375 376 377 378 379 380 381
/*
 * Wait for page writeback to be completed.
 *
 * Since fuse doesn't rely on the VM writeback tracking, this has to
 * use some other means.
 */
static int fuse_wait_on_page_writeback(struct inode *inode, pgoff_t index)
{
	struct fuse_inode *fi = get_fuse_inode(inode);

	wait_event(fi->page_waitq, !fuse_page_is_writeback(inode, index));
	return 0;
}

382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
/*
 * Wait for all pending writepages on the inode to finish.
 *
 * This is currently done by blocking further writes with FUSE_NOWRITE
 * and waiting for all sent writes to complete.
 *
 * This must be called under i_mutex, otherwise the FUSE_NOWRITE usage
 * could conflict with truncation.
 */
static void fuse_sync_writes(struct inode *inode)
{
	fuse_set_nowrite(inode);
	fuse_release_nowrite(inode);
}

397
static int fuse_flush(struct file *file, fl_owner_t id)
398
{
399
	struct inode *inode = file_inode(file);
400 401 402 403 404 405
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_file *ff = file->private_data;
	struct fuse_req *req;
	struct fuse_flush_in inarg;
	int err;

406 407 408
	if (is_bad_inode(inode))
		return -EIO;

409 410 411
	if (fc->no_flush)
		return 0;

Miklos Szeredi's avatar
Miklos Szeredi committed
412
	err = write_inode_now(inode, 1);
413 414 415
	if (err)
		return err;

Al Viro's avatar
Al Viro committed
416
	inode_lock(inode);
417
	fuse_sync_writes(inode);
Al Viro's avatar
Al Viro committed
418
	inode_unlock(inode);
419

420 421 422 423 424 425 426 427 428
	if (test_bit(AS_ENOSPC, &file->f_mapping->flags) &&
	    test_and_clear_bit(AS_ENOSPC, &file->f_mapping->flags))
		err = -ENOSPC;
	if (test_bit(AS_EIO, &file->f_mapping->flags) &&
	    test_and_clear_bit(AS_EIO, &file->f_mapping->flags))
		err = -EIO;
	if (err)
		return err;

429
	req = fuse_get_req_nofail_nopages(fc, file);
430 431
	memset(&inarg, 0, sizeof(inarg));
	inarg.fh = ff->fh;
432
	inarg.lock_owner = fuse_lock_owner_id(fc, id);
433 434 435 436 437
	req->in.h.opcode = FUSE_FLUSH;
	req->in.h.nodeid = get_node_id(inode);
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(inarg);
	req->in.args[0].value = &inarg;
Miklos Szeredi's avatar
Miklos Szeredi committed
438
	__set_bit(FR_FORCE, &req->flags);
439
	fuse_request_send(fc, req);
440 441 442 443 444 445 446 447 448
	err = req->out.h.error;
	fuse_put_request(fc, req);
	if (err == -ENOSYS) {
		fc->no_flush = 1;
		err = 0;
	}
	return err;
}

449 450
int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
		      int datasync, int isdir)
451
{
452
	struct inode *inode = file->f_mapping->host;
453 454
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_file *ff = file->private_data;
455
	FUSE_ARGS(args);
456 457 458
	struct fuse_fsync_in inarg;
	int err;

459 460 461
	if (is_bad_inode(inode))
		return -EIO;

Al Viro's avatar
Al Viro committed
462
	inode_lock(inode);
463

Miklos Szeredi's avatar
Miklos Szeredi committed
464 465 466 467 468
	/*
	 * Start writeback against all dirty pages of the inode, then
	 * wait for all outstanding writes, before sending the FSYNC
	 * request.
	 */
Miklos Szeredi's avatar
Miklos Szeredi committed
469
	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
Miklos Szeredi's avatar
Miklos Szeredi committed
470
	if (err)
471
		goto out;
Miklos Szeredi's avatar
Miklos Szeredi committed
472 473

	fuse_sync_writes(inode);
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488

	/*
	 * Due to implementation of fuse writeback
	 * filemap_write_and_wait_range() does not catch errors.
	 * We have to do this directly after fuse_sync_writes()
	 */
	if (test_bit(AS_ENOSPC, &file->f_mapping->flags) &&
	    test_and_clear_bit(AS_ENOSPC, &file->f_mapping->flags))
		err = -ENOSPC;
	if (test_bit(AS_EIO, &file->f_mapping->flags) &&
	    test_and_clear_bit(AS_EIO, &file->f_mapping->flags))
		err = -EIO;
	if (err)
		goto out;

Miklos Szeredi's avatar
Miklos Szeredi committed
489 490 491
	err = sync_inode_metadata(inode, 1);
	if (err)
		goto out;
Miklos Szeredi's avatar
Miklos Szeredi committed
492

Miklos Szeredi's avatar
Miklos Szeredi committed
493 494
	if ((!isdir && fc->no_fsync) || (isdir && fc->no_fsyncdir))
		goto out;
495

496 497 498
	memset(&inarg, 0, sizeof(inarg));
	inarg.fh = ff->fh;
	inarg.fsync_flags = datasync ? 1 : 0;
499 500 501 502 503 504
	args.in.h.opcode = isdir ? FUSE_FSYNCDIR : FUSE_FSYNC;
	args.in.h.nodeid = get_node_id(inode);
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	err = fuse_simple_request(fc, &args);
505
	if (err == -ENOSYS) {
506 507 508 509
		if (isdir)
			fc->no_fsyncdir = 1;
		else
			fc->no_fsync = 1;
510 511
		err = 0;
	}
512
out:
Al Viro's avatar
Al Viro committed
513
	inode_unlock(inode);
514 515 516
	return err;
}

517 518
static int fuse_fsync(struct file *file, loff_t start, loff_t end,
		      int datasync)
519
{
520
	return fuse_fsync_common(file, start, end, datasync, 0);
521 522
}

523 524
void fuse_read_fill(struct fuse_req *req, struct file *file, loff_t pos,
		    size_t count, int opcode)
525
{
526
	struct fuse_read_in *inarg = &req->misc.read.in;
527
	struct fuse_file *ff = file->private_data;
528

529 530 531
	inarg->fh = ff->fh;
	inarg->offset = pos;
	inarg->size = count;
532
	inarg->flags = file->f_flags;
533
	req->in.h.opcode = opcode;
534
	req->in.h.nodeid = ff->nodeid;
535 536
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(struct fuse_read_in);
537
	req->in.args[0].value = inarg;
538 539 540 541 542
	req->out.argvar = 1;
	req->out.numargs = 1;
	req->out.args[0].size = count;
}

543 544 545 546 547 548 549 550 551 552 553 554
static void fuse_release_user_pages(struct fuse_req *req, int write)
{
	unsigned i;

	for (i = 0; i < req->num_pages; i++) {
		struct page *page = req->pages[i];
		if (write)
			set_page_dirty_lock(page);
		put_page(page);
	}
}

555 556 557 558 559
static void fuse_io_release(struct kref *kref)
{
	kfree(container_of(kref, struct fuse_io_priv, refcnt));
}

560 561 562 563 564 565 566 567 568 569 570
static ssize_t fuse_get_res_by_io(struct fuse_io_priv *io)
{
	if (io->err)
		return io->err;

	if (io->bytes >= 0 && io->write)
		return -EIO;

	return io->bytes < 0 ? io->size : io->bytes;
}

571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588
/**
 * In case of short read, the caller sets 'pos' to the position of
 * actual end of fuse request in IO request. Otherwise, if bytes_requested
 * == bytes_transferred or rw == WRITE, the caller sets 'pos' to -1.
 *
 * An example:
 * User requested DIO read of 64K. It was splitted into two 32K fuse requests,
 * both submitted asynchronously. The first of them was ACKed by userspace as
 * fully completed (req->out.args[0].size == 32K) resulting in pos == -1. The
 * second request was ACKed as short, e.g. only 1K was read, resulting in
 * pos == 33K.
 *
 * Thus, when all fuse requests are completed, the minimal non-negative 'pos'
 * will be equal to the length of the longest contiguous fragment of
 * transferred data starting from the beginning of IO request.
 */
static void fuse_aio_complete(struct fuse_io_priv *io, int err, ssize_t pos)
{
589
	bool is_sync = is_sync_kiocb(io->iocb);
590 591 592 593 594 595 596 597 598
	int left;

	spin_lock(&io->lock);
	if (err)
		io->err = io->err ? : err;
	else if (pos >= 0 && (io->bytes < 0 || pos < io->bytes))
		io->bytes = pos;

	left = --io->reqs;
599 600
	if (!left && is_sync)
		complete(io->done);
601 602
	spin_unlock(&io->lock);

603 604
	if (!left && !is_sync) {
		ssize_t res = fuse_get_res_by_io(io);
605

606 607 608 609
		if (res >= 0) {
			struct inode *inode = file_inode(io->iocb->ki_filp);
			struct fuse_conn *fc = get_fuse_conn(inode);
			struct fuse_inode *fi = get_fuse_inode(inode);
610

611 612 613
			spin_lock(&fc->lock);
			fi->attr_version = ++fc->attr_version;
			spin_unlock(&fc->lock);
614 615
		}

616
		io->iocb->ki_complete(io->iocb, res, 0);
617
	}
618 619

	kref_put(&io->refcnt, fuse_io_release);
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
}

static void fuse_aio_complete_req(struct fuse_conn *fc, struct fuse_req *req)
{
	struct fuse_io_priv *io = req->io;
	ssize_t pos = -1;

	fuse_release_user_pages(req, !io->write);

	if (io->write) {
		if (req->misc.write.in.size != req->misc.write.out.size)
			pos = req->misc.write.in.offset - io->offset +
				req->misc.write.out.size;
	} else {
		if (req->misc.read.in.size != req->out.args[0].size)
			pos = req->misc.read.in.offset - io->offset +
				req->out.args[0].size;
	}

	fuse_aio_complete(io, req->out.h.error, pos);
}

static size_t fuse_async_req_send(struct fuse_conn *fc, struct fuse_req *req,
		size_t num_bytes, struct fuse_io_priv *io)
{
	spin_lock(&io->lock);
646
	kref_get(&io->refcnt);
647 648 649 650 651 652 653
	io->size += num_bytes;
	io->reqs++;
	spin_unlock(&io->lock);

	req->io = io;
	req->end = fuse_aio_complete_req;

654
	__fuse_get_request(req);
655 656 657 658 659
	fuse_request_send_background(fc, req);

	return num_bytes;
}

660
static size_t fuse_send_read(struct fuse_req *req, struct fuse_io_priv *io,
661
			     loff_t pos, size_t count, fl_owner_t owner)
662
{
663
	struct file *file = io->file;
664 665
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
666

667
	fuse_read_fill(req, file, pos, count, FUSE_READ);
668
	if (owner != NULL) {
669
		struct fuse_read_in *inarg = &req->misc.read.in;
670 671 672 673

		inarg->read_flags |= FUSE_READ_LOCKOWNER;
		inarg->lock_owner = fuse_lock_owner_id(fc, owner);
	}
674 675 676 677

	if (io->async)
		return fuse_async_req_send(fc, req, count, io);

678
	fuse_request_send(fc, req);
679
	return req->out.args[0].size;
680 681
}

682 683 684 685 686 687 688
static void fuse_read_update_size(struct inode *inode, loff_t size,
				  u64 attr_ver)
{
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);

	spin_lock(&fc->lock);
689 690
	if (attr_ver == fi->attr_version && size < inode->i_size &&
	    !test_bit(FUSE_I_SIZE_UNSTABLE, &fi->state)) {
691 692 693 694 695 696
		fi->attr_version = ++fc->attr_version;
		i_size_write(inode, size);
	}
	spin_unlock(&fc->lock);
}

697 698 699 700
static void fuse_short_read(struct fuse_req *req, struct inode *inode,
			    u64 attr_ver)
{
	size_t num_read = req->out.args[0].size;
701 702 703 704 705 706 707 708 709
	struct fuse_conn *fc = get_fuse_conn(inode);

	if (fc->writeback_cache) {
		/*
		 * A hole in a file. Some data after the hole are in page cache,
		 * but have not reached the client fs yet. So, the hole is not
		 * present there.
		 */
		int i;
710 711
		int start_idx = num_read >> PAGE_SHIFT;
		size_t off = num_read & (PAGE_SIZE - 1);
712

713
		for (i = start_idx; i < req->num_pages; i++) {
714
			zero_user_segment(req->pages[i], off, PAGE_SIZE);
715 716 717 718 719 720
			off = 0;
		}
	} else {
		loff_t pos = page_offset(req->pages[0]) + num_read;
		fuse_read_update_size(inode, pos, attr_ver);
	}
721 722
}

723
static int fuse_do_readpage(struct file *file, struct page *page)
724
{
725
	struct fuse_io_priv io = FUSE_IO_PRIV_SYNC(file);
726 727
	struct inode *inode = page->mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
728
	struct fuse_req *req;
729 730
	size_t num_read;
	loff_t pos = page_offset(page);
731
	size_t count = PAGE_SIZE;
732
	u64 attr_ver;
733 734
	int err;

Miklos Szeredi's avatar
Miklos Szeredi committed
735
	/*
Lucas De Marchi's avatar
Lucas De Marchi committed
736
	 * Page writeback can extend beyond the lifetime of the
Miklos Szeredi's avatar
Miklos Szeredi committed
737 738 739 740 741
	 * page-cache page, so make sure we read a properly synced
	 * page.
	 */
	fuse_wait_on_page_writeback(inode, page->index);

742
	req = fuse_get_req(fc, 1);
743
	if (IS_ERR(req))
744
		return PTR_ERR(req);
745

746 747
	attr_ver = fuse_get_attr_version(fc);

748
	req->out.page_zeroing = 1;
749
	req->out.argpages = 1;
750 751
	req->num_pages = 1;
	req->pages[0] = page;
752
	req->page_descs[0].length = count;
753
	num_read = fuse_send_read(req, &io, pos, count, NULL);
754
	err = req->out.h.error;
755 756 757 758 759 760

	if (!err) {
		/*
		 * Short read means EOF.  If file size is larger, truncate it
		 */
		if (num_read < count)
761
			fuse_short_read(req, inode, attr_ver);
762

763
		SetPageUptodate(page);
764 765
	}

766
	fuse_put_request(fc, req);
767 768 769 770 771 772 773 774 775 776 777 778 779 780

	return err;
}

static int fuse_readpage(struct file *file, struct page *page)
{
	struct inode *inode = page->mapping->host;
	int err;

	err = -EIO;
	if (is_bad_inode(inode))
		goto out;

	err = fuse_do_readpage(file, page);
781
	fuse_invalidate_atime(inode);
782 783 784 785 786
 out:
	unlock_page(page);
	return err;
}

787
static void fuse_readpages_end(struct fuse_conn *fc, struct fuse_req *req)
788
{
789
	int i;
790 791
	size_t count = req->misc.read.in.size;
	size_t num_read = req->out.args[0].size;
792
	struct address_space *mapping = NULL;
793

794 795
	for (i = 0; mapping == NULL && i < req->num_pages; i++)
		mapping = req->pages[i]->mapping;
796

797 798 799 800 801 802
	if (mapping) {
		struct inode *inode = mapping->host;

		/*
		 * Short read means EOF. If file size is larger, truncate it
		 */
803 804
		if (!req->out.h.error && num_read < count)
			fuse_short_read(req, inode, req->misc.read.attr_ver);
805

806
		fuse_invalidate_atime(inode);
807
	}
808

809 810 811 812
	for (i = 0; i < req->num_pages; i++) {
		struct page *page = req->pages[i];
		if (!req->out.h.error)
			SetPageUptodate(page);
813 814
		else
			SetPageError(page);
815
		unlock_page(page);
816
		put_page(page);
817
	}
818
	if (req->ff)
819
		fuse_file_put(req->ff, false);
820 821
}

822
static void fuse_send_readpages(struct fuse_req *req, struct file *file)
823
{
824 825
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
826
	loff_t pos = page_offset(req->pages[0]);
827
	size_t count = req->num_pages << PAGE_SHIFT;
828 829

	req->out.argpages = 1;
830
	req->out.page_zeroing = 1;
831
	req->out.page_replace = 1;
832
	fuse_read_fill(req, file, pos, count, FUSE_READ);
833
	req->misc.read.attr_ver = fuse_get_attr_version(fc);
834
	if (fc->async_read) {
835
		req->ff = fuse_file_get(ff);
836
		req->end = fuse_readpages_end;
837
		fuse_request_send_background(fc, req);
838
	} else {
839
		fuse_request_send(fc, req);
840
		fuse_readpages_end(fc, req);
841
		fuse_put_request(fc, req);
842
	}
843 844
}

845
struct fuse_fill_data {
846
	struct fuse_req *req;
847
	struct file *file;
848
	struct inode *inode;
849
	unsigned nr_pages;
850 851 852 853
};

static int fuse_readpages_fill(void *_data, struct page *page)
{
854
	struct fuse_fill_data *data = _data;
855 856 857 858
	struct fuse_req *req = data->req;
	struct inode *inode = data->inode;
	struct fuse_conn *fc = get_fuse_conn(inode);

Miklos Szeredi's avatar
Miklos Szeredi committed
859 860
	fuse_wait_on_page_writeback(inode, page->index);

861 862
	if (req->num_pages &&
	    (req->num_pages == FUSE_MAX_PAGES_PER_REQ ||
863
	     (req->num_pages + 1) * PAGE_SIZE > fc->max_read ||
864
	     req->pages[req->num_pages - 1]->index + 1 != page->index)) {
865 866
		int nr_alloc = min_t(unsigned, data->nr_pages,
				     FUSE_MAX_PAGES_PER_REQ);
867
		fuse_send_readpages(req, data->file);
868 869 870 871 872 873
		if (fc->async_read)
			req = fuse_get_req_for_background(fc, nr_alloc);
		else
			req = fuse_get_req(fc, nr_alloc);

		data->req = req;
874
		if (IS_ERR(req)) {
875
			unlock_page(page);
876
			return PTR_ERR(req);
877 878
		}
	}
879 880 881 882 883 884

	if (WARN_ON(req->num_pages >= req->max_pages)) {
		fuse_put_request(fc, req);
		return -EIO;
	}

885
	get_page(page);
886
	req->pages[req->num_pages] = page;
887
	req->page_descs[req->num_pages].length = PAGE_SIZE;
Miklos Szeredi's avatar
Miklos Szeredi committed
888
	req->num_pages++;
889
	data->nr_pages--;
890 891 892 893 894 895 896 897
	return 0;
}

static int fuse_readpages(struct file *file, struct address_space *mapping,
			  struct list_head *pages, unsigned nr_pages)
{
	struct inode *inode = mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
898
	struct fuse_fill_data data;
899
	int err;
900
	int nr_alloc = min_t(unsigned, nr_pages, FUSE_MAX_PAGES_PER_REQ);
901

902
	err = -EIO;
903
	if (is_bad_inode(inode))
904
		goto out;
905

906
	data.file = file;
907
	data.inode = inode;
908 909 910 911
	if (fc->async_read)
		data.req = fuse_get_req_for_background(fc, nr_alloc);
	else
		data.req = fuse_get_req(fc, nr_alloc);
912
	data.nr_pages = nr_pages;
913
	err = PTR_ERR(data.req);
914
	if (IS_ERR(data.req))
915
		goto out;
916 917

	err = read_cache_pages(mapping, pages, fuse_readpages_fill, &data);
918 919
	if (!err) {
		if (data.req->num_pages)
920
			fuse_send_readpages(data.req, file);
921 922 923
		else
			fuse_put_request(fc, data.req);
	}
924
out:
925
	return err;
926 927
}

928
static ssize_t fuse_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
Miklos Szeredi's avatar
Miklos Szeredi committed
929 930
{
	struct inode *inode = iocb->ki_filp->f_mapping->host;
931
	struct fuse_conn *fc = get_fuse_conn(inode);
Miklos Szeredi's avatar
Miklos Szeredi committed
932

933 934 935 936 937 938
	/*
	 * In auto invalidate mode, always update attributes on read.
	 * Otherwise, only update if we attempt to read past EOF (to ensure
	 * i_size is up to date).
	 */
	if (fc->auto_inval_data ||
939
	    (iocb->ki_pos + iov_iter_count(to) > i_size_read(inode))) {
Miklos Szeredi's avatar
Miklos Szeredi committed
940 941 942 943 944 945
		int err;
		err = fuse_update_attributes(inode, NULL, iocb->ki_filp, NULL);
		if (err)
			return err;
	}

946
	return generic_file_read_iter(iocb, to);
Miklos Szeredi's avatar
Miklos Szeredi committed
947 948
}

949
static void fuse_write_fill(struct fuse_req *req, struct fuse_file *ff,
950
			    loff_t pos, size_t count)
951
{
952 953
	struct fuse_write_in *inarg = &req->misc.write.in;
	struct fuse_write_out *outarg = &req->misc.write.out;
954

955 956 957
	inarg->fh = ff->fh;
	inarg->offset = pos;
	inarg->size = count;
958
	req->in.h.opcode = FUSE_WRITE;
959
	req->in.h.nodeid = ff->nodeid;
960
	req->in.numargs = 2;
961
	if (ff->fc->minor < 9)
962 963 964
		req->in.args[0].size = FUSE_COMPAT_WRITE_IN_SIZE;
	else
		req->in.args[0].size = sizeof(struct fuse_write_in);
965
	req->in.args[0].value = inarg;
966 967 968
	req->in.args[1].size = count;
	req->out.numargs = 1;
	req->out.args[0].size = sizeof(struct fuse_write_out);
969 970 971
	req->out.args[0].value = outarg;
}

972
static size_t fuse_send_write(struct fuse_req *req, struct fuse_io_priv *io,
973
			      loff_t pos, size_t count, fl_owner_t owner)
974
{
975
	struct file *file = io->file;
976 977
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
978 979
	struct fuse_write_in *inarg = &req->misc.write.in;

980
	fuse_write_fill(req, ff, pos, count);
981
	inarg->flags = file->f_flags;
982 983 984 985
	if (owner != NULL) {
		inarg->write_flags |= FUSE_WRITE_LOCKOWNER;
		inarg->lock_owner = fuse_lock_owner_id(fc, owner);
	}
986 987 988 989

	if (io->async)
		return fuse_async_req_send(fc, req, count, io);

990
	fuse_request_send(fc, req);
991
	return req->misc.write.out.size;
992 993
}

994
bool fuse_write_update_size(struct inode *inode, loff_t pos)
995 996 997
{
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
998
	bool ret = false;
999 1000 1001

	spin_lock(&fc->lock);
	fi->attr_version = ++fc->attr_version;
1002
	if (pos > inode->i_size) {
1003
		i_size_write(inode, pos);
1004 1005
		ret = true;
	}
1006
	spin_unlock(&fc->lock);
1007 1008

	return ret;
1009 1010
}

Nick Piggin's avatar
Nick Piggin committed
1011 1012 1013 1014 1015 1016 1017
static size_t fuse_send_write_pages(struct fuse_req *req, struct file *file,
				    struct inode *inode, loff_t pos,
				    size_t count)
{
	size_t res;
	unsigned offset;
	unsigned i;
1018
	struct fuse_io_priv io = FUSE_IO_PRIV_SYNC(file);
Nick Piggin's avatar
Nick Piggin committed
1019 1020 1021 1022

	for (i = 0; i < req->num_pages; i++)
		fuse_wait_on_page_writeback(inode, req->pages[i]->index);

1023
	res = fuse_send_write(req, &io, pos, count, NULL);
Nick Piggin's avatar
Nick Piggin committed
1024

1025
	offset = req->page_descs[0].offset;
Nick Piggin's avatar
Nick Piggin committed
1026 1027 1028 1029
	count = res;
	for (i = 0; i < req->num_pages; i++) {
		struct page *page = req->pages[i];

1030
		if (!req->out.h.error && !offset && count >= PAGE_SIZE)
Nick Piggin's avatar
Nick Piggin committed
1031 1032
			SetPageUptodate(page);

1033 1034
		if (count > PAGE_SIZE - offset)
			count -= PAGE_SIZE - offset;
Nick Piggin's avatar
Nick Piggin committed
1035 1036 1037 1038 1039
		else
			count = 0;
		offset = 0;

		unlock_page(page);
1040
		put_page(page);
Nick Piggin's avatar
Nick Piggin committed
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
	}

	return res;
}

static ssize_t fuse_fill_write_pages(struct fuse_req *req,
			       struct address_space *mapping,
			       struct iov_iter *ii, loff_t pos)
{
	struct fuse_conn *fc = get_fuse_conn(mapping->host);
1051
	unsigned offset = pos & (PAGE_SIZE - 1);
Nick Piggin's avatar
Nick Piggin committed
1052 1053 1054
	size_t count = 0;
	int err;

1055
	req->in.argpages = 1;
1056
	req->page_descs[0].offset = offset;
Nick Piggin's avatar
Nick Piggin committed
1057 1058 1059 1060

	do {
		size_t tmp;
		struct page *page;
1061 1062
		pgoff_t index = pos >> PAGE_SHIFT;
		size_t bytes = min_t(size_t, PAGE_SIZE - offset,
Nick Piggin's avatar
Nick Piggin committed
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
				     iov_iter_count(ii));

		bytes = min_t(size_t, bytes, fc->max_write - count);

 again:
		err = -EFAULT;
		if (iov_iter_fault_in_readable(ii, bytes))
			break;

		err = -ENOMEM;
1073
		page = grab_cache_page_write_begin(mapping, index, 0);
Nick Piggin's avatar
Nick Piggin committed
1074 1075 1076
		if (!page)
			break;

1077 1078 1079
		if (mapping_writably_mapped(mapping))
			flush_dcache_page(page);

Nick Piggin's avatar
Nick Piggin committed
1080 1081 1082
		tmp = iov_iter_copy_from_user_atomic(page, ii, offset, bytes);
		flush_dcache_page(page);

1083
		iov_iter_advance(ii, tmp);
Nick Piggin's avatar
Nick Piggin committed
1084 1085
		if (!tmp) {
			unlock_page(page);
1086
			put_page(page);
Nick Piggin's avatar
Nick Piggin committed
1087 1088 1089 1090 1091 1092
			bytes = min(bytes, iov_iter_single_seg_count(ii));
			goto again;
		}

		err = 0;
		req->pages[req->num_pages] = page;
1093
		req->page_descs[req->num_pages].length = tmp;
Nick Piggin's avatar
Nick Piggin committed
1094 1095 1096 1097 1098
		req->num_pages++;

		count += tmp;
		pos += tmp;
		offset += tmp;
1099
		if (offset == PAGE_SIZE)
Nick Piggin's avatar
Nick Piggin committed
1100 1101
			offset = 0;

1102 1103
		if (!fc->big_writes)
			break;
Nick Piggin's avatar
Nick Piggin committed
1104
	} while (iov_iter_count(ii) && count < fc->max_write &&
1105
		 req->num_pages < req->max_pages && offset == 0);
Nick Piggin's avatar
Nick Piggin committed
1106 1107 1108 1109

	return count > 0 ? count : err;
}

1110 1111 1112
static inline unsigned fuse_wr_pages(loff_t pos, size_t len)
{
	return min_t(unsigned,
1113 1114
		     ((pos + len - 1) >> PAGE_SHIFT) -
		     (pos >> PAGE_SHIFT) + 1,
1115 1116 1117
		     FUSE_MAX_PAGES_PER_REQ);
}

Nick Piggin's avatar
Nick Piggin committed
1118 1119 1120 1121 1122 1123
static ssize_t fuse_perform_write(struct file *file,
				  struct address_space *mapping,
				  struct iov_iter *ii, loff_t pos)
{
	struct inode *inode = mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
1124
	struct fuse_inode *fi = get_fuse_inode(inode);
Nick Piggin's avatar
Nick Piggin committed
1125 1126 1127 1128 1129 1130
	int err = 0;
	ssize_t res = 0;

	if (is_bad_inode(inode))
		return -EIO;

1131 1132 1133
	if (inode->i_size < pos + iov_iter_count(ii))
		set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);

Nick Piggin's avatar
Nick Piggin committed
1134 1135 1136
	do {
		struct fuse_req *req;
		ssize_t count;
1137
		unsigned nr_pages = fuse_wr_pages(pos, iov_iter_count(ii));
Nick Piggin's avatar
Nick Piggin committed
1138

1139
		req = fuse_get_req(fc, nr_pages);
Nick Piggin's avatar
Nick Piggin committed
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
		if (IS_ERR(req)) {
			err = PTR_ERR(req);
			break;
		}

		count = fuse_fill_write_pages(req, mapping, ii, pos);
		if (count <= 0) {
			err = count;
		} else {
			size_t num_written;

			num_written = fuse_send_write_pages(req, file, inode,
							    pos, count);
			err = req->out.h.error;
			if (!err) {
				res += num_written;
				pos += num_written;

				/* break out of the loop on short write */
				if (num_written != count)
					err = -EIO;
			}
		}
		fuse_put_request(fc, req);
	} while (!err && iov_iter_count(ii));

	if (res > 0)
		fuse_write_update_size(inode, pos);

1169
	clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
Nick Piggin's avatar
Nick Piggin committed
1170 1171 1172 1173 1174
	fuse_invalidate_attr(inode);

	return res > 0 ? res : err;
}

Al Viro's avatar
Al Viro committed
1175
static ssize_t fuse_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
Nick Piggin's avatar
Nick Piggin committed
1176 1177 1178 1179
{
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	ssize_t written = 0;
1180
	ssize_t written_buffered = 0;
Nick Piggin's avatar
Nick Piggin committed
1181 1182
	struct inode *inode = mapping->host;
	ssize_t err;
1183
	loff_t endbyte = 0;
Nick Piggin's avatar
Nick Piggin committed
1184

1185 1186 1187 1188 1189 1190
	if (get_fuse_conn(inode)->writeback_cache) {
		/* Update size (EOF optimization) and mode (SUID clearing) */
		err = fuse_update_attributes(mapping->host, NULL, file, NULL);
		if (err)
			return err;

Al Viro's avatar
Al Viro committed
1191
		return generic_file_write_iter(iocb, from);
1192 1193
	}

Al Viro's avatar
Al Viro committed
1194
	inode_lock(inode);
Nick Piggin's avatar
Nick Piggin committed
1195 1196

	/* We can write back this queue in page reclaim */
1197
	current->backing_dev_info = inode_to_bdi(inode);
Nick Piggin's avatar
Nick Piggin committed
1198

1199 1200
	err = generic_write_checks(iocb, from);
	if (err <= 0)
Nick Piggin's avatar
Nick Piggin committed
1201 1202
		goto out;

1203
	err = file_remove_privs(file);
Nick Piggin's avatar
Nick Piggin committed
1204 1205 1206
	if (err)
		goto out;

1207 1208 1209
	err = file_update_time(file);
	if (err)
		goto out;
Nick Piggin's avatar
Nick Piggin committed
1210

1211
	if (iocb->ki_flags & IOCB_DIRECT) {
1212
		loff_t pos = iocb->ki_pos;
Al Viro's avatar
Al Viro committed
1213 1214
		written = generic_file_direct_write(iocb, from, pos);
		if (written < 0 || !iov_iter_count(from))
1215 1216 1217
			goto out;

		pos += written;
Nick Piggin's avatar
Nick Piggin committed
1218

Al Viro's avatar
Al Viro committed
1219
		written_buffered = fuse_perform_write(file, mapping, from, pos);
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
		if (written_buffered < 0) {
			err = written_buffered;
			goto out;
		}
		endbyte = pos + written_buffered - 1;

		err = filemap_write_and_wait_range(file->f_mapping, pos,
						   endbyte);
		if (err)
			goto out;

		invalidate_mapping_pages(file->f_mapping,
1232 1233
					 pos >> PAGE_SHIFT,
					 endbyte >> PAGE_SHIFT);
1234 1235 1236 1237

		written += written_buffered;
		iocb->ki_pos = pos + written_buffered;
	} else {
1238
		written = fuse_perform_write(file, mapping, from, iocb->ki_pos);
1239
		if (written >= 0)
1240
			iocb->ki_pos += written;
1241
	}
Nick Piggin's avatar
Nick Piggin committed
1242 1243
out:
	current->backing_dev_info = NULL;
Al Viro's avatar
Al Viro committed
1244
	inode_unlock(inode);
Nick Piggin's avatar
Nick Piggin committed
1245 1246 1247 1248

	return written ? written : err;
}

1249 1250
static inline void fuse_page_descs_length_init(struct fuse_req *req,
		unsigned index, unsigned nr_pages)
1251 1252 1253
{
	int i;

1254
	for (i = index; i < index + nr_pages; i++)
1255 1256 1257 1258
		req->page_descs[i].length = PAGE_SIZE -
			req->page_descs[i].offset;
}

1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
static inline unsigned long fuse_get_user_addr(const struct iov_iter *ii)
{
	return (unsigned long)ii->iov->iov_base + ii->iov_offset;
}

static inline size_t fuse_get_frag_size(const struct iov_iter *ii,
					size_t max_size)
{
	return min(iov_iter_single_seg_count(ii), max_size);
}

1270
static int fuse_get_user_pages(struct fuse_req *req, struct iov_iter *ii,
1271
			       size_t *nbytesp, int write)
Miklos Szeredi's avatar
Miklos Szeredi committed
1272
{
1273
	size_t nbytes = 0;  /* # bytes already packed in req */
1274
	ssize_t ret = 0;
1275

1276
	/* Special case for kernel I/O: can copy directly into the buffer */
Al Viro's avatar
Al Viro committed
1277
	if (ii->type & ITER_KVEC) {
1278 1279 1280
		unsigned long user_addr = fuse_get_user_addr(ii);
		size_t frag_size = fuse_get_frag_size(ii, *nbytesp);

1281 1282 1283 1284 1285
		if (write)
			req->in.args[1].value = (void *) user_addr;
		else
			req->out.args[0].value = (void *) user_addr;

1286 1287
		iov_iter_advance(ii, frag_size);
		*nbytesp = frag_size;
1288 1289
		return 0;
	}
Miklos Szeredi's avatar
Miklos Szeredi committed
1290

1291
	while (nbytes < *nbytesp && req->num_pages < req->max_pages) {
1292
		unsigned npages;
Al Viro's avatar
Al Viro committed
1293
		size_t start;
1294
		ret = iov_iter_get_pages(ii, &req->pages[req->num_pages],
1295
					*nbytesp - nbytes,
1296 1297
					req->max_pages - req->num_pages,
					&start);
1298
		if (ret < 0)
1299
			break;
1300

1301 1302
		iov_iter_advance(ii, ret);
		nbytes += ret;
1303

1304 1305
		ret += start;
		npages = (ret + PAGE_SIZE - 1) / PAGE_SIZE;
1306

1307
		req->page_descs[req->num_pages].offset = start;
1308 1309 1310 1311
		fuse_page_descs_length_init(req, req->num_pages, npages);

		req->num_pages += npages;
		req->page_descs[req->num_pages - 1].length -=
1312
			(PAGE_SIZE - ret) & (PAGE_SIZE - 1);
1313
	}
1314 1315 1316 1317 1318 1319

	if (write)
		req->in.argpages = 1;
	else
		req->out.argpages = 1;

1320
	*nbytesp = nbytes;
1321

1322
	return ret < 0 ? ret : 0;
Miklos Szeredi's avatar
Miklos Szeredi committed
1323 1324
}

1325 1326
static inline int fuse_iter_npages(const struct iov_iter *ii_p)
{
Al Viro's avatar
Al Viro committed
1327
	return iov_iter_npages(ii_p, FUSE_MAX_PAGES_PER_REQ);
1328 1329
}

1330 1331
ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
		       loff_t *ppos, int flags)
Miklos Szeredi's avatar
Miklos Szeredi committed
1332
{
1333 1334
	int write = flags & FUSE_DIO_WRITE;
	int cuse = flags & FUSE_DIO_CUSE;
1335
	struct file *file = io->file;
1336
	struct inode *inode = file->f_mapping->host;
1337 1338
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
Miklos Szeredi's avatar
Miklos Szeredi committed
1339 1340
	size_t nmax = write ? fc->max_write : fc->max_read;
	loff_t pos = *ppos;
1341
	size_t count = iov_iter_count(iter);
1342 1343
	pgoff_t idx_from = pos >> PAGE_SHIFT;
	pgoff_t idx_to = (pos + count - 1) >> PAGE_SHIFT;
Miklos Szeredi's avatar
Miklos Szeredi committed
1344
	ssize_t res = 0;
1345
	struct fuse_req *req;
1346
	int err = 0;
1347

1348
	if (io->async)
1349
		req = fuse_get_req_for_background(fc, fuse_iter_npages(iter));
1350
	else
1351
		req = fuse_get_req(fc, fuse_iter_npages(iter));
1352 1353
	if (IS_ERR(req))
		return PTR_ERR(req);
Miklos Szeredi's avatar
Miklos Szeredi committed
1354

1355 1356
	if (!cuse && fuse_range_is_writeback(inode, idx_from, idx_to)) {
		if (!write)
Al Viro's avatar
Al Viro committed
1357
			inode_lock(inode);
1358 1359
		fuse_sync_writes(inode);
		if (!write)
Al Viro's avatar
Al Viro committed
1360
			inode_unlock(inode);
1361 1362
	}

Miklos Szeredi's avatar
Miklos Szeredi committed
1363 1364
	while (count) {
		size_t nres;
1365
		fl_owner_t owner = current->files;
1366
		size_t nbytes = min(count, nmax);
1367 1368
		err = fuse_get_user_pages(req, iter, &nbytes, write);
		if (err && !nbytes)
Miklos Szeredi's avatar
Miklos Szeredi committed
1369
			break;
1370

Miklos Szeredi's avatar
Miklos Szeredi committed
1371
		if (write)
1372
			nres = fuse_send_write(req, io, pos, nbytes, owner);
Miklos Szeredi's avatar
Miklos Szeredi committed
1373
		else
1374
			nres = fuse_send_read(req, io, pos, nbytes, owner);
1375

1376 1377
		if (!io->async)
			fuse_release_user_pages(req, !write);
Miklos Szeredi's avatar
Miklos Szeredi committed
1378
		if (req->out.h.error) {
1379
			err = req->out.h.error;
Miklos Szeredi's avatar
Miklos Szeredi committed
1380 1381
			break;
		} else if (nres > nbytes) {
1382 1383
			res = 0;
			err = -EIO;
Miklos Szeredi's avatar
Miklos Szeredi committed
1384 1385 1386 1387 1388 1389 1390
			break;
		}
		count -= nres;
		res += nres;
		pos += nres;
		if (nres != nbytes)
			break;
1391 1392
		if (count) {
			fuse_put_request(fc, req);
1393 1394
			if (io->async)
				req = fuse_get_req_for_background(fc,
1395
					fuse_iter_npages(iter));
1396
			else
1397
				req = fuse_get_req(fc, fuse_iter_npages(iter));
1398 1399 1400
			if (IS_ERR(req))
				break;
		}
Miklos Szeredi's avatar
Miklos Szeredi committed
1401
	}
1402 1403
	if (!IS_ERR(req))
		fuse_put_request(fc, req);
1404
	if (res > 0)
Miklos Szeredi's avatar
Miklos Szeredi committed
1405 1406
		*ppos = pos;

1407
	return res > 0 ? res : err;
Miklos Szeredi's avatar
Miklos Szeredi committed
1408
}
1409
EXPORT_SYMBOL_GPL(fuse_direct_io);
Miklos Szeredi's avatar
Miklos Szeredi committed
1410

1411
static ssize_t __fuse_direct_read(struct fuse_io_priv *io,
1412 1413
				  struct iov_iter *iter,
				  loff_t *ppos)
Miklos Szeredi's avatar
Miklos Szeredi committed
1414
{
1415
	ssize_t res;
1416
	struct file *file = io->file;
1417
	struct inode *inode = file_inode(file);
1418 1419 1420 1421

	if (is_bad_inode(inode))
		return -EIO;

1422
	res = fuse_direct_io(io, iter, ppos, 0);
1423 1424 1425 1426

	fuse_invalidate_attr(inode);

	return res;
Miklos Szeredi's avatar
Miklos Szeredi committed
1427 1428
}

1429
static ssize_t fuse_direct_read_iter(struct kiocb *iocb, struct iov_iter *to)
1430
{
1431
	struct fuse_io_priv io = FUSE_IO_PRIV_SYNC(iocb->ki_filp);
1432
	return __fuse_direct_read(&io, to, &iocb->ki_pos);
1433 1434
}

1435
static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from)
1436
{
1437
	struct file *file = iocb->ki_filp;
1438
	struct inode *inode = file_inode(file);
1439
	struct fuse_io_priv io = FUSE_IO_PRIV_SYNC(file);
1440
	ssize_t res;
1441 1442 1443 1444 1445

	if (is_bad_inode(inode))
		return -EIO;

	/* Don't allow parallel writes to the same file */
Al Viro's avatar
Al Viro committed
1446
	inode_lock(inode);
1447 1448
	res = generic_write_checks(iocb, from);
	if (res > 0)
1449 1450
		res = fuse_direct_io(&io, from, &iocb->ki_pos, FUSE_DIO_WRITE);
	fuse_invalidate_attr(inode);
1451
	if (res > 0)
1452
		fuse_write_update_size(inode, iocb->ki_pos);
Al Viro's avatar
Al Viro committed
1453
	inode_unlock(inode);
1454 1455 1456 1457

	return res;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
1458
static void fuse_writepage_free(struct fuse_conn *fc, struct fuse_req *req)
1459
{
1460 1461 1462 1463
	int i;

	for (i = 0; i < req->num_pages; i++)
		__free_page(req->pages[i]);
1464 1465 1466

	if (req->ff)
		fuse_file_put(req->ff, false);
Miklos Szeredi's avatar
Miklos Szeredi committed
1467 1468 1469 1470 1471 1472
}

static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req)
{
	struct inode *inode = req->inode;
	struct fuse_inode *fi = get_fuse_inode(inode);
1473
	struct backing_dev_info *bdi = inode_to_bdi(inode);
1474
	int i;
Miklos Szeredi's avatar
Miklos Szeredi committed
1475 1476

	list_del(&req->writepages_entry);
1477
	for (i = 0; i < req->num_pages; i++) {
1478
		dec_wb_stat(&bdi->wb, WB_WRITEBACK);
1479
		dec_zone_page_state(req->pages[i], NR_WRITEBACK_TEMP);
1480
		wb_writeout_inc(&bdi->wb);
1481
	}
Miklos Szeredi's avatar
Miklos Szeredi committed
1482 1483 1484 1485
	wake_up(&fi->page_waitq);
}

/* Called under fc->lock, may release and reacquire it */
1486 1487
static void fuse_send_writepage(struct fuse_conn *fc, struct fuse_req *req,
				loff_t size)
Miklos Szeredi's avatar
Miklos Szeredi committed
1488 1489
__releases(fc->lock)
__acquires(fc->lock)
Miklos Szeredi's avatar
Miklos Szeredi committed
1490 1491 1492
{
	struct fuse_inode *fi = get_fuse_inode(req->inode);
	struct fuse_write_in *inarg = &req->misc.write.in;
1493
	__u64 data_size = req->num_pages * PAGE_SIZE;
Miklos Szeredi's avatar
Miklos Szeredi committed
1494 1495 1496 1497

	if (!fc->connected)
		goto out_free;

1498 1499
	if (inarg->offset + data_size <= size) {
		inarg->size = data_size;
Miklos Szeredi's avatar
Miklos Szeredi committed
1500
	} else if (inarg->offset < size) {
1501
		inarg->size = size - inarg->offset;
Miklos Szeredi's avatar
Miklos Szeredi committed
1502 1503 1504
	} else {
		/* Got truncated off completely */
		goto out_free;
1505
	}
Miklos Szeredi's avatar
Miklos Szeredi committed
1506 1507 1508

	req->in.args[1].size = inarg->size;
	fi->writectr++;
1509
	fuse_request_send_background_locked(fc, req);
Miklos Szeredi's avatar
Miklos Szeredi committed
1510 1511 1512 1513 1514 1515
	return;

 out_free:
	fuse_writepage_finish(fc, req);
	spin_unlock(&fc->lock);
	fuse_writepage_free(fc, req);
1516
	fuse_put_request(fc, req);
Miklos Szeredi's avatar
Miklos Szeredi committed
1517
	spin_lock(&fc->lock);
1518 1519
}

Miklos Szeredi's avatar
Miklos Szeredi committed
1520 1521 1522 1523 1524 1525 1526
/*
 * If fi->writectr is positive (no truncate or fsync going on) send
 * all queued writepage requests.
 *
 * Called with fc->lock
 */
void fuse_flush_writepages(struct inode *inode)
Miklos Szeredi's avatar
Miklos Szeredi committed
1527 1528
__releases(fc->lock)
__acquires(fc->lock)
1529
{
Miklos Szeredi's avatar
Miklos Szeredi committed
1530 1531
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
1532
	size_t crop = i_size_read(inode);
Miklos Szeredi's avatar
Miklos Szeredi committed
1533 1534 1535 1536 1537
	struct fuse_req *req;

	while (fi->writectr >= 0 && !list_empty(&fi->queued_writes)) {
		req = list_entry(fi->queued_writes.next, struct fuse_req, list);
		list_del_init(&req->list);
1538
		fuse_send_writepage(fc, req, crop);
Miklos Szeredi's avatar
Miklos Szeredi committed
1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
	}
}

static void fuse_writepage_end(struct fuse_conn *fc, struct fuse_req *req)
{
	struct inode *inode = req->inode;
	struct fuse_inode *fi = get_fuse_inode(inode);

	mapping_set_error(inode->i_mapping, req->out.h.error);
	spin_lock(&fc->lock);
1549
	while (req->misc.write.next) {
1550 1551
		struct fuse_conn *fc = get_fuse_conn(inode);
		struct fuse_write_in *inarg = &req->misc.write.in;
1552 1553 1554
		struct fuse_req *next = req->misc.write.next;
		req->misc.write.next = next->misc.write.next;
		next->misc.write.next = NULL;
1555
		next->ff = fuse_file_get(req->ff);
1556
		list_add(&next->writepages_entry, &fi->writepages);
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581

		/*
		 * Skip fuse_flush_writepages() to make it easy to crop requests
		 * based on primary request size.
		 *
		 * 1st case (trivial): there are no concurrent activities using
		 * fuse_set/release_nowrite.  Then we're on safe side because
		 * fuse_flush_writepages() would call fuse_send_writepage()
		 * anyway.
		 *
		 * 2nd case: someone called fuse_set_nowrite and it is waiting
		 * now for completion of all in-flight requests.  This happens
		 * rarely and no more than once per page, so this should be
		 * okay.
		 *
		 * 3rd case: someone (e.g. fuse_do_setattr()) is in the middle
		 * of fuse_set_nowrite..fuse_release_nowrite section.  The fact
		 * that fuse_set_nowrite returned implies that all in-flight
		 * requests were completed along with all of their secondary
		 * requests.  Further primary requests are blocked by negative
		 * writectr.  Hence there cannot be any in-flight requests and
		 * no invocations of fuse_writepage_end() while we're in
		 * fuse_set_nowrite..fuse_release_nowrite section.
		 */
		fuse_send_writepage(fc, next, inarg->offset + inarg->size);
1582
	}
Miklos Szeredi's avatar
Miklos Szeredi committed
1583 1584 1585 1586 1587 1588
	fi->writectr--;
	fuse_writepage_finish(fc, req);
	spin_unlock(&fc->lock);
	fuse_writepage_free(fc, req);
}

Miklos Szeredi's avatar
Miklos Szeredi committed
1589 1590
static struct fuse_file *__fuse_write_file_get(struct fuse_conn *fc,
					       struct fuse_inode *fi)
1591
{
1592
	struct fuse_file *ff = NULL;
1593 1594

	spin_lock(&fc->lock);
Miklos Szeredi's avatar
Miklos Szeredi committed
1595
	if (!list_empty(&fi->write_files)) {
1596 1597 1598 1599
		ff = list_entry(fi->write_files.next, struct fuse_file,
				write_entry);
		fuse_file_get(ff);
	}
1600 1601 1602 1603 1604
	spin_unlock(&fc->lock);

	return ff;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620
static struct fuse_file *fuse_write_file_get(struct fuse_conn *fc,
					     struct fuse_inode *fi)
{
	struct fuse_file *ff = __fuse_write_file_get(fc, fi);
	WARN_ON(!ff);
	return ff;
}

int fuse_write_inode(struct inode *inode, struct writeback_control *wbc)
{
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
	struct fuse_file *ff;
	int err;

	ff = __fuse_write_file_get(fc, fi);
1621
	err = fuse_flush_times(inode, ff);
Miklos Szeredi's avatar
Miklos Szeredi committed
1622 1623 1624 1625 1626 1627
	if (ff)
		fuse_file_put(ff, 0);

	return err;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
1628 1629 1630 1631 1632 1633 1634 1635
static int fuse_writepage_locked(struct page *page)
{
	struct address_space *mapping = page->mapping;
	struct inode *inode = mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
	struct fuse_req *req;
	struct page *tmp_page;
1636
	int error = -ENOMEM;
Miklos Szeredi's avatar
Miklos Szeredi committed
1637 1638 1639

	set_page_writeback(page);

1640
	req = fuse_request_alloc_nofs(1);
Miklos Szeredi's avatar
Miklos Szeredi committed
1641 1642 1643
	if (!req)
		goto err;

Miklos Szeredi's avatar
Miklos Szeredi committed
1644 1645
	/* writeback always goes to bg_queue */
	__set_bit(FR_BACKGROUND, &req->flags);
Miklos Szeredi's avatar
Miklos Szeredi committed
1646 1647 1648 1649
	tmp_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
	if (!tmp_page)
		goto err_free;

1650
	error = -EIO;
1651
	req->ff = fuse_write_file_get(fc, fi);
1652
	if (!req->ff)
1653
		goto err_nofile;
1654

1655
	fuse_write_fill(req, req->ff, page_offset(page), 0);
Miklos Szeredi's avatar
Miklos Szeredi committed
1656 1657

	copy_highpage(tmp_page, page);
1658
	req->misc.write.in.write_flags |= FUSE_WRITE_CACHE;
1659
	req->misc.write.next = NULL;
1660
	req->in.argpages = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
1661 1662
	req->num_pages = 1;
	req->pages[0] = tmp_page;
1663
	req->page_descs[0].offset = 0;
1664
	req->page_descs[0].length = PAGE_SIZE;
Miklos Szeredi's avatar
Miklos Szeredi committed
1665 1666 1667
	req->end = fuse_writepage_end;
	req->inode = inode;

1668
	inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
Miklos Szeredi's avatar
Miklos Szeredi committed
1669 1670 1671 1672 1673 1674 1675 1676
	inc_zone_page_state(tmp_page, NR_WRITEBACK_TEMP);

	spin_lock(&fc->lock);
	list_add(&req->writepages_entry, &fi->writepages);
	list_add_tail(&req->list, &fi->queued_writes);
	fuse_flush_writepages(inode);
	spin_unlock(&fc->lock);

1677 1678
	end_page_writeback(page);

Miklos Szeredi's avatar
Miklos Szeredi committed
1679 1680
	return 0;

1681 1682
err_nofile:
	__free_page(tmp_page);
Miklos Szeredi's avatar
Miklos Szeredi committed
1683 1684 1685 1686
err_free:
	fuse_request_free(req);
err:
	end_page_writeback(page);
1687
	return error;
Miklos Szeredi's avatar
Miklos Szeredi committed
1688 1689 1690 1691 1692 1693
}

static int fuse_writepage(struct page *page, struct writeback_control *wbc)
{
	int err;

1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
	if (fuse_page_is_writeback(page->mapping->host, page->index)) {
		/*
		 * ->writepages() should be called for sync() and friends.  We
		 * should only get here on direct reclaim and then we are
		 * allowed to skip a page which is already in flight
		 */
		WARN_ON(wbc->sync_mode == WB_SYNC_ALL);

		redirty_page_for_writepage(wbc, page);
		return 0;
	}

Miklos Szeredi's avatar
Miklos Szeredi committed
1706 1707 1708 1709 1710 1711
	err = fuse_writepage_locked(page);
	unlock_page(page);

	return err;
}

1712 1713 1714 1715
struct fuse_fill_wb_data {
	struct fuse_req *req;
	struct fuse_file *ff;
	struct inode *inode;
1716
	struct page **orig_pages;
1717 1718 1719 1720 1721 1722 1723 1724
};

static void fuse_writepages_send(struct fuse_fill_wb_data *data)
{
	struct fuse_req *req = data->req;
	struct inode *inode = data->inode;
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
1725 1726
	int num_pages = req->num_pages;
	int i;
1727 1728 1729 1730 1731 1732

	req->ff = fuse_file_get(data->ff);
	spin_lock(&fc->lock);
	list_add_tail(&req->list, &fi->queued_writes);
	fuse_flush_writepages(inode);
	spin_unlock(&fc->lock);
1733 1734 1735

	for (i = 0; i < num_pages; i++)
		end_page_writeback(data->orig_pages[i]);
1736 1737
}

1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753
static bool fuse_writepage_in_flight(struct fuse_req *new_req,
				     struct page *page)
{
	struct fuse_conn *fc = get_fuse_conn(new_req->inode);
	struct fuse_inode *fi = get_fuse_inode(new_req->inode);
	struct fuse_req *tmp;
	struct fuse_req *old_req;
	bool found = false;
	pgoff_t curr_index;

	BUG_ON(new_req->num_pages != 0);

	spin_lock(&fc->lock);
	list_del(&new_req->writepages_entry);
	list_for_each_entry(old_req, &fi->writepages, writepages_entry) {
		BUG_ON(old_req->inode != new_req->inode);
1754
		curr_index = old_req->misc.write.in.offset >> PAGE_SHIFT;
1755 1756 1757 1758 1759 1760
		if (curr_index <= page->index &&
		    page->index < curr_index + old_req->num_pages) {
			found = true;
			break;
		}
	}
1761 1762
	if (!found) {
		list_add(&new_req->writepages_entry, &fi->writepages);
1763
		goto out_unlock;
1764
	}
1765

1766
	new_req->num_pages = 1;
1767 1768
	for (tmp = old_req; tmp != NULL; tmp = tmp->misc.write.next) {
		BUG_ON(tmp->inode != new_req->inode);
1769
		curr_index = tmp->misc.write.in.offset >> PAGE_SHIFT;
1770 1771 1772 1773 1774 1775
		if (tmp->num_pages == 1 &&
		    curr_index == page->index) {
			old_req = tmp;
		}
	}

Miklos Szeredi's avatar
Miklos Szeredi committed
1776
	if (old_req->num_pages == 1 && test_bit(FR_PENDING, &old_req->flags)) {
1777
		struct backing_dev_info *bdi = inode_to_bdi(page->mapping->host);
1778

1779 1780 1781
		copy_highpage(old_req->pages[0], page);
		spin_unlock(&fc->lock);

1782
		dec_wb_stat(&bdi->wb, WB_WRITEBACK);
1783
		dec_zone_page_state(page, NR_WRITEBACK_TEMP);
1784
		wb_writeout_inc(&bdi->wb);
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797
		fuse_writepage_free(fc, new_req);
		fuse_request_free(new_req);
		goto out;
	} else {
		new_req->misc.write.next = old_req->misc.write.next;
		old_req->misc.write.next = new_req;
	}
out_unlock:
	spin_unlock(&fc->lock);
out:
	return found;
}

1798 1799 1800 1801 1802 1803 1804 1805
static int fuse_writepages_fill(struct page *page,
		struct writeback_control *wbc, void *_data)
{
	struct fuse_fill_wb_data *data = _data;
	struct fuse_req *req = data->req;
	struct inode *inode = data->inode;
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct page *tmp_page;
1806
	bool is_writeback;
1807 1808 1809 1810 1811 1812 1813 1814 1815
	int err;

	if (!data->ff) {
		err = -EIO;
		data->ff = fuse_write_file_get(fc, get_fuse_inode(inode));
		if (!data->ff)
			goto out_unlock;
	}

1816 1817 1818 1819 1820 1821 1822 1823 1824 1825
	/*
	 * Being under writeback is unlikely but possible.  For example direct
	 * read to an mmaped fuse file will set the page dirty twice; once when
	 * the pages are faulted with get_user_pages(), and then after the read
	 * completed.
	 */
	is_writeback = fuse_page_is_writeback(inode, page->index);

	if (req && req->num_pages &&
	    (is_writeback || req->num_pages == FUSE_MAX_PAGES_PER_REQ ||
1826
	     (req->num_pages + 1) * PAGE_SIZE > fc->max_write ||
1827 1828 1829
	     data->orig_pages[req->num_pages - 1]->index + 1 != page->index)) {
		fuse_writepages_send(data);
		data->req = NULL;
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
	}
	err = -ENOMEM;
	tmp_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
	if (!tmp_page)
		goto out_unlock;

	/*
	 * The page must not be redirtied until the writeout is completed
	 * (i.e. userspace has sent a reply to the write request).  Otherwise
	 * there could be more than one temporary page instance for each real
	 * page.
	 *
	 * This is ensured by holding the page lock in page_mkwrite() while
	 * checking fuse_page_is_writeback().  We already hold the page lock
	 * since clear_page_dirty_for_io() and keep it held until we add the
	 * request to the fi->writepages list and increment req->num_pages.
	 * After this fuse_page_is_writeback() will indicate that the page is
	 * under writeback, so we can release the page lock.
	 */
	if (data->req == NULL) {
		struct fuse_inode *fi = get_fuse_inode(inode);

		err = -ENOMEM;
		req = fuse_request_alloc_nofs(FUSE_MAX_PAGES_PER_REQ);
		if (!req) {
			__free_page(tmp_page);
			goto out_unlock;
		}

		fuse_write_fill(req, data->ff, page_offset(page), 0);
		req->misc.write.in.write_flags |= FUSE_WRITE_CACHE;
1861
		req->misc.write.next = NULL;
1862
		req->in.argpages = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
1863
		__set_bit(FR_BACKGROUND, &req->flags);
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880
		req->num_pages = 0;
		req->end = fuse_writepage_end;
		req->inode = inode;

		spin_lock(&fc->lock);
		list_add(&req->writepages_entry, &fi->writepages);
		spin_unlock(&fc->lock);

		data->req = req;
	}
	set_page_writeback(page);

	copy_highpage(tmp_page, page);
	req->pages[req->num_pages] = tmp_page;
	req->page_descs[req->num_pages].offset = 0;
	req->page_descs[req->num_pages].length = PAGE_SIZE;

1881
	inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
1882
	inc_zone_page_state(tmp_page, NR_WRITEBACK_TEMP);
1883 1884 1885 1886 1887 1888 1889

	err = 0;
	if (is_writeback && fuse_writepage_in_flight(req, page)) {
		end_page_writeback(page);
		data->req = NULL;
		goto out_unlock;
	}
1890
	data->orig_pages[req->num_pages] = page;
1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920

	/*
	 * Protected by fc->lock against concurrent access by
	 * fuse_page_is_writeback().
	 */
	spin_lock(&fc->lock);
	req->num_pages++;
	spin_unlock(&fc->lock);

out_unlock:
	unlock_page(page);

	return err;
}

static int fuse_writepages(struct address_space *mapping,
			   struct writeback_control *wbc)
{
	struct inode *inode = mapping->host;
	struct fuse_fill_wb_data data;
	int err;

	err = -EIO;
	if (is_bad_inode(inode))
		goto out;

	data.inode = inode;
	data.req = NULL;
	data.ff = NULL;

1921
	err = -ENOMEM;
1922 1923
	data.orig_pages = kcalloc(FUSE_MAX_PAGES_PER_REQ,
				  sizeof(struct page *),
1924 1925 1926 1927
				  GFP_NOFS);
	if (!data.orig_pages)
		goto out;

1928 1929 1930 1931 1932 1933 1934 1935 1936
	err = write_cache_pages(mapping, wbc, fuse_writepages_fill, &data);
	if (data.req) {
		/* Ignore errors if we can write at least one page */
		BUG_ON(!data.req->num_pages);
		fuse_writepages_send(&data);
		err = 0;
	}
	if (data.ff)
		fuse_file_put(data.ff, false);
1937 1938

	kfree(data.orig_pages);
1939 1940 1941 1942
out:
	return err;
}

1943 1944 1945 1946 1947 1948 1949 1950
/*
 * It's worthy to make sure that space is reserved on disk for the write,
 * but how to implement it without killing performance need more thinking.
 */
static int fuse_write_begin(struct file *file, struct address_space *mapping,
		loff_t pos, unsigned len, unsigned flags,
		struct page **pagep, void **fsdata)
{
1951
	pgoff_t index = pos >> PAGE_SHIFT;
Al Viro's avatar
Al Viro committed
1952
	struct fuse_conn *fc = get_fuse_conn(file_inode(file));
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964
	struct page *page;
	loff_t fsize;
	int err = -ENOMEM;

	WARN_ON(!fc->writeback_cache);

	page = grab_cache_page_write_begin(mapping, index, flags);
	if (!page)
		goto error;

	fuse_wait_on_page_writeback(mapping->host, page->index);

1965
	if (PageUptodate(page) || len == PAGE_SIZE)
1966 1967 1968 1969 1970 1971
		goto success;
	/*
	 * Check if the start this page comes after the end of file, in which
	 * case the readpage can be optimized away.
	 */
	fsize = i_size_read(mapping->host);
1972 1973
	if (fsize <= (pos & PAGE_MASK)) {
		size_t off = pos & ~PAGE_MASK;
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
		if (off)
			zero_user_segment(page, 0, off);
		goto success;
	}
	err = fuse_do_readpage(file, page);
	if (err)
		goto cleanup;
success:
	*pagep = page;
	return 0;

cleanup:
	unlock_page(page);
1987
	put_page(page);
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
error:
	return err;
}

static int fuse_write_end(struct file *file, struct address_space *mapping,
		loff_t pos, unsigned len, unsigned copied,
		struct page *page, void *fsdata)
{
	struct inode *inode = page->mapping->host;

	if (!PageUptodate(page)) {
		/* Zero any unwritten bytes at the end of the page */
2000
		size_t endoff = (pos + copied) & ~PAGE_MASK;
2001
		if (endoff)
2002
			zero_user_segment(page, endoff, PAGE_SIZE);
2003 2004 2005 2006 2007 2008
		SetPageUptodate(page);
	}

	fuse_write_update_size(inode, pos + copied);
	set_page_dirty(page);
	unlock_page(page);
2009
	put_page(page);
2010 2011 2012 2013

	return copied;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049
static int fuse_launder_page(struct page *page)
{
	int err = 0;
	if (clear_page_dirty_for_io(page)) {
		struct inode *inode = page->mapping->host;
		err = fuse_writepage_locked(page);
		if (!err)
			fuse_wait_on_page_writeback(inode, page->index);
	}
	return err;
}

/*
 * Write back dirty pages now, because there may not be any suitable
 * open files later
 */
static void fuse_vma_close(struct vm_area_struct *vma)
{
	filemap_write_and_wait(vma->vm_file->f_mapping);
}

/*
 * Wait for writeback against this page to complete before allowing it
 * to be marked dirty again, and hence written back again, possibly
 * before the previous writepage completed.
 *
 * Block here, instead of in ->writepage(), so that the userspace fs
 * can only block processes actually operating on the filesystem.
 *
 * Otherwise unprivileged userspace fs would be able to block
 * unrelated:
 *
 * - page migration
 * - sync(2)
 * - try_to_free_pages() with order > PAGE_ALLOC_COSTLY_ORDER
 */
2050
static int fuse_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
Miklos Szeredi's avatar
Miklos Szeredi committed
2051
{
2052
	struct page *page = vmf->page;
Miklos Szeredi's avatar
Miklos Szeredi committed
2053 2054 2055 2056 2057 2058 2059 2060
	struct inode *inode = file_inode(vma->vm_file);

	file_update_time(vma->vm_file);
	lock_page(page);
	if (page->mapping != inode->i_mapping) {
		unlock_page(page);
		return VM_FAULT_NOPAGE;
	}
Miklos Szeredi's avatar
Miklos Szeredi committed
2061 2062

	fuse_wait_on_page_writeback(inode, page->index);
Miklos Szeredi's avatar
Miklos Szeredi committed
2063
	return VM_FAULT_LOCKED;
Miklos Szeredi's avatar
Miklos Szeredi committed
2064 2065
}

2066
static const struct vm_operations_struct fuse_file_vm_ops = {
Miklos Szeredi's avatar
Miklos Szeredi committed
2067 2068
	.close		= fuse_vma_close,
	.fault		= filemap_fault,
2069
	.map_pages	= filemap_map_pages,
Miklos Szeredi's avatar
Miklos Szeredi committed
2070 2071 2072 2073 2074
	.page_mkwrite	= fuse_page_mkwrite,
};

static int fuse_file_mmap(struct file *file, struct vm_area_struct *vma)
{
2075 2076 2077
	if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE))
		fuse_link_write_file(file);

Miklos Szeredi's avatar
Miklos Szeredi committed
2078 2079
	file_accessed(file);
	vma->vm_ops = &fuse_file_vm_ops;
2080 2081 2082
	return 0;
}

2083 2084 2085 2086 2087 2088
static int fuse_direct_mmap(struct file *file, struct vm_area_struct *vma)
{
	/* Can't provide the coherency needed for MAP_SHARED */
	if (vma->vm_flags & VM_MAYSHARE)
		return -ENODEV;

2089 2090
	invalidate_inode_pages2(file->f_mapping);

2091 2092 2093
	return generic_file_mmap(file, vma);
}

2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118
static int convert_fuse_file_lock(const struct fuse_file_lock *ffl,
				  struct file_lock *fl)
{
	switch (ffl->type) {
	case F_UNLCK:
		break;

	case F_RDLCK:
	case F_WRLCK:
		if (ffl->start > OFFSET_MAX || ffl->end > OFFSET_MAX ||
		    ffl->end < ffl->start)
			return -EIO;

		fl->fl_start = ffl->start;
		fl->fl_end = ffl->end;
		fl->fl_pid = ffl->pid;
		break;

	default:
		return -EIO;
	}
	fl->fl_type = ffl->type;
	return 0;
}

2119
static void fuse_lk_fill(struct fuse_args *args, struct file *file,
2120
			 const struct file_lock *fl, int opcode, pid_t pid,
2121
			 int flock, struct fuse_lk_in *inarg)
2122
{
2123
	struct inode *inode = file_inode(file);
2124
	struct fuse_conn *fc = get_fuse_conn(inode);
2125
	struct fuse_file *ff = file->private_data;
2126 2127 2128 2129 2130 2131 2132 2133

	memset(inarg, 0, sizeof(*inarg));
	inarg->fh = ff->fh;
	inarg->owner = fuse_lock_owner_id(fc, fl->fl_owner);
	inarg->lk.start = fl->fl_start;
	inarg->lk.end = fl->fl_end;
	inarg->lk.type = fl->fl_type;
	inarg->lk.pid = pid;
2134
	if (flock)
2135 2136 2137 2138 2139 2140
		inarg->lk_flags |= FUSE_LK_FLOCK;
	args->in.h.opcode = opcode;
	args->in.h.nodeid = get_node_id(inode);
	args->in.numargs = 1;
	args->in.args[0].size = sizeof(*inarg);
	args->in.args[0].value = inarg;
2141 2142 2143 2144
}

static int fuse_getlk(struct file *file, struct file_lock *fl)
{
2145
	struct inode *inode = file_inode(file);
2146
	struct fuse_conn *fc = get_fuse_conn(inode);
2147 2148
	FUSE_ARGS(args);
	struct fuse_lk_in inarg;
2149 2150 2151
	struct fuse_lk_out outarg;
	int err;

2152 2153 2154 2155 2156
	fuse_lk_fill(&args, file, fl, FUSE_GETLK, 0, 0, &inarg);
	args.out.numargs = 1;
	args.out.args[0].size = sizeof(outarg);
	args.out.args[0].value = &outarg;
	err = fuse_simple_request(fc, &args);
2157 2158 2159 2160 2161 2162
	if (!err)
		err = convert_fuse_file_lock(&outarg.lk, fl);

	return err;
}

2163
static int fuse_setlk(struct file *file, struct file_lock *fl, int flock)
2164
{
2165
	struct inode *inode = file_inode(file);
2166
	struct fuse_conn *fc = get_fuse_conn(inode);
2167 2168
	FUSE_ARGS(args);
	struct fuse_lk_in inarg;
2169 2170 2171 2172
	int opcode = (fl->fl_flags & FL_SLEEP) ? FUSE_SETLKW : FUSE_SETLK;
	pid_t pid = fl->fl_type != F_UNLCK ? current->tgid : 0;
	int err;

2173
	if (fl->fl_lmops && fl->fl_lmops->lm_grant) {
Miklos Szeredi's avatar
Miklos Szeredi committed
2174 2175 2176 2177
		/* NLM needs asynchronous locks, which we don't support yet */
		return -ENOLCK;
	}

2178 2179 2180 2181
	/* Unlock on close is handled by the flush method */
	if (fl->fl_flags & FL_CLOSE)
		return 0;

2182 2183
	fuse_lk_fill(&args, file, fl, opcode, pid, flock, &inarg);
	err = fuse_simple_request(fc, &args);
2184

2185 2186 2187
	/* locking is restartable */
	if (err == -EINTR)
		err = -ERESTARTSYS;
2188

2189 2190 2191 2192 2193
	return err;
}

static int fuse_file_lock(struct file *file, int cmd, struct file_lock *fl)
{
2194
	struct inode *inode = file_inode(file);
2195 2196 2197
	struct fuse_conn *fc = get_fuse_conn(inode);
	int err;

Miklos Szeredi's avatar
Miklos Szeredi committed
2198 2199 2200
	if (cmd == F_CANCELLK) {
		err = 0;
	} else if (cmd == F_GETLK) {
2201
		if (fc->no_lock) {
2202
			posix_test_lock(file, fl);
2203 2204 2205 2206 2207
			err = 0;
		} else
			err = fuse_getlk(file, fl);
	} else {
		if (fc->no_lock)
Miklos Szeredi's avatar
Miklos Szeredi committed
2208
			err = posix_lock_file(file, fl, NULL);
2209
		else
2210
			err = fuse_setlk(file, fl, 0);
2211 2212 2213 2214
	}
	return err;
}

2215 2216
static int fuse_file_flock(struct file *file, int cmd, struct file_lock *fl)
{
2217
	struct inode *inode = file_inode(file);
2218 2219 2220
	struct fuse_conn *fc = get_fuse_conn(inode);
	int err;

Miklos Szeredi's avatar
Miklos Szeredi committed
2221
	if (fc->no_flock) {
2222
		err = locks_lock_file_wait(file, fl);
2223
	} else {
Miklos Szeredi's avatar
Miklos Szeredi committed
2224 2225
		struct fuse_file *ff = file->private_data;

2226
		/* emulate flock with POSIX locks */
Miklos Szeredi's avatar
Miklos Szeredi committed
2227
		ff->flock = true;
2228 2229 2230 2231 2232 2233
		err = fuse_setlk(file, fl, 1);
	}

	return err;
}

2234 2235 2236 2237
static sector_t fuse_bmap(struct address_space *mapping, sector_t block)
{
	struct inode *inode = mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
2238
	FUSE_ARGS(args);
2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
	struct fuse_bmap_in inarg;
	struct fuse_bmap_out outarg;
	int err;

	if (!inode->i_sb->s_bdev || fc->no_bmap)
		return 0;

	memset(&inarg, 0, sizeof(inarg));
	inarg.block = block;
	inarg.blocksize = inode->i_sb->s_blocksize;
2249 2250 2251 2252 2253 2254 2255 2256 2257
	args.in.h.opcode = FUSE_BMAP;
	args.in.h.nodeid = get_node_id(inode);
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.out.numargs = 1;
	args.out.args[0].size = sizeof(outarg);
	args.out.args[0].value = &outarg;
	err = fuse_simple_request(fc, &args);
2258 2259 2260 2261 2262 2263
	if (err == -ENOSYS)
		fc->no_bmap = 1;

	return err ? 0 : outarg.block;
}

2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307
static loff_t fuse_lseek(struct file *file, loff_t offset, int whence)
{
	struct inode *inode = file->f_mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_file *ff = file->private_data;
	FUSE_ARGS(args);
	struct fuse_lseek_in inarg = {
		.fh = ff->fh,
		.offset = offset,
		.whence = whence
	};
	struct fuse_lseek_out outarg;
	int err;

	if (fc->no_lseek)
		goto fallback;

	args.in.h.opcode = FUSE_LSEEK;
	args.in.h.nodeid = ff->nodeid;
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.out.numargs = 1;
	args.out.args[0].size = sizeof(outarg);
	args.out.args[0].value = &outarg;
	err = fuse_simple_request(fc, &args);
	if (err) {
		if (err == -ENOSYS) {
			fc->no_lseek = 1;
			goto fallback;
		}
		return err;
	}

	return vfs_setpos(file, outarg.offset, inode->i_sb->s_maxbytes);

fallback:
	err = fuse_update_attributes(inode, NULL, file, NULL);
	if (!err)
		return generic_file_llseek(file, offset, whence);
	else
		return err;
}

2308
static loff_t fuse_file_llseek(struct file *file, loff_t offset, int whence)
Miklos Szeredi's avatar
Miklos Szeredi committed
2309 2310
{
	loff_t retval;
2311
	struct inode *inode = file_inode(file);
Miklos Szeredi's avatar
Miklos Szeredi committed
2312

2313 2314 2315 2316
	switch (whence) {
	case SEEK_SET:
	case SEEK_CUR:
		 /* No i_mutex protection necessary for SEEK_CUR and SEEK_SET */
2317
		retval = generic_file_llseek(file, offset, whence);
2318 2319
		break;
	case SEEK_END:
Al Viro's avatar
Al Viro committed
2320
		inode_lock(inode);
2321 2322 2323
		retval = fuse_update_attributes(inode, NULL, file, NULL);
		if (!retval)
			retval = generic_file_llseek(file, offset, whence);
Al Viro's avatar
Al Viro committed
2324
		inode_unlock(inode);
2325 2326 2327
		break;
	case SEEK_HOLE:
	case SEEK_DATA:
Al Viro's avatar
Al Viro committed
2328
		inode_lock(inode);
2329
		retval = fuse_lseek(file, offset, whence);
Al Viro's avatar
Al Viro committed
2330
		inode_unlock(inode);
2331 2332 2333 2334
		break;
	default:
		retval = -EINVAL;
	}
2335

Miklos Szeredi's avatar
Miklos Szeredi committed
2336 2337 2338
	return retval;
}

Tejun Heo's avatar
Tejun Heo committed
2339 2340 2341 2342 2343 2344 2345 2346 2347
static int fuse_ioctl_copy_user(struct page **pages, struct iovec *iov,
			unsigned int nr_segs, size_t bytes, bool to_user)
{
	struct iov_iter ii;
	int page_idx = 0;

	if (!bytes)
		return 0;

Al Viro's avatar
Al Viro committed
2348
	iov_iter_init(&ii, to_user ? READ : WRITE, iov, nr_segs, bytes);
Tejun Heo's avatar
Tejun Heo committed
2349 2350 2351 2352

	while (iov_iter_count(&ii)) {
		struct page *page = pages[page_idx++];
		size_t todo = min_t(size_t, PAGE_SIZE, iov_iter_count(&ii));
2353
		void *kaddr;
Tejun Heo's avatar
Tejun Heo committed
2354

2355
		kaddr = kmap(page);
Tejun Heo's avatar
Tejun Heo committed
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375

		while (todo) {
			char __user *uaddr = ii.iov->iov_base + ii.iov_offset;
			size_t iov_len = ii.iov->iov_len - ii.iov_offset;
			size_t copy = min(todo, iov_len);
			size_t left;

			if (!to_user)
				left = copy_from_user(kaddr, uaddr, copy);
			else
				left = copy_to_user(uaddr, kaddr, copy);

			if (unlikely(left))
				return -EFAULT;

			iov_iter_advance(&ii, copy);
			todo -= copy;
			kaddr += copy;
		}

2376
		kunmap(page);
Tejun Heo's avatar
Tejun Heo committed
2377 2378 2379 2380 2381
	}

	return 0;
}

2382 2383 2384 2385 2386 2387
/*
 * CUSE servers compiled on 32bit broke on 64bit kernels because the
 * ABI was defined to be 'struct iovec' which is different on 32bit
 * and 64bit.  Fortunately we can determine which structure the server
 * used from the size of the reply.
 */
Miklos Szeredi's avatar
Miklos Szeredi committed
2388 2389 2390
static int fuse_copy_ioctl_iovec_old(struct iovec *dst, void *src,
				     size_t transferred, unsigned count,
				     bool is_compat)
2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419
{
#ifdef CONFIG_COMPAT
	if (count * sizeof(struct compat_iovec) == transferred) {
		struct compat_iovec *ciov = src;
		unsigned i;

		/*
		 * With this interface a 32bit server cannot support
		 * non-compat (i.e. ones coming from 64bit apps) ioctl
		 * requests
		 */
		if (!is_compat)
			return -EINVAL;

		for (i = 0; i < count; i++) {
			dst[i].iov_base = compat_ptr(ciov[i].iov_base);
			dst[i].iov_len = ciov[i].iov_len;
		}
		return 0;
	}
#endif

	if (count * sizeof(struct iovec) != transferred)
		return -EIO;

	memcpy(dst, src, transferred);
	return 0;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
2420 2421 2422 2423 2424 2425
/* Make sure iov_length() won't overflow */
static int fuse_verify_ioctl_iov(struct iovec *iov, size_t count)
{
	size_t n;
	u32 max = FUSE_MAX_PAGES_PER_REQ << PAGE_SHIFT;

2426
	for (n = 0; n < count; n++, iov++) {
Miklos Szeredi's avatar
Miklos Szeredi committed
2427 2428 2429 2430 2431 2432 2433
		if (iov->iov_len > (size_t) max)
			return -ENOMEM;
		max -= iov->iov_len;
	}
	return 0;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469
static int fuse_copy_ioctl_iovec(struct fuse_conn *fc, struct iovec *dst,
				 void *src, size_t transferred, unsigned count,
				 bool is_compat)
{
	unsigned i;
	struct fuse_ioctl_iovec *fiov = src;

	if (fc->minor < 16) {
		return fuse_copy_ioctl_iovec_old(dst, src, transferred,
						 count, is_compat);
	}

	if (count * sizeof(struct fuse_ioctl_iovec) != transferred)
		return -EIO;

	for (i = 0; i < count; i++) {
		/* Did the server supply an inappropriate value? */
		if (fiov[i].base != (unsigned long) fiov[i].base ||
		    fiov[i].len != (unsigned long) fiov[i].len)
			return -EIO;

		dst[i].iov_base = (void __user *) (unsigned long) fiov[i].base;
		dst[i].iov_len = (size_t) fiov[i].len;

#ifdef CONFIG_COMPAT
		if (is_compat &&
		    (ptr_to_compat(dst[i].iov_base) != fiov[i].base ||
		     (compat_size_t) dst[i].iov_len != fiov[i].len))
			return -EIO;
#endif
	}

	return 0;
}


Tejun Heo's avatar
Tejun Heo committed
2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515
/*
 * For ioctls, there is no generic way to determine how much memory
 * needs to be read and/or written.  Furthermore, ioctls are allowed
 * to dereference the passed pointer, so the parameter requires deep
 * copying but FUSE has no idea whatsoever about what to copy in or
 * out.
 *
 * This is solved by allowing FUSE server to retry ioctl with
 * necessary in/out iovecs.  Let's assume the ioctl implementation
 * needs to read in the following structure.
 *
 * struct a {
 *	char	*buf;
 *	size_t	buflen;
 * }
 *
 * On the first callout to FUSE server, inarg->in_size and
 * inarg->out_size will be NULL; then, the server completes the ioctl
 * with FUSE_IOCTL_RETRY set in out->flags, out->in_iovs set to 1 and
 * the actual iov array to
 *
 * { { .iov_base = inarg.arg,	.iov_len = sizeof(struct a) } }
 *
 * which tells FUSE to copy in the requested area and retry the ioctl.
 * On the second round, the server has access to the structure and
 * from that it can tell what to look for next, so on the invocation,
 * it sets FUSE_IOCTL_RETRY, out->in_iovs to 2 and iov array to
 *
 * { { .iov_base = inarg.arg,	.iov_len = sizeof(struct a)	},
 *   { .iov_base = a.buf,	.iov_len = a.buflen		} }
 *
 * FUSE will copy both struct a and the pointed buffer from the
 * process doing the ioctl and retry ioctl with both struct a and the
 * buffer.
 *
 * This time, FUSE server has everything it needs and completes ioctl
 * without FUSE_IOCTL_RETRY which finishes the ioctl call.
 *
 * Copying data out works the same way.
 *
 * Note that if FUSE_IOCTL_UNRESTRICTED is clear, the kernel
 * automatically initializes in and out iovs by decoding @cmd with
 * _IOC_* macros and the server is not allowed to request RETRY.  This
 * limits ioctl data transfers to well-formed ioctls and is the forced
 * behavior for all FUSE servers.
 */
2516 2517
long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
		   unsigned int flags)
Tejun Heo's avatar
Tejun Heo committed
2518 2519
{
	struct fuse_file *ff = file->private_data;
2520
	struct fuse_conn *fc = ff->fc;
Tejun Heo's avatar
Tejun Heo committed
2521 2522 2523 2524 2525 2526 2527 2528 2529
	struct fuse_ioctl_in inarg = {
		.fh = ff->fh,
		.cmd = cmd,
		.arg = arg,
		.flags = flags
	};
	struct fuse_ioctl_out outarg;
	struct fuse_req *req = NULL;
	struct page **pages = NULL;
Miklos Szeredi's avatar
Miklos Szeredi committed
2530
	struct iovec *iov_page = NULL;
Tejun Heo's avatar
Tejun Heo committed
2531 2532 2533 2534 2535
	struct iovec *in_iov = NULL, *out_iov = NULL;
	unsigned int in_iovs = 0, out_iovs = 0, num_pages = 0, max_pages;
	size_t in_size, out_size, transferred;
	int err;

Miklos Szeredi's avatar
Miklos Szeredi committed
2536 2537 2538 2539 2540 2541 2542
#if BITS_PER_LONG == 32
	inarg.flags |= FUSE_IOCTL_32BIT;
#else
	if (flags & FUSE_IOCTL_COMPAT)
		inarg.flags |= FUSE_IOCTL_32BIT;
#endif

Tejun Heo's avatar
Tejun Heo committed
2543
	/* assume all the iovs returned by client always fits in a page */
Miklos Szeredi's avatar
Miklos Szeredi committed
2544
	BUILD_BUG_ON(sizeof(struct fuse_ioctl_iovec) * FUSE_IOCTL_MAX_IOV > PAGE_SIZE);
Tejun Heo's avatar
Tejun Heo committed
2545 2546

	err = -ENOMEM;
2547
	pages = kcalloc(FUSE_MAX_PAGES_PER_REQ, sizeof(pages[0]), GFP_KERNEL);
Miklos Szeredi's avatar
Miklos Szeredi committed
2548
	iov_page = (struct iovec *) __get_free_page(GFP_KERNEL);
Tejun Heo's avatar
Tejun Heo committed
2549 2550 2551 2552 2553 2554 2555 2556
	if (!pages || !iov_page)
		goto out;

	/*
	 * If restricted, initialize IO parameters as encoded in @cmd.
	 * RETRY from server is not allowed.
	 */
	if (!(flags & FUSE_IOCTL_UNRESTRICTED)) {
Miklos Szeredi's avatar
Miklos Szeredi committed
2557
		struct iovec *iov = iov_page;
Tejun Heo's avatar
Tejun Heo committed
2558

2559
		iov->iov_base = (void __user *)arg;
Tejun Heo's avatar
Tejun Heo committed
2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594
		iov->iov_len = _IOC_SIZE(cmd);

		if (_IOC_DIR(cmd) & _IOC_WRITE) {
			in_iov = iov;
			in_iovs = 1;
		}

		if (_IOC_DIR(cmd) & _IOC_READ) {
			out_iov = iov;
			out_iovs = 1;
		}
	}

 retry:
	inarg.in_size = in_size = iov_length(in_iov, in_iovs);
	inarg.out_size = out_size = iov_length(out_iov, out_iovs);

	/*
	 * Out data can be used either for actual out data or iovs,
	 * make sure there always is at least one page.
	 */
	out_size = max_t(size_t, out_size, PAGE_SIZE);
	max_pages = DIV_ROUND_UP(max(in_size, out_size), PAGE_SIZE);

	/* make sure there are enough buffer pages and init request with them */
	err = -ENOMEM;
	if (max_pages > FUSE_MAX_PAGES_PER_REQ)
		goto out;
	while (num_pages < max_pages) {
		pages[num_pages] = alloc_page(GFP_KERNEL | __GFP_HIGHMEM);
		if (!pages[num_pages])
			goto out;
		num_pages++;
	}

2595
	req = fuse_get_req(fc, num_pages);
Tejun Heo's avatar
Tejun Heo committed
2596 2597 2598 2599 2600 2601 2602
	if (IS_ERR(req)) {
		err = PTR_ERR(req);
		req = NULL;
		goto out;
	}
	memcpy(req->pages, pages, sizeof(req->pages[0]) * num_pages);
	req->num_pages = num_pages;
2603
	fuse_page_descs_length_init(req, 0, req->num_pages);
Tejun Heo's avatar
Tejun Heo committed
2604 2605 2606

	/* okay, let's send it to the client */
	req->in.h.opcode = FUSE_IOCTL;
2607
	req->in.h.nodeid = ff->nodeid;
Tejun Heo's avatar
Tejun Heo committed
2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(inarg);
	req->in.args[0].value = &inarg;
	if (in_size) {
		req->in.numargs++;
		req->in.args[1].size = in_size;
		req->in.argpages = 1;

		err = fuse_ioctl_copy_user(pages, in_iov, in_iovs, in_size,
					   false);
		if (err)
			goto out;
	}

	req->out.numargs = 2;
	req->out.args[0].size = sizeof(outarg);
	req->out.args[0].value = &outarg;
	req->out.args[1].size = out_size;
	req->out.argpages = 1;
	req->out.argvar = 1;

2629
	fuse_request_send(fc, req);
Tejun Heo's avatar
Tejun Heo committed
2630 2631 2632 2633 2634 2635 2636 2637 2638
	err = req->out.h.error;
	transferred = req->out.args[1].size;
	fuse_put_request(fc, req);
	req = NULL;
	if (err)
		goto out;

	/* did it ask for retry? */
	if (outarg.flags & FUSE_IOCTL_RETRY) {
Miklos Szeredi's avatar
Miklos Szeredi committed
2639
		void *vaddr;
Tejun Heo's avatar
Tejun Heo committed
2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658

		/* no retry if in restricted mode */
		err = -EIO;
		if (!(flags & FUSE_IOCTL_UNRESTRICTED))
			goto out;

		in_iovs = outarg.in_iovs;
		out_iovs = outarg.out_iovs;

		/*
		 * Make sure things are in boundary, separate checks
		 * are to protect against overflow.
		 */
		err = -ENOMEM;
		if (in_iovs > FUSE_IOCTL_MAX_IOV ||
		    out_iovs > FUSE_IOCTL_MAX_IOV ||
		    in_iovs + out_iovs > FUSE_IOCTL_MAX_IOV)
			goto out;

2659
		vaddr = kmap_atomic(pages[0]);
Miklos Szeredi's avatar
Miklos Szeredi committed
2660
		err = fuse_copy_ioctl_iovec(fc, iov_page, vaddr,
2661 2662
					    transferred, in_iovs + out_iovs,
					    (flags & FUSE_IOCTL_COMPAT) != 0);
2663
		kunmap_atomic(vaddr);
2664 2665
		if (err)
			goto out;
Tejun Heo's avatar
Tejun Heo committed
2666

Miklos Szeredi's avatar
Miklos Szeredi committed
2667
		in_iov = iov_page;
Tejun Heo's avatar
Tejun Heo committed
2668 2669
		out_iov = in_iov + in_iovs;

Miklos Szeredi's avatar
Miklos Szeredi committed
2670 2671 2672 2673 2674 2675 2676 2677
		err = fuse_verify_ioctl_iov(in_iov, in_iovs);
		if (err)
			goto out;

		err = fuse_verify_ioctl_iov(out_iov, out_iovs);
		if (err)
			goto out;

Tejun Heo's avatar
Tejun Heo committed
2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688
		goto retry;
	}

	err = -EIO;
	if (transferred > inarg.out_size)
		goto out;

	err = fuse_ioctl_copy_user(pages, out_iov, out_iovs, transferred, true);
 out:
	if (req)
		fuse_put_request(fc, req);
Miklos Szeredi's avatar
Miklos Szeredi committed
2689
	free_page((unsigned long) iov_page);
Tejun Heo's avatar
Tejun Heo committed
2690 2691 2692 2693 2694 2695
	while (num_pages)
		__free_page(pages[--num_pages]);
	kfree(pages);

	return err ? err : outarg.result;
}
2696
EXPORT_SYMBOL_GPL(fuse_do_ioctl);
Tejun Heo's avatar
Tejun Heo committed
2697

2698 2699
long fuse_ioctl_common(struct file *file, unsigned int cmd,
		       unsigned long arg, unsigned int flags)
2700
{
2701
	struct inode *inode = file_inode(file);
2702 2703
	struct fuse_conn *fc = get_fuse_conn(inode);

2704
	if (!fuse_allow_current_process(fc))
2705 2706 2707 2708 2709 2710 2711 2712
		return -EACCES;

	if (is_bad_inode(inode))
		return -EIO;

	return fuse_do_ioctl(file, cmd, arg, flags);
}

Tejun Heo's avatar
Tejun Heo committed
2713 2714 2715
static long fuse_file_ioctl(struct file *file, unsigned int cmd,
			    unsigned long arg)
{
2716
	return fuse_ioctl_common(file, cmd, arg, 0);
Tejun Heo's avatar
Tejun Heo committed
2717 2718 2719 2720 2721
}

static long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
				   unsigned long arg)
{
2722
	return fuse_ioctl_common(file, cmd, arg, FUSE_IOCTL_COMPAT);
Tejun Heo's avatar
Tejun Heo committed
2723 2724
}

Tejun Heo's avatar
Tejun Heo committed
2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765
/*
 * All files which have been polled are linked to RB tree
 * fuse_conn->polled_files which is indexed by kh.  Walk the tree and
 * find the matching one.
 */
static struct rb_node **fuse_find_polled_node(struct fuse_conn *fc, u64 kh,
					      struct rb_node **parent_out)
{
	struct rb_node **link = &fc->polled_files.rb_node;
	struct rb_node *last = NULL;

	while (*link) {
		struct fuse_file *ff;

		last = *link;
		ff = rb_entry(last, struct fuse_file, polled_node);

		if (kh < ff->kh)
			link = &last->rb_left;
		else if (kh > ff->kh)
			link = &last->rb_right;
		else
			return link;
	}

	if (parent_out)
		*parent_out = last;
	return link;
}

/*
 * The file is about to be polled.  Make sure it's on the polled_files
 * RB tree.  Note that files once added to the polled_files tree are
 * not removed before the file is released.  This is because a file
 * polled once is likely to be polled again.
 */
static void fuse_register_polled_file(struct fuse_conn *fc,
				      struct fuse_file *ff)
{
	spin_lock(&fc->lock);
	if (RB_EMPTY_NODE(&ff->polled_node)) {
2766
		struct rb_node **link, *uninitialized_var(parent);
Tejun Heo's avatar
Tejun Heo committed
2767 2768 2769 2770 2771 2772 2773 2774 2775

		link = fuse_find_polled_node(fc, ff->kh, &parent);
		BUG_ON(*link);
		rb_link_node(&ff->polled_node, parent, link);
		rb_insert_color(&ff->polled_node, &fc->polled_files);
	}
	spin_unlock(&fc->lock);
}

2776
unsigned fuse_file_poll(struct file *file, poll_table *wait)
Tejun Heo's avatar
Tejun Heo committed
2777 2778
{
	struct fuse_file *ff = file->private_data;
2779
	struct fuse_conn *fc = ff->fc;
Tejun Heo's avatar
Tejun Heo committed
2780 2781
	struct fuse_poll_in inarg = { .fh = ff->fh, .kh = ff->kh };
	struct fuse_poll_out outarg;
2782
	FUSE_ARGS(args);
Tejun Heo's avatar
Tejun Heo committed
2783 2784 2785 2786 2787 2788
	int err;

	if (fc->no_poll)
		return DEFAULT_POLLMASK;

	poll_wait(file, &ff->poll_wait, wait);
Enke Chen's avatar
Enke Chen committed
2789
	inarg.events = (__u32)poll_requested_events(wait);
Tejun Heo's avatar
Tejun Heo committed
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799

	/*
	 * Ask for notification iff there's someone waiting for it.
	 * The client may ignore the flag and always notify.
	 */
	if (waitqueue_active(&ff->poll_wait)) {
		inarg.flags |= FUSE_POLL_SCHEDULE_NOTIFY;
		fuse_register_polled_file(fc, ff);
	}

2800 2801 2802 2803 2804 2805 2806 2807 2808
	args.in.h.opcode = FUSE_POLL;
	args.in.h.nodeid = ff->nodeid;
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.out.numargs = 1;
	args.out.args[0].size = sizeof(outarg);
	args.out.args[0].value = &outarg;
	err = fuse_simple_request(fc, &args);
Tejun Heo's avatar
Tejun Heo committed
2809 2810 2811 2812 2813 2814 2815 2816 2817

	if (!err)
		return outarg.revents;
	if (err == -ENOSYS) {
		fc->no_poll = 1;
		return DEFAULT_POLLMASK;
	}
	return POLLERR;
}
2818
EXPORT_SYMBOL_GPL(fuse_file_poll);
Tejun Heo's avatar
Tejun Heo committed
2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843

/*
 * This is called from fuse_handle_notify() on FUSE_NOTIFY_POLL and
 * wakes up the poll waiters.
 */
int fuse_notify_poll_wakeup(struct fuse_conn *fc,
			    struct fuse_notify_poll_wakeup_out *outarg)
{
	u64 kh = outarg->kh;
	struct rb_node **link;

	spin_lock(&fc->lock);

	link = fuse_find_polled_node(fc, kh, NULL);
	if (*link) {
		struct fuse_file *ff;

		ff = rb_entry(*link, struct fuse_file, polled_node);
		wake_up_interruptible_sync(&ff->poll_wait);
	}

	spin_unlock(&fc->lock);
	return 0;
}

2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857
static void fuse_do_truncate(struct file *file)
{
	struct inode *inode = file->f_mapping->host;
	struct iattr attr;

	attr.ia_valid = ATTR_SIZE;
	attr.ia_size = i_size_read(inode);

	attr.ia_file = file;
	attr.ia_valid |= ATTR_FILE;

	fuse_do_setattr(inode, &attr, file);
}

2858 2859 2860 2861 2862
static inline loff_t fuse_round_up(loff_t off)
{
	return round_up(off, FUSE_MAX_PAGES_PER_REQ << PAGE_SHIFT);
}

2863
static ssize_t
2864
fuse_direct_IO(struct kiocb *iocb, struct iov_iter *iter, loff_t offset)
2865
{
2866
	DECLARE_COMPLETION_ONSTACK(wait);
2867
	ssize_t ret = 0;
2868 2869
	struct file *file = iocb->ki_filp;
	struct fuse_file *ff = file->private_data;
2870
	bool async_dio = ff->fc->async_dio;
2871
	loff_t pos = 0;
2872 2873
	struct inode *inode;
	loff_t i_size;
2874
	size_t count = iov_iter_count(iter);
2875
	struct fuse_io_priv *io;
2876
	bool is_sync = is_sync_kiocb(iocb);
2877 2878

	pos = offset;
2879 2880
	inode = file->f_mapping->host;
	i_size = i_size_read(inode);
2881

2882
	if ((iov_iter_rw(iter) == READ) && (offset > i_size))
2883 2884
		return 0;

2885
	/* optimization for short read */
2886
	if (async_dio && iov_iter_rw(iter) != WRITE && offset + count > i_size) {
2887 2888
		if (offset >= i_size)
			return 0;
2889 2890
		iov_iter_truncate(iter, fuse_round_up(i_size - offset));
		count = iov_iter_count(iter);
2891 2892
	}

2893
	io = kmalloc(sizeof(struct fuse_io_priv), GFP_KERNEL);
2894 2895
	if (!io)
		return -ENOMEM;
2896
	spin_lock_init(&io->lock);
2897
	kref_init(&io->refcnt);
2898 2899 2900 2901
	io->reqs = 1;
	io->bytes = -1;
	io->size = 0;
	io->offset = offset;
2902
	io->write = (iov_iter_rw(iter) == WRITE);
2903
	io->err = 0;
2904
	io->file = file;
2905 2906
	/*
	 * By default, we want to optimize all I/Os with async request
2907
	 * submission to the client filesystem if supported.
2908
	 */
2909
	io->async = async_dio;
2910 2911 2912 2913 2914 2915 2916
	io->iocb = iocb;

	/*
	 * We cannot asynchronously extend the size of a file. We have no method
	 * to wait on real async I/O requests, so we must submit this request
	 * synchronously.
	 */
2917
	if (!is_sync && (offset + count > i_size) &&
2918
	    iov_iter_rw(iter) == WRITE)
2919
		io->async = false;
2920

2921 2922 2923 2924 2925 2926
	if (io->async && is_sync) {
		/*
		 * Additional reference to keep io around after
		 * calling fuse_aio_complete()
		 */
		kref_get(&io->refcnt);
2927
		io->done = &wait;
2928
	}
2929

2930
	if (iov_iter_rw(iter) == WRITE) {
2931
		ret = fuse_direct_io(io, iter, &pos, FUSE_DIO_WRITE);
2932 2933
		fuse_invalidate_attr(inode);
	} else {
2934
		ret = __fuse_direct_read(io, iter, &pos);
2935
	}
2936

2937 2938 2939 2940
	if (io->async) {
		fuse_aio_complete(io, ret < 0 ? ret : 0, -1);

		/* we have a non-extending, async request, so return */
2941
		if (!is_sync)
2942 2943
			return -EIOCBQUEUED;

2944 2945
		wait_for_completion(&wait);
		ret = fuse_get_res_by_io(io);
2946 2947
	}

2948
	kref_put(&io->refcnt, fuse_io_release);
2949

2950
	if (iov_iter_rw(iter) == WRITE) {
2951 2952 2953 2954 2955
		if (ret > 0)
			fuse_write_update_size(inode, pos);
		else if (ret < 0 && offset + count > i_size)
			fuse_do_truncate(file);
	}
2956 2957 2958 2959

	return ret;
}

2960 2961
static long fuse_file_fallocate(struct file *file, int mode, loff_t offset,
				loff_t length)
2962 2963
{
	struct fuse_file *ff = file->private_data;
2964
	struct inode *inode = file_inode(file);
2965
	struct fuse_inode *fi = get_fuse_inode(inode);
2966
	struct fuse_conn *fc = ff->fc;
2967
	FUSE_ARGS(args);
2968 2969 2970 2971 2972 2973 2974
	struct fuse_fallocate_in inarg = {
		.fh = ff->fh,
		.offset = offset,
		.length = length,
		.mode = mode
	};
	int err;
2975 2976
	bool lock_inode = !(mode & FALLOC_FL_KEEP_SIZE) ||
			   (mode & FALLOC_FL_PUNCH_HOLE);
2977

Miklos Szeredi's avatar
Miklos Szeredi committed
2978 2979 2980
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

2981 2982 2983
	if (fc->no_fallocate)
		return -EOPNOTSUPP;

2984
	if (lock_inode) {
Al Viro's avatar
Al Viro committed
2985
		inode_lock(inode);
2986 2987 2988 2989 2990 2991 2992 2993 2994
		if (mode & FALLOC_FL_PUNCH_HOLE) {
			loff_t endbyte = offset + length - 1;
			err = filemap_write_and_wait_range(inode->i_mapping,
							   offset, endbyte);
			if (err)
				goto out;

			fuse_sync_writes(inode);
		}
2995 2996
	}

2997 2998 2999
	if (!(mode & FALLOC_FL_KEEP_SIZE))
		set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);

3000 3001 3002 3003 3004 3005
	args.in.h.opcode = FUSE_FALLOCATE;
	args.in.h.nodeid = ff->nodeid;
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	err = fuse_simple_request(fc, &args);
3006 3007 3008 3009
	if (err == -ENOSYS) {
		fc->no_fallocate = 1;
		err = -EOPNOTSUPP;
	}
3010 3011 3012 3013
	if (err)
		goto out;

	/* we could have extended the file */
3014 3015 3016
	if (!(mode & FALLOC_FL_KEEP_SIZE)) {
		bool changed = fuse_write_update_size(inode, offset + length);

3017 3018
		if (changed && fc->writeback_cache)
			file_update_time(file);
3019
	}
3020 3021 3022 3023 3024 3025

	if (mode & FALLOC_FL_PUNCH_HOLE)
		truncate_pagecache_range(inode, offset, offset + length - 1);

	fuse_invalidate_attr(inode);

3026
out:
3027 3028 3029
	if (!(mode & FALLOC_FL_KEEP_SIZE))
		clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);

3030
	if (lock_inode)
Al Viro's avatar
Al Viro committed
3031
		inode_unlock(inode);
3032

3033 3034 3035
	return err;
}

3036
static const struct file_operations fuse_file_operations = {
Miklos Szeredi's avatar
Miklos Szeredi committed
3037
	.llseek		= fuse_file_llseek,
3038
	.read_iter	= fuse_file_read_iter,
Al Viro's avatar
Al Viro committed
3039
	.write_iter	= fuse_file_write_iter,
3040 3041 3042 3043 3044
	.mmap		= fuse_file_mmap,
	.open		= fuse_open,
	.flush		= fuse_flush,
	.release	= fuse_release,
	.fsync		= fuse_fsync,
3045
	.lock		= fuse_file_lock,
3046
	.flock		= fuse_file_flock,
3047
	.splice_read	= generic_file_splice_read,
Tejun Heo's avatar
Tejun Heo committed
3048 3049
	.unlocked_ioctl	= fuse_file_ioctl,
	.compat_ioctl	= fuse_file_compat_ioctl,
Tejun Heo's avatar
Tejun Heo committed
3050
	.poll		= fuse_file_poll,
3051
	.fallocate	= fuse_file_fallocate,
3052 3053
};

3054
static const struct file_operations fuse_direct_io_file_operations = {
Miklos Szeredi's avatar
Miklos Szeredi committed
3055
	.llseek		= fuse_file_llseek,
3056 3057
	.read_iter	= fuse_direct_read_iter,
	.write_iter	= fuse_direct_write_iter,
3058
	.mmap		= fuse_direct_mmap,
Miklos Szeredi's avatar
Miklos Szeredi committed
3059 3060 3061 3062
	.open		= fuse_open,
	.flush		= fuse_flush,
	.release	= fuse_release,
	.fsync		= fuse_fsync,
3063
	.lock		= fuse_file_lock,
3064
	.flock		= fuse_file_flock,
Tejun Heo's avatar
Tejun Heo committed
3065 3066
	.unlocked_ioctl	= fuse_file_ioctl,
	.compat_ioctl	= fuse_file_compat_ioctl,
Tejun Heo's avatar
Tejun Heo committed
3067
	.poll		= fuse_file_poll,
3068
	.fallocate	= fuse_file_fallocate,
3069
	/* no splice_read */
Miklos Szeredi's avatar
Miklos Szeredi committed
3070 3071
};

3072
static const struct address_space_operations fuse_file_aops  = {
3073
	.readpage	= fuse_readpage,
Miklos Szeredi's avatar
Miklos Szeredi committed
3074
	.writepage	= fuse_writepage,
3075
	.writepages	= fuse_writepages,
Miklos Szeredi's avatar
Miklos Szeredi committed
3076
	.launder_page	= fuse_launder_page,
3077
	.readpages	= fuse_readpages,
Miklos Szeredi's avatar
Miklos Szeredi committed
3078
	.set_page_dirty	= __set_page_dirty_nobuffers,
3079
	.bmap		= fuse_bmap,
3080
	.direct_IO	= fuse_direct_IO,
3081 3082
	.write_begin	= fuse_write_begin,
	.write_end	= fuse_write_end,
3083 3084 3085 3086
};

void fuse_init_file_inode(struct inode *inode)
{
3087 3088
	inode->i_fop = &fuse_file_operations;
	inode->i_data.a_ops = &fuse_file_aops;
3089
}