file.c 76.3 KB
Newer Older
M
Miklos Szeredi 已提交
1 2
/*
  FUSE: Filesystem in Userspace
M
Miklos Szeredi 已提交
3
  Copyright (C) 2001-2008  Miklos Szeredi <miklos@szeredi.hu>
M
Miklos Szeredi 已提交
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>
A
Alexey Dobriyan 已提交
14
#include <linux/sched.h>
15
#include <linux/module.h>
16
#include <linux/compat.h>
17
#include <linux/swap.h>
18
#include <linux/aio.h>
19
#include <linux/falloc.h>
M
Miklos Szeredi 已提交
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)
M
Miklos Szeredi 已提交
25 26
{
	struct fuse_open_in inarg;
M
Miklos Szeredi 已提交
27 28 29
	struct fuse_req *req;
	int err;

M
Maxim Patlasov 已提交
30
	req = fuse_get_req_nopages(fc);
31 32
	if (IS_ERR(req))
		return PTR_ERR(req);
M
Miklos Szeredi 已提交
33 34

	memset(&inarg, 0, sizeof(inarg));
35 36 37
	inarg.flags = file->f_flags & ~(O_CREAT | O_EXCL | O_NOCTTY);
	if (!fc->atomic_o_trunc)
		inarg.flags &= ~O_TRUNC;
38 39
	req->in.h.opcode = opcode;
	req->in.h.nodeid = nodeid;
M
Miklos Szeredi 已提交
40 41 42 43 44 45
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(inarg);
	req->in.args[0].value = &inarg;
	req->out.numargs = 1;
	req->out.args[0].size = sizeof(*outargp);
	req->out.args[0].value = outargp;
46
	fuse_request_send(fc, req);
M
Miklos Szeredi 已提交
47 48 49 50 51 52
	err = req->out.h.error;
	fuse_put_request(fc, req);

	return err;
}

T
Tejun Heo 已提交
53
struct fuse_file *fuse_file_alloc(struct fuse_conn *fc)
M
Miklos Szeredi 已提交
54 55
{
	struct fuse_file *ff;
T
Tejun Heo 已提交
56

M
Miklos Szeredi 已提交
57
	ff = kmalloc(sizeof(struct fuse_file), GFP_KERNEL);
T
Tejun Heo 已提交
58 59 60
	if (unlikely(!ff))
		return NULL;

61
	ff->fc = fc;
62
	ff->reserved_req = fuse_request_alloc(0);
T
Tejun Heo 已提交
63 64 65
	if (unlikely(!ff->reserved_req)) {
		kfree(ff);
		return NULL;
M
Miklos Szeredi 已提交
66
	}
T
Tejun Heo 已提交
67 68 69 70 71 72 73 74 75 76

	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);

M
Miklos Szeredi 已提交
77 78 79 80 81
	return ff;
}

void fuse_file_free(struct fuse_file *ff)
{
82
	fuse_request_free(ff->reserved_req);
M
Miklos Szeredi 已提交
83 84 85
	kfree(ff);
}

86
struct fuse_file *fuse_file_get(struct fuse_file *ff)
87 88 89 90 91
{
	atomic_inc(&ff->count);
	return ff;
}

92 93 94 95 96 97 98 99 100 101 102 103 104 105
static void fuse_release_async(struct work_struct *work)
{
	struct fuse_req *req;
	struct fuse_conn *fc;
	struct path path;

	req = container_of(work, struct fuse_req, misc.release.work);
	path = req->misc.release.path;
	fc = get_fuse_conn(path.dentry->d_inode);

	fuse_put_request(fc, req);
	path_put(&path);
}

M
Miklos Szeredi 已提交
106 107
static void fuse_release_end(struct fuse_conn *fc, struct fuse_req *req)
{
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122
	if (fc->destroy_req) {
		/*
		 * If this is a fuseblk mount, then it's possible that
		 * releasing the path will result in releasing the
		 * super block and sending the DESTROY request.  If
		 * the server is single threaded, this would hang.
		 * For this reason do the path_put() in a separate
		 * thread.
		 */
		atomic_inc(&req->count);
		INIT_WORK(&req->misc.release.work, fuse_release_async);
		schedule_work(&req->misc.release.work);
	} else {
		path_put(&req->misc.release.path);
	}
M
Miklos Szeredi 已提交
123 124
}

125
static void fuse_file_put(struct fuse_file *ff, bool sync)
126 127 128
{
	if (atomic_dec_and_test(&ff->count)) {
		struct fuse_req *req = ff->reserved_req;
129

130 131 132 133 134 135 136 137 138
		if (ff->fc->no_open) {
			/*
			 * Drop the release request when client does not
			 * implement 'open'
			 */
			req->background = 0;
			path_put(&req->misc.release.path);
			fuse_put_request(ff->fc, req);
		} else if (sync) {
139
			req->background = 0;
140 141 142 143 144
			fuse_request_send(ff->fc, req);
			path_put(&req->misc.release.path);
			fuse_put_request(ff->fc, req);
		} else {
			req->end = fuse_release_end;
145
			req->background = 1;
146 147
			fuse_request_send_background(ff->fc, req);
		}
148 149 150 151
		kfree(ff);
	}
}

152 153
int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
		 bool isdir)
154 155 156 157 158 159 160 161
{
	struct fuse_file *ff;
	int opcode = isdir ? FUSE_OPENDIR : FUSE_OPEN;

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

162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
	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;
		}
179 180 181
	}

	if (isdir)
182
		ff->open_flags &= ~FOPEN_DIRECT_IO;
183 184 185 186 187 188

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

	return 0;
}
189
EXPORT_SYMBOL_GPL(fuse_do_open);
190

191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
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);
}

207
void fuse_finish_open(struct inode *inode, struct file *file)
M
Miklos Szeredi 已提交
208
{
209
	struct fuse_file *ff = file->private_data;
210
	struct fuse_conn *fc = get_fuse_conn(inode);
211 212

	if (ff->open_flags & FOPEN_DIRECT_IO)
M
Miklos Szeredi 已提交
213
		file->f_op = &fuse_direct_io_file_operations;
214
	if (!(ff->open_flags & FOPEN_KEEP_CACHE))
M
Miklos Szeredi 已提交
215
		invalidate_inode_pages2(inode->i_mapping);
216
	if (ff->open_flags & FOPEN_NONSEEKABLE)
T
Tejun Heo 已提交
217
		nonseekable_open(inode, file);
218 219 220 221 222 223 224 225
	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);
226 227
		if (fc->writeback_cache)
			file_update_time(file);
228
	}
P
Pavel Emelyanov 已提交
229 230
	if ((file->f_mode & FMODE_WRITE) && fc->writeback_cache)
		fuse_link_write_file(file);
M
Miklos Szeredi 已提交
231 232
}

233
int fuse_open_common(struct inode *inode, struct file *file, bool isdir)
M
Miklos Szeredi 已提交
234
{
T
Tejun Heo 已提交
235
	struct fuse_conn *fc = get_fuse_conn(inode);
M
Miklos Szeredi 已提交
236
	int err;
237 238 239
	bool lock_inode = (file->f_flags & O_TRUNC) &&
			  fc->atomic_o_trunc &&
			  fc->writeback_cache;
M
Miklos Szeredi 已提交
240 241 242 243 244

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

245 246 247
	if (lock_inode)
		mutex_lock(&inode->i_mutex);

248
	err = fuse_do_open(fc, get_node_id(inode), file, isdir);
M
Miklos Szeredi 已提交
249

250 251
	if (!err)
		fuse_finish_open(inode, file);
252

253 254 255 256
	if (lock_inode)
		mutex_unlock(&inode->i_mutex);

	return err;
M
Miklos Szeredi 已提交
257 258
}

259
static void fuse_prepare_release(struct fuse_file *ff, int flags, int opcode)
260
{
261
	struct fuse_conn *fc = ff->fc;
262
	struct fuse_req *req = ff->reserved_req;
263
	struct fuse_release_in *inarg = &req->misc.release.in;
M
Miklos Szeredi 已提交
264

265 266 267 268 269 270
	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);

271
	wake_up_interruptible_all(&ff->poll_wait);
272

M
Miklos Szeredi 已提交
273
	inarg->fh = ff->fh;
M
Miklos Szeredi 已提交
274
	inarg->flags = flags;
275
	req->in.h.opcode = opcode;
276
	req->in.h.nodeid = ff->nodeid;
M
Miklos Szeredi 已提交
277 278 279
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(struct fuse_release_in);
	req->in.args[0].value = inarg;
M
Miklos Szeredi 已提交
280 281
}

282
void fuse_release_common(struct file *file, int opcode)
M
Miklos Szeredi 已提交
283
{
T
Tejun Heo 已提交
284 285
	struct fuse_file *ff;
	struct fuse_req *req;
M
Miklos Szeredi 已提交
286

T
Tejun Heo 已提交
287 288
	ff = file->private_data;
	if (unlikely(!ff))
289
		return;
T
Tejun Heo 已提交
290 291

	req = ff->reserved_req;
292
	fuse_prepare_release(ff, file->f_flags, opcode);
T
Tejun Heo 已提交
293

M
Miklos Szeredi 已提交
294 295 296 297 298 299
	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);
	}
T
Tejun Heo 已提交
300
	/* Hold vfsmount and dentry until release is finished */
301 302
	path_get(&file->f_path);
	req->misc.release.path = file->f_path;
T
Tejun Heo 已提交
303 304 305 306 307

	/*
	 * Normally this will send the RELEASE request, however if
	 * some asynchronous READ or WRITE requests are outstanding,
	 * the sending will be delayed.
308 309 310 311
	 *
	 * 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.
T
Tejun Heo 已提交
312
	 */
313
	fuse_file_put(ff, ff->fc->destroy_req != NULL);
M
Miklos Szeredi 已提交
314 315
}

316 317
static int fuse_open(struct inode *inode, struct file *file)
{
318
	return fuse_open_common(inode, file, false);
319 320 321 322
}

static int fuse_release(struct inode *inode, struct file *file)
{
P
Pavel Emelyanov 已提交
323 324 325 326
	struct fuse_conn *fc = get_fuse_conn(inode);

	/* see fuse_vma_close() for !writeback_cache case */
	if (fc->writeback_cache)
M
Miklos Szeredi 已提交
327
		write_inode_now(inode, 1);
M
Maxim Patlasov 已提交
328

329 330 331 332 333 334 335 336 337 338 339
	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);
	ff->reserved_req->force = 1;
340
	ff->reserved_req->background = 0;
341 342 343
	fuse_request_send(ff->fc, ff->reserved_req);
	fuse_put_request(ff->fc, ff->reserved_req);
	kfree(ff);
344
}
345
EXPORT_SYMBOL_GPL(fuse_sync_release);
346

347
/*
348 349
 * Scramble the ID space with XTEA, so that the value of the files_struct
 * pointer is not exposed to userspace.
350
 */
351
u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id)
352
{
353 354 355 356 357 358 359 360 361 362 363 364 365 366
	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);
367 368
}

M
Miklos Szeredi 已提交
369
/*
370
 * Check if any page in a range is under writeback
M
Miklos Szeredi 已提交
371 372 373 374
 *
 * This is currently done by walking the list of writepage requests
 * for the inode, which can be pretty inefficient.
 */
375 376
static bool fuse_range_is_writeback(struct inode *inode, pgoff_t idx_from,
				   pgoff_t idx_to)
M
Miklos Szeredi 已提交
377 378 379 380 381 382 383 384 385 386 387 388
{
	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);
		curr_index = req->misc.write.in.offset >> PAGE_CACHE_SHIFT;
389 390
		if (idx_from < curr_index + req->num_pages &&
		    curr_index <= idx_to) {
M
Miklos Szeredi 已提交
391 392 393 394 395 396 397 398 399
			found = true;
			break;
		}
	}
	spin_unlock(&fc->lock);

	return found;
}

400 401 402 403 404
static inline bool fuse_page_is_writeback(struct inode *inode, pgoff_t index)
{
	return fuse_range_is_writeback(inode, index, index);
}

M
Miklos Szeredi 已提交
405 406 407 408 409 410 411 412 413 414 415 416 417 418
/*
 * 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;
}

419 420 421 422 423 424 425 426 427 428 429 430 431 432 433
/*
 * 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);
}

434
static int fuse_flush(struct file *file, fl_owner_t id)
M
Miklos Szeredi 已提交
435
{
A
Al Viro 已提交
436
	struct inode *inode = file_inode(file);
M
Miklos Szeredi 已提交
437 438 439 440 441 442
	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;

443 444 445
	if (is_bad_inode(inode))
		return -EIO;

M
Miklos Szeredi 已提交
446 447 448
	if (fc->no_flush)
		return 0;

M
Miklos Szeredi 已提交
449
	err = write_inode_now(inode, 1);
450 451 452 453 454 455 456
	if (err)
		return err;

	mutex_lock(&inode->i_mutex);
	fuse_sync_writes(inode);
	mutex_unlock(&inode->i_mutex);

M
Maxim Patlasov 已提交
457
	req = fuse_get_req_nofail_nopages(fc, file);
M
Miklos Szeredi 已提交
458 459
	memset(&inarg, 0, sizeof(inarg));
	inarg.fh = ff->fh;
460
	inarg.lock_owner = fuse_lock_owner_id(fc, id);
M
Miklos Szeredi 已提交
461 462 463 464 465
	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;
466
	req->force = 1;
467
	fuse_request_send(fc, req);
M
Miklos Szeredi 已提交
468 469 470 471 472 473 474 475 476
	err = req->out.h.error;
	fuse_put_request(fc, req);
	if (err == -ENOSYS) {
		fc->no_flush = 1;
		err = 0;
	}
	return err;
}

477 478
int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
		      int datasync, int isdir)
M
Miklos Szeredi 已提交
479
{
480
	struct inode *inode = file->f_mapping->host;
M
Miklos Szeredi 已提交
481 482 483 484 485 486
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_file *ff = file->private_data;
	struct fuse_req *req;
	struct fuse_fsync_in inarg;
	int err;

487 488 489
	if (is_bad_inode(inode))
		return -EIO;

490 491
	mutex_lock(&inode->i_mutex);

M
Miklos Szeredi 已提交
492 493 494 495 496
	/*
	 * Start writeback against all dirty pages of the inode, then
	 * wait for all outstanding writes, before sending the FSYNC
	 * request.
	 */
M
Miklos Szeredi 已提交
497
	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
M
Miklos Szeredi 已提交
498
	if (err)
499
		goto out;
M
Miklos Szeredi 已提交
500 501

	fuse_sync_writes(inode);
M
Miklos Szeredi 已提交
502 503 504
	err = sync_inode_metadata(inode, 1);
	if (err)
		goto out;
M
Miklos Szeredi 已提交
505

M
Miklos Szeredi 已提交
506 507
	if ((!isdir && fc->no_fsync) || (isdir && fc->no_fsyncdir))
		goto out;
M
Maxim Patlasov 已提交
508

M
Maxim Patlasov 已提交
509
	req = fuse_get_req_nopages(fc);
510 511 512 513
	if (IS_ERR(req)) {
		err = PTR_ERR(req);
		goto out;
	}
M
Miklos Szeredi 已提交
514 515 516 517

	memset(&inarg, 0, sizeof(inarg));
	inarg.fh = ff->fh;
	inarg.fsync_flags = datasync ? 1 : 0;
518
	req->in.h.opcode = isdir ? FUSE_FSYNCDIR : FUSE_FSYNC;
M
Miklos Szeredi 已提交
519 520 521 522
	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;
523
	fuse_request_send(fc, req);
M
Miklos Szeredi 已提交
524 525 526
	err = req->out.h.error;
	fuse_put_request(fc, req);
	if (err == -ENOSYS) {
527 528 529 530
		if (isdir)
			fc->no_fsyncdir = 1;
		else
			fc->no_fsync = 1;
M
Miklos Szeredi 已提交
531 532
		err = 0;
	}
533 534
out:
	mutex_unlock(&inode->i_mutex);
M
Miklos Szeredi 已提交
535 536 537
	return err;
}

538 539
static int fuse_fsync(struct file *file, loff_t start, loff_t end,
		      int datasync)
540
{
541
	return fuse_fsync_common(file, start, end, datasync, 0);
542 543
}

544 545
void fuse_read_fill(struct fuse_req *req, struct file *file, loff_t pos,
		    size_t count, int opcode)
M
Miklos Szeredi 已提交
546
{
547
	struct fuse_read_in *inarg = &req->misc.read.in;
548
	struct fuse_file *ff = file->private_data;
M
Miklos Szeredi 已提交
549

550 551 552
	inarg->fh = ff->fh;
	inarg->offset = pos;
	inarg->size = count;
553
	inarg->flags = file->f_flags;
554
	req->in.h.opcode = opcode;
555
	req->in.h.nodeid = ff->nodeid;
M
Miklos Szeredi 已提交
556 557
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(struct fuse_read_in);
558
	req->in.args[0].value = inarg;
M
Miklos Szeredi 已提交
559 560 561 562 563
	req->out.argvar = 1;
	req->out.numargs = 1;
	req->out.args[0].size = count;
}

564 565 566 567 568 569 570 571 572 573 574 575
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);
	}
}

M
Maxim Patlasov 已提交
576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
/**
 * 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)
{
	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;
	spin_unlock(&io->lock);

	if (!left) {
		long res;

		if (io->err)
			res = io->err;
		else if (io->bytes >= 0 && io->write)
			res = -EIO;
		else {
			res = io->bytes < 0 ? io->size : io->bytes;

			if (!is_sync_kiocb(io->iocb)) {
A
Al Viro 已提交
616
				struct inode *inode = file_inode(io->iocb->ki_filp);
M
Maxim Patlasov 已提交
617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
				struct fuse_conn *fc = get_fuse_conn(inode);
				struct fuse_inode *fi = get_fuse_inode(inode);

				spin_lock(&fc->lock);
				fi->attr_version = ++fc->attr_version;
				spin_unlock(&fc->lock);
			}
		}

		aio_complete(io->iocb, res, 0);
		kfree(io);
	}
}

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);
	io->size += num_bytes;
	io->reqs++;
	spin_unlock(&io->lock);

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

662
	__fuse_get_request(req);
M
Maxim Patlasov 已提交
663 664 665 666 667
	fuse_request_send_background(fc, req);

	return num_bytes;
}

668
static size_t fuse_send_read(struct fuse_req *req, struct fuse_io_priv *io,
669
			     loff_t pos, size_t count, fl_owner_t owner)
670
{
671
	struct file *file = io->file;
672 673
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
674

675
	fuse_read_fill(req, file, pos, count, FUSE_READ);
676
	if (owner != NULL) {
677
		struct fuse_read_in *inarg = &req->misc.read.in;
678 679 680 681

		inarg->read_flags |= FUSE_READ_LOCKOWNER;
		inarg->lock_owner = fuse_lock_owner_id(fc, owner);
	}
682 683 684 685

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

686
	fuse_request_send(fc, req);
687
	return req->out.args[0].size;
688 689
}

690 691 692 693 694 695 696
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);
697 698
	if (attr_ver == fi->attr_version && size < inode->i_size &&
	    !test_bit(FUSE_I_SIZE_UNSTABLE, &fi->state)) {
699 700 701 702 703 704
		fi->attr_version = ++fc->attr_version;
		i_size_write(inode, size);
	}
	spin_unlock(&fc->lock);
}

705 706 707 708
static void fuse_short_read(struct fuse_req *req, struct inode *inode,
			    u64 attr_ver)
{
	size_t num_read = req->out.args[0].size;
P
Pavel Emelyanov 已提交
709 710 711 712 713 714 715 716 717 718 719
	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;
		int start_idx = num_read >> PAGE_CACHE_SHIFT;
		size_t off = num_read & (PAGE_CACHE_SIZE - 1);
720

P
Pavel Emelyanov 已提交
721 722 723 724 725 726 727 728
		for (i = start_idx; i < req->num_pages; i++) {
			zero_user_segment(req->pages[i], off, PAGE_CACHE_SIZE);
			off = 0;
		}
	} else {
		loff_t pos = page_offset(req->pages[0]) + num_read;
		fuse_read_update_size(inode, pos, attr_ver);
	}
729 730
}

731
static int fuse_do_readpage(struct file *file, struct page *page)
M
Miklos Szeredi 已提交
732
{
733
	struct fuse_io_priv io = { .async = 0, .file = file };
M
Miklos Szeredi 已提交
734 735
	struct inode *inode = page->mapping->host;
	struct fuse_conn *fc = get_fuse_conn(inode);
736
	struct fuse_req *req;
737 738 739 740
	size_t num_read;
	loff_t pos = page_offset(page);
	size_t count = PAGE_CACHE_SIZE;
	u64 attr_ver;
741 742
	int err;

M
Miklos Szeredi 已提交
743
	/*
L
Lucas De Marchi 已提交
744
	 * Page writeback can extend beyond the lifetime of the
M
Miklos Szeredi 已提交
745 746 747 748 749
	 * page-cache page, so make sure we read a properly synced
	 * page.
	 */
	fuse_wait_on_page_writeback(inode, page->index);

M
Maxim Patlasov 已提交
750
	req = fuse_get_req(fc, 1);
751
	if (IS_ERR(req))
752
		return PTR_ERR(req);
M
Miklos Szeredi 已提交
753

754 755
	attr_ver = fuse_get_attr_version(fc);

M
Miklos Szeredi 已提交
756
	req->out.page_zeroing = 1;
757
	req->out.argpages = 1;
M
Miklos Szeredi 已提交
758 759
	req->num_pages = 1;
	req->pages[0] = page;
760
	req->page_descs[0].length = count;
761
	num_read = fuse_send_read(req, &io, pos, count, NULL);
M
Miklos Szeredi 已提交
762
	err = req->out.h.error;
763 764 765 766 767 768

	if (!err) {
		/*
		 * Short read means EOF.  If file size is larger, truncate it
		 */
		if (num_read < count)
769
			fuse_short_read(req, inode, attr_ver);
770

M
Miklos Szeredi 已提交
771
		SetPageUptodate(page);
772 773
	}

774
	fuse_put_request(fc, req);
775 776 777 778 779 780 781 782 783 784 785 786 787 788

	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);
789
	fuse_invalidate_atime(inode);
M
Miklos Szeredi 已提交
790 791 792 793 794
 out:
	unlock_page(page);
	return err;
}

795
static void fuse_readpages_end(struct fuse_conn *fc, struct fuse_req *req)
796
{
797
	int i;
798 799
	size_t count = req->misc.read.in.size;
	size_t num_read = req->out.args[0].size;
800
	struct address_space *mapping = NULL;
801

802 803
	for (i = 0; mapping == NULL && i < req->num_pages; i++)
		mapping = req->pages[i]->mapping;
804

805 806 807 808 809 810
	if (mapping) {
		struct inode *inode = mapping->host;

		/*
		 * Short read means EOF. If file size is larger, truncate it
		 */
811 812
		if (!req->out.h.error && num_read < count)
			fuse_short_read(req, inode, req->misc.read.attr_ver);
813

814
		fuse_invalidate_atime(inode);
815
	}
816

817 818 819 820
	for (i = 0; i < req->num_pages; i++) {
		struct page *page = req->pages[i];
		if (!req->out.h.error)
			SetPageUptodate(page);
821 822
		else
			SetPageError(page);
823
		unlock_page(page);
824
		page_cache_release(page);
825
	}
826
	if (req->ff)
827
		fuse_file_put(req->ff, false);
828 829
}

830
static void fuse_send_readpages(struct fuse_req *req, struct file *file)
831
{
832 833
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
834 835
	loff_t pos = page_offset(req->pages[0]);
	size_t count = req->num_pages << PAGE_CACHE_SHIFT;
836 837

	req->out.argpages = 1;
838
	req->out.page_zeroing = 1;
839
	req->out.page_replace = 1;
840
	fuse_read_fill(req, file, pos, count, FUSE_READ);
841
	req->misc.read.attr_ver = fuse_get_attr_version(fc);
842
	if (fc->async_read) {
843
		req->ff = fuse_file_get(ff);
844
		req->end = fuse_readpages_end;
845
		fuse_request_send_background(fc, req);
846
	} else {
847
		fuse_request_send(fc, req);
848
		fuse_readpages_end(fc, req);
849
		fuse_put_request(fc, req);
850
	}
851 852
}

853
struct fuse_fill_data {
854
	struct fuse_req *req;
855
	struct file *file;
856
	struct inode *inode;
M
Maxim Patlasov 已提交
857
	unsigned nr_pages;
858 859 860 861
};

static int fuse_readpages_fill(void *_data, struct page *page)
{
862
	struct fuse_fill_data *data = _data;
863 864 865 866
	struct fuse_req *req = data->req;
	struct inode *inode = data->inode;
	struct fuse_conn *fc = get_fuse_conn(inode);

M
Miklos Szeredi 已提交
867 868
	fuse_wait_on_page_writeback(inode, page->index);

869 870 871 872
	if (req->num_pages &&
	    (req->num_pages == FUSE_MAX_PAGES_PER_REQ ||
	     (req->num_pages + 1) * PAGE_CACHE_SIZE > fc->max_read ||
	     req->pages[req->num_pages - 1]->index + 1 != page->index)) {
M
Maxim Patlasov 已提交
873 874
		int nr_alloc = min_t(unsigned, data->nr_pages,
				     FUSE_MAX_PAGES_PER_REQ);
875
		fuse_send_readpages(req, data->file);
876 877 878 879 880 881
		if (fc->async_read)
			req = fuse_get_req_for_background(fc, nr_alloc);
		else
			req = fuse_get_req(fc, nr_alloc);

		data->req = req;
882
		if (IS_ERR(req)) {
883
			unlock_page(page);
884
			return PTR_ERR(req);
885 886
		}
	}
M
Maxim Patlasov 已提交
887 888 889 890 891 892

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

893
	page_cache_get(page);
894
	req->pages[req->num_pages] = page;
895
	req->page_descs[req->num_pages].length = PAGE_SIZE;
M
Miklos Szeredi 已提交
896
	req->num_pages++;
M
Maxim Patlasov 已提交
897
	data->nr_pages--;
898 899 900 901 902 903 904 905
	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);
906
	struct fuse_fill_data data;
907
	int err;
M
Maxim Patlasov 已提交
908
	int nr_alloc = min_t(unsigned, nr_pages, FUSE_MAX_PAGES_PER_REQ);
909

910
	err = -EIO;
911
	if (is_bad_inode(inode))
912
		goto out;
913

914
	data.file = file;
915
	data.inode = inode;
916 917 918 919
	if (fc->async_read)
		data.req = fuse_get_req_for_background(fc, nr_alloc);
	else
		data.req = fuse_get_req(fc, nr_alloc);
M
Maxim Patlasov 已提交
920
	data.nr_pages = nr_pages;
921
	err = PTR_ERR(data.req);
922
	if (IS_ERR(data.req))
923
		goto out;
924 925

	err = read_cache_pages(mapping, pages, fuse_readpages_fill, &data);
926 927
	if (!err) {
		if (data.req->num_pages)
928
			fuse_send_readpages(data.req, file);
929 930 931
		else
			fuse_put_request(fc, data.req);
	}
932
out:
933
	return err;
934 935
}

M
Miklos Szeredi 已提交
936 937 938 939
static ssize_t fuse_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
				  unsigned long nr_segs, loff_t pos)
{
	struct inode *inode = iocb->ki_filp->f_mapping->host;
940
	struct fuse_conn *fc = get_fuse_conn(inode);
M
Miklos Szeredi 已提交
941

942 943 944 945 946 947 948
	/*
	 * 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 ||
	    (pos + iov_length(iov, nr_segs) > i_size_read(inode))) {
M
Miklos Szeredi 已提交
949 950 951 952 953 954 955 956 957
		int err;
		err = fuse_update_attributes(inode, NULL, iocb->ki_filp, NULL);
		if (err)
			return err;
	}

	return generic_file_aio_read(iocb, iov, nr_segs, pos);
}

958
static void fuse_write_fill(struct fuse_req *req, struct fuse_file *ff,
959
			    loff_t pos, size_t count)
M
Miklos Szeredi 已提交
960
{
961 962
	struct fuse_write_in *inarg = &req->misc.write.in;
	struct fuse_write_out *outarg = &req->misc.write.out;
M
Miklos Szeredi 已提交
963

964 965 966
	inarg->fh = ff->fh;
	inarg->offset = pos;
	inarg->size = count;
M
Miklos Szeredi 已提交
967
	req->in.h.opcode = FUSE_WRITE;
968
	req->in.h.nodeid = ff->nodeid;
M
Miklos Szeredi 已提交
969
	req->in.numargs = 2;
970
	if (ff->fc->minor < 9)
971 972 973
		req->in.args[0].size = FUSE_COMPAT_WRITE_IN_SIZE;
	else
		req->in.args[0].size = sizeof(struct fuse_write_in);
974
	req->in.args[0].value = inarg;
M
Miklos Szeredi 已提交
975 976 977
	req->in.args[1].size = count;
	req->out.numargs = 1;
	req->out.args[0].size = sizeof(struct fuse_write_out);
978 979 980
	req->out.args[0].value = outarg;
}

981
static size_t fuse_send_write(struct fuse_req *req, struct fuse_io_priv *io,
982
			      loff_t pos, size_t count, fl_owner_t owner)
983
{
984
	struct file *file = io->file;
985 986
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
987 988
	struct fuse_write_in *inarg = &req->misc.write.in;

989
	fuse_write_fill(req, ff, pos, count);
990
	inarg->flags = file->f_flags;
991 992 993 994
	if (owner != NULL) {
		inarg->write_flags |= FUSE_WRITE_LOCKOWNER;
		inarg->lock_owner = fuse_lock_owner_id(fc, owner);
	}
995 996 997 998

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

999
	fuse_request_send(fc, req);
1000
	return req->misc.write.out.size;
M
Miklos Szeredi 已提交
1001 1002
}

M
Maxim Patlasov 已提交
1003
bool fuse_write_update_size(struct inode *inode, loff_t pos)
1004 1005 1006
{
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
M
Maxim Patlasov 已提交
1007
	bool ret = false;
1008 1009 1010

	spin_lock(&fc->lock);
	fi->attr_version = ++fc->attr_version;
M
Maxim Patlasov 已提交
1011
	if (pos > inode->i_size) {
1012
		i_size_write(inode, pos);
M
Maxim Patlasov 已提交
1013 1014
		ret = true;
	}
1015
	spin_unlock(&fc->lock);
M
Maxim Patlasov 已提交
1016 1017

	return ret;
1018 1019
}

N
Nick Piggin 已提交
1020 1021 1022 1023 1024 1025 1026
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;
1027
	struct fuse_io_priv io = { .async = 0, .file = file };
N
Nick Piggin 已提交
1028 1029 1030 1031

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

1032
	res = fuse_send_write(req, &io, pos, count, NULL);
N
Nick Piggin 已提交
1033

1034
	offset = req->page_descs[0].offset;
N
Nick Piggin 已提交
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
	count = res;
	for (i = 0; i < req->num_pages; i++) {
		struct page *page = req->pages[i];

		if (!req->out.h.error && !offset && count >= PAGE_CACHE_SIZE)
			SetPageUptodate(page);

		if (count > PAGE_CACHE_SIZE - offset)
			count -= PAGE_CACHE_SIZE - offset;
		else
			count = 0;
		offset = 0;

		unlock_page(page);
		page_cache_release(page);
	}

	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);
	unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
	size_t count = 0;
	int err;

1064
	req->in.argpages = 1;
1065
	req->page_descs[0].offset = offset;
N
Nick Piggin 已提交
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081

	do {
		size_t tmp;
		struct page *page;
		pgoff_t index = pos >> PAGE_CACHE_SHIFT;
		size_t bytes = min_t(size_t, PAGE_CACHE_SIZE - offset,
				     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;
1082
		page = grab_cache_page_write_begin(mapping, index, 0);
N
Nick Piggin 已提交
1083 1084 1085
		if (!page)
			break;

1086 1087 1088
		if (mapping_writably_mapped(mapping))
			flush_dcache_page(page);

N
Nick Piggin 已提交
1089 1090 1091
		tmp = iov_iter_copy_from_user_atomic(page, ii, offset, bytes);
		flush_dcache_page(page);

1092 1093
		mark_page_accessed(page);

N
Nick Piggin 已提交
1094 1095 1096 1097 1098 1099 1100 1101 1102
		if (!tmp) {
			unlock_page(page);
			page_cache_release(page);
			bytes = min(bytes, iov_iter_single_seg_count(ii));
			goto again;
		}

		err = 0;
		req->pages[req->num_pages] = page;
1103
		req->page_descs[req->num_pages].length = tmp;
N
Nick Piggin 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112
		req->num_pages++;

		iov_iter_advance(ii, tmp);
		count += tmp;
		pos += tmp;
		offset += tmp;
		if (offset == PAGE_CACHE_SIZE)
			offset = 0;

1113 1114
		if (!fc->big_writes)
			break;
N
Nick Piggin 已提交
1115
	} while (iov_iter_count(ii) && count < fc->max_write &&
1116
		 req->num_pages < req->max_pages && offset == 0);
N
Nick Piggin 已提交
1117 1118 1119 1120

	return count > 0 ? count : err;
}

1121 1122 1123 1124 1125 1126 1127 1128
static inline unsigned fuse_wr_pages(loff_t pos, size_t len)
{
	return min_t(unsigned,
		     ((pos + len - 1) >> PAGE_CACHE_SHIFT) -
		     (pos >> PAGE_CACHE_SHIFT) + 1,
		     FUSE_MAX_PAGES_PER_REQ);
}

N
Nick Piggin 已提交
1129 1130 1131 1132 1133 1134
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);
1135
	struct fuse_inode *fi = get_fuse_inode(inode);
N
Nick Piggin 已提交
1136 1137 1138 1139 1140 1141
	int err = 0;
	ssize_t res = 0;

	if (is_bad_inode(inode))
		return -EIO;

1142 1143 1144
	if (inode->i_size < pos + iov_iter_count(ii))
		set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);

N
Nick Piggin 已提交
1145 1146 1147
	do {
		struct fuse_req *req;
		ssize_t count;
1148
		unsigned nr_pages = fuse_wr_pages(pos, iov_iter_count(ii));
N
Nick Piggin 已提交
1149

1150
		req = fuse_get_req(fc, nr_pages);
N
Nick Piggin 已提交
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
		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);

1180
	clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
N
Nick Piggin 已提交
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191
	fuse_invalidate_attr(inode);

	return res > 0 ? res : err;
}

static ssize_t fuse_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
				   unsigned long nr_segs, loff_t pos)
{
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	size_t count = 0;
A
Anand Avati 已提交
1192
	size_t ocount = 0;
N
Nick Piggin 已提交
1193
	ssize_t written = 0;
A
Anand Avati 已提交
1194
	ssize_t written_buffered = 0;
N
Nick Piggin 已提交
1195 1196 1197
	struct inode *inode = mapping->host;
	ssize_t err;
	struct iov_iter i;
A
Anand Avati 已提交
1198
	loff_t endbyte = 0;
N
Nick Piggin 已提交
1199

P
Pavel Emelyanov 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208
	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;

		return generic_file_aio_write(iocb, iov, nr_segs, pos);
	}

N
Nick Piggin 已提交
1209 1210
	WARN_ON(iocb->ki_pos != pos);

A
Anand Avati 已提交
1211 1212
	ocount = 0;
	err = generic_segment_checks(iov, &nr_segs, &ocount, VERIFY_READ);
N
Nick Piggin 已提交
1213 1214 1215
	if (err)
		return err;

A
Anand Avati 已提交
1216
	count = ocount;
N
Nick Piggin 已提交
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
	mutex_lock(&inode->i_mutex);

	/* We can write back this queue in page reclaim */
	current->backing_dev_info = mapping->backing_dev_info;

	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
	if (err)
		goto out;

	if (count == 0)
		goto out;

1229
	err = file_remove_suid(file);
N
Nick Piggin 已提交
1230 1231 1232
	if (err)
		goto out;

1233 1234 1235
	err = file_update_time(file);
	if (err)
		goto out;
N
Nick Piggin 已提交
1236

A
Anand Avati 已提交
1237
	if (file->f_flags & O_DIRECT) {
1238
		written = generic_file_direct_write(iocb, iov, &nr_segs, pos, 
A
Anand Avati 已提交
1239 1240 1241 1242 1243 1244
						    count, ocount);
		if (written < 0 || written == count)
			goto out;

		pos += written;
		count -= written;
N
Nick Piggin 已提交
1245

A
Anand Avati 已提交
1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
		iov_iter_init(&i, iov, nr_segs, count, written);
		written_buffered = fuse_perform_write(file, mapping, &i, pos);
		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,
					 pos >> PAGE_CACHE_SHIFT,
					 endbyte >> PAGE_CACHE_SHIFT);

		written += written_buffered;
		iocb->ki_pos = pos + written_buffered;
	} else {
		iov_iter_init(&i, iov, nr_segs, count, 0);
		written = fuse_perform_write(file, mapping, &i, pos);
		if (written >= 0)
			iocb->ki_pos = pos + written;
	}
N
Nick Piggin 已提交
1271 1272 1273 1274 1275 1276 1277
out:
	current->backing_dev_info = NULL;
	mutex_unlock(&inode->i_mutex);

	return written ? written : err;
}

1278 1279
static inline void fuse_page_descs_length_init(struct fuse_req *req,
		unsigned index, unsigned nr_pages)
1280 1281 1282
{
	int i;

1283
	for (i = index; i < index + nr_pages; i++)
1284 1285 1286 1287
		req->page_descs[i].length = PAGE_SIZE -
			req->page_descs[i].offset;
}

1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
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);
}

1299
static int fuse_get_user_pages(struct fuse_req *req, struct iov_iter *ii,
1300
			       size_t *nbytesp, int write)
M
Miklos Szeredi 已提交
1301
{
1302
	size_t nbytes = 0;  /* # bytes already packed in req */
1303

1304 1305
	/* Special case for kernel I/O: can copy directly into the buffer */
	if (segment_eq(get_fs(), KERNEL_DS)) {
1306 1307 1308
		unsigned long user_addr = fuse_get_user_addr(ii);
		size_t frag_size = fuse_get_frag_size(ii, *nbytesp);

1309 1310 1311 1312 1313
		if (write)
			req->in.args[1].value = (void *) user_addr;
		else
			req->out.args[0].value = (void *) user_addr;

1314 1315
		iov_iter_advance(ii, frag_size);
		*nbytesp = frag_size;
1316 1317
		return 0;
	}
M
Miklos Szeredi 已提交
1318

1319
	while (nbytes < *nbytesp && req->num_pages < req->max_pages) {
1320 1321 1322 1323 1324
		unsigned npages;
		unsigned long user_addr = fuse_get_user_addr(ii);
		unsigned offset = user_addr & ~PAGE_MASK;
		size_t frag_size = fuse_get_frag_size(ii, *nbytesp - nbytes);
		int ret;
M
Miklos Szeredi 已提交
1325

1326
		unsigned n = req->max_pages - req->num_pages;
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
		frag_size = min_t(size_t, frag_size, n << PAGE_SHIFT);

		npages = (frag_size + offset + PAGE_SIZE - 1) >> PAGE_SHIFT;
		npages = clamp(npages, 1U, n);

		ret = get_user_pages_fast(user_addr, npages, !write,
					  &req->pages[req->num_pages]);
		if (ret < 0)
			return ret;

		npages = ret;
		frag_size = min_t(size_t, frag_size,
				  (npages << PAGE_SHIFT) - offset);
		iov_iter_advance(ii, frag_size);

		req->page_descs[req->num_pages].offset = offset;
		fuse_page_descs_length_init(req, req->num_pages, npages);

		req->num_pages += npages;
		req->page_descs[req->num_pages - 1].length -=
			(npages << PAGE_SHIFT) - offset - frag_size;

		nbytes += frag_size;
	}
1351 1352 1353 1354 1355 1356

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

1357
	*nbytesp = nbytes;
1358

M
Miklos Szeredi 已提交
1359 1360 1361
	return 0;
}

1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
static inline int fuse_iter_npages(const struct iov_iter *ii_p)
{
	struct iov_iter ii = *ii_p;
	int npages = 0;

	while (iov_iter_count(&ii) && npages < FUSE_MAX_PAGES_PER_REQ) {
		unsigned long user_addr = fuse_get_user_addr(&ii);
		unsigned offset = user_addr & ~PAGE_MASK;
		size_t frag_size = iov_iter_single_seg_count(&ii);

		npages += (frag_size + offset + PAGE_SIZE - 1) >> PAGE_SHIFT;
		iov_iter_advance(&ii, frag_size);
	}

	return min(npages, FUSE_MAX_PAGES_PER_REQ);
}

1379
ssize_t fuse_direct_io(struct fuse_io_priv *io, const struct iovec *iov,
M
Miklos Szeredi 已提交
1380
		       unsigned long nr_segs, size_t count, loff_t *ppos,
1381
		       int flags)
M
Miklos Szeredi 已提交
1382
{
1383 1384
	int write = flags & FUSE_DIO_WRITE;
	int cuse = flags & FUSE_DIO_CUSE;
1385
	struct file *file = io->file;
1386
	struct inode *inode = file->f_mapping->host;
1387 1388
	struct fuse_file *ff = file->private_data;
	struct fuse_conn *fc = ff->fc;
M
Miklos Szeredi 已提交
1389 1390
	size_t nmax = write ? fc->max_write : fc->max_read;
	loff_t pos = *ppos;
1391 1392
	pgoff_t idx_from = pos >> PAGE_CACHE_SHIFT;
	pgoff_t idx_to = (pos + count - 1) >> PAGE_CACHE_SHIFT;
M
Miklos Szeredi 已提交
1393
	ssize_t res = 0;
1394
	struct fuse_req *req;
1395 1396 1397
	struct iov_iter ii;

	iov_iter_init(&ii, iov, nr_segs, count, 0);
1398

1399 1400 1401 1402
	if (io->async)
		req = fuse_get_req_for_background(fc, fuse_iter_npages(&ii));
	else
		req = fuse_get_req(fc, fuse_iter_npages(&ii));
1403 1404
	if (IS_ERR(req))
		return PTR_ERR(req);
M
Miklos Szeredi 已提交
1405

1406 1407 1408 1409 1410 1411 1412 1413
	if (!cuse && fuse_range_is_writeback(inode, idx_from, idx_to)) {
		if (!write)
			mutex_lock(&inode->i_mutex);
		fuse_sync_writes(inode);
		if (!write)
			mutex_unlock(&inode->i_mutex);
	}

M
Miklos Szeredi 已提交
1414 1415
	while (count) {
		size_t nres;
1416
		fl_owner_t owner = current->files;
1417
		size_t nbytes = min(count, nmax);
1418
		int err = fuse_get_user_pages(req, &ii, &nbytes, write);
M
Miklos Szeredi 已提交
1419 1420 1421 1422
		if (err) {
			res = err;
			break;
		}
1423

M
Miklos Szeredi 已提交
1424
		if (write)
1425
			nres = fuse_send_write(req, io, pos, nbytes, owner);
M
Miklos Szeredi 已提交
1426
		else
1427
			nres = fuse_send_read(req, io, pos, nbytes, owner);
1428

1429 1430
		if (!io->async)
			fuse_release_user_pages(req, !write);
M
Miklos Szeredi 已提交
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443
		if (req->out.h.error) {
			if (!res)
				res = req->out.h.error;
			break;
		} else if (nres > nbytes) {
			res = -EIO;
			break;
		}
		count -= nres;
		res += nres;
		pos += nres;
		if (nres != nbytes)
			break;
1444 1445
		if (count) {
			fuse_put_request(fc, req);
1446 1447 1448 1449 1450
			if (io->async)
				req = fuse_get_req_for_background(fc,
					fuse_iter_npages(&ii));
			else
				req = fuse_get_req(fc, fuse_iter_npages(&ii));
1451 1452 1453
			if (IS_ERR(req))
				break;
		}
M
Miklos Szeredi 已提交
1454
	}
1455 1456
	if (!IS_ERR(req))
		fuse_put_request(fc, req);
1457
	if (res > 0)
M
Miklos Szeredi 已提交
1458 1459 1460 1461
		*ppos = pos;

	return res;
}
1462
EXPORT_SYMBOL_GPL(fuse_direct_io);
M
Miklos Szeredi 已提交
1463

1464 1465
static ssize_t __fuse_direct_read(struct fuse_io_priv *io,
				  const struct iovec *iov,
1466 1467
				  unsigned long nr_segs, loff_t *ppos,
				  size_t count)
M
Miklos Szeredi 已提交
1468
{
1469
	ssize_t res;
1470
	struct file *file = io->file;
A
Al Viro 已提交
1471
	struct inode *inode = file_inode(file);
1472 1473 1474 1475

	if (is_bad_inode(inode))
		return -EIO;

1476
	res = fuse_direct_io(io, iov, nr_segs, count, ppos, 0);
1477 1478 1479 1480

	fuse_invalidate_attr(inode);

	return res;
M
Miklos Szeredi 已提交
1481 1482
}

1483 1484 1485
static ssize_t fuse_direct_read(struct file *file, char __user *buf,
				     size_t count, loff_t *ppos)
{
1486
	struct fuse_io_priv io = { .async = 0, .file = file };
M
Miklos Szeredi 已提交
1487
	struct iovec iov = { .iov_base = buf, .iov_len = count };
1488
	return __fuse_direct_read(&io, &iov, 1, ppos, count);
1489 1490
}

1491 1492
static ssize_t __fuse_direct_write(struct fuse_io_priv *io,
				   const struct iovec *iov,
1493
				   unsigned long nr_segs, loff_t *ppos)
M
Miklos Szeredi 已提交
1494
{
1495
	struct file *file = io->file;
A
Al Viro 已提交
1496
	struct inode *inode = file_inode(file);
1497
	size_t count = iov_length(iov, nr_segs);
M
Miklos Szeredi 已提交
1498
	ssize_t res;
1499

1500
	res = generic_write_checks(file, ppos, &count, 0);
1501
	if (!res)
1502 1503
		res = fuse_direct_io(io, iov, nr_segs, count, ppos,
				     FUSE_DIO_WRITE);
1504 1505 1506

	fuse_invalidate_attr(inode);

M
Miklos Szeredi 已提交
1507 1508 1509
	return res;
}

A
Anand Avati 已提交
1510 1511 1512
static ssize_t fuse_direct_write(struct file *file, const char __user *buf,
				 size_t count, loff_t *ppos)
{
M
Miklos Szeredi 已提交
1513
	struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = count };
A
Al Viro 已提交
1514
	struct inode *inode = file_inode(file);
A
Anand Avati 已提交
1515
	ssize_t res;
1516
	struct fuse_io_priv io = { .async = 0, .file = file };
A
Anand Avati 已提交
1517 1518 1519 1520 1521 1522

	if (is_bad_inode(inode))
		return -EIO;

	/* Don't allow parallel writes to the same file */
	mutex_lock(&inode->i_mutex);
1523
	res = __fuse_direct_write(&io, &iov, 1, ppos);
1524 1525
	if (res > 0)
		fuse_write_update_size(inode, *ppos);
A
Anand Avati 已提交
1526 1527 1528 1529 1530
	mutex_unlock(&inode->i_mutex);

	return res;
}

M
Miklos Szeredi 已提交
1531
static void fuse_writepage_free(struct fuse_conn *fc, struct fuse_req *req)
M
Miklos Szeredi 已提交
1532
{
1533 1534 1535 1536
	int i;

	for (i = 0; i < req->num_pages; i++)
		__free_page(req->pages[i]);
1537 1538 1539

	if (req->ff)
		fuse_file_put(req->ff, false);
M
Miklos Szeredi 已提交
1540 1541 1542 1543 1544 1545 1546
}

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);
	struct backing_dev_info *bdi = inode->i_mapping->backing_dev_info;
1547
	int i;
M
Miklos Szeredi 已提交
1548 1549

	list_del(&req->writepages_entry);
1550 1551 1552 1553 1554
	for (i = 0; i < req->num_pages; i++) {
		dec_bdi_stat(bdi, BDI_WRITEBACK);
		dec_zone_page_state(req->pages[i], NR_WRITEBACK_TEMP);
		bdi_writeout_inc(bdi);
	}
M
Miklos Szeredi 已提交
1555 1556 1557 1558
	wake_up(&fi->page_waitq);
}

/* Called under fc->lock, may release and reacquire it */
1559 1560
static void fuse_send_writepage(struct fuse_conn *fc, struct fuse_req *req,
				loff_t size)
M
Miklos Szeredi 已提交
1561 1562
__releases(fc->lock)
__acquires(fc->lock)
M
Miklos Szeredi 已提交
1563 1564 1565
{
	struct fuse_inode *fi = get_fuse_inode(req->inode);
	struct fuse_write_in *inarg = &req->misc.write.in;
1566
	__u64 data_size = req->num_pages * PAGE_CACHE_SIZE;
M
Miklos Szeredi 已提交
1567 1568 1569 1570

	if (!fc->connected)
		goto out_free;

1571 1572
	if (inarg->offset + data_size <= size) {
		inarg->size = data_size;
M
Miklos Szeredi 已提交
1573
	} else if (inarg->offset < size) {
1574
		inarg->size = size - inarg->offset;
M
Miklos Szeredi 已提交
1575 1576 1577
	} else {
		/* Got truncated off completely */
		goto out_free;
M
Miklos Szeredi 已提交
1578
	}
M
Miklos Szeredi 已提交
1579 1580 1581

	req->in.args[1].size = inarg->size;
	fi->writectr++;
1582
	fuse_request_send_background_locked(fc, req);
M
Miklos Szeredi 已提交
1583 1584 1585 1586 1587 1588
	return;

 out_free:
	fuse_writepage_finish(fc, req);
	spin_unlock(&fc->lock);
	fuse_writepage_free(fc, req);
1589
	fuse_put_request(fc, req);
M
Miklos Szeredi 已提交
1590
	spin_lock(&fc->lock);
M
Miklos Szeredi 已提交
1591 1592
}

M
Miklos Szeredi 已提交
1593 1594 1595 1596 1597 1598 1599
/*
 * 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)
M
Miklos Szeredi 已提交
1600 1601
__releases(fc->lock)
__acquires(fc->lock)
M
Miklos Szeredi 已提交
1602
{
M
Miklos Szeredi 已提交
1603 1604
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);
1605
	size_t crop = i_size_read(inode);
M
Miklos Szeredi 已提交
1606 1607 1608 1609 1610
	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);
1611
		fuse_send_writepage(fc, req, crop);
M
Miklos Szeredi 已提交
1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
	}
}

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);
1622
	while (req->misc.write.next) {
1623 1624
		struct fuse_conn *fc = get_fuse_conn(inode);
		struct fuse_write_in *inarg = &req->misc.write.in;
1625 1626 1627
		struct fuse_req *next = req->misc.write.next;
		req->misc.write.next = next->misc.write.next;
		next->misc.write.next = NULL;
1628
		next->ff = fuse_file_get(req->ff);
1629
		list_add(&next->writepages_entry, &fi->writepages);
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654

		/*
		 * 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);
1655
	}
M
Miklos Szeredi 已提交
1656 1657 1658 1659 1660 1661
	fi->writectr--;
	fuse_writepage_finish(fc, req);
	spin_unlock(&fc->lock);
	fuse_writepage_free(fc, req);
}

M
Miklos Szeredi 已提交
1662 1663
static struct fuse_file *__fuse_write_file_get(struct fuse_conn *fc,
					       struct fuse_inode *fi)
1664
{
M
Miklos Szeredi 已提交
1665
	struct fuse_file *ff = NULL;
1666 1667

	spin_lock(&fc->lock);
M
Miklos Szeredi 已提交
1668
	if (!list_empty(&fi->write_files)) {
M
Miklos Szeredi 已提交
1669 1670 1671 1672
		ff = list_entry(fi->write_files.next, struct fuse_file,
				write_entry);
		fuse_file_get(ff);
	}
1673 1674 1675 1676 1677
	spin_unlock(&fc->lock);

	return ff;
}

M
Miklos Szeredi 已提交
1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
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);
	err = fuse_flush_mtime(inode, ff);
	if (ff)
		fuse_file_put(ff, 0);

	return err;
}

M
Miklos Szeredi 已提交
1701 1702 1703 1704 1705 1706 1707 1708
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;
M
Miklos Szeredi 已提交
1709
	int error = -ENOMEM;
M
Miklos Szeredi 已提交
1710 1711 1712

	set_page_writeback(page);

1713
	req = fuse_request_alloc_nofs(1);
M
Miklos Szeredi 已提交
1714 1715 1716
	if (!req)
		goto err;

1717
	req->background = 1; /* writeback always goes to bg_queue */
M
Miklos Szeredi 已提交
1718 1719 1720 1721
	tmp_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
	if (!tmp_page)
		goto err_free;

M
Miklos Szeredi 已提交
1722
	error = -EIO;
1723
	req->ff = fuse_write_file_get(fc, fi);
M
Miklos Szeredi 已提交
1724 1725 1726
	if (!req->ff)
		goto err_free;

1727
	fuse_write_fill(req, req->ff, page_offset(page), 0);
M
Miklos Szeredi 已提交
1728 1729

	copy_highpage(tmp_page, page);
1730
	req->misc.write.in.write_flags |= FUSE_WRITE_CACHE;
1731
	req->misc.write.next = NULL;
1732
	req->in.argpages = 1;
M
Miklos Szeredi 已提交
1733 1734
	req->num_pages = 1;
	req->pages[0] = tmp_page;
1735
	req->page_descs[0].offset = 0;
1736
	req->page_descs[0].length = PAGE_SIZE;
M
Miklos Szeredi 已提交
1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
	req->end = fuse_writepage_end;
	req->inode = inode;

	inc_bdi_stat(mapping->backing_dev_info, BDI_WRITEBACK);
	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);

1749 1750
	end_page_writeback(page);

M
Miklos Szeredi 已提交
1751 1752 1753 1754 1755 1756
	return 0;

err_free:
	fuse_request_free(req);
err:
	end_page_writeback(page);
M
Miklos Szeredi 已提交
1757
	return error;
M
Miklos Szeredi 已提交
1758 1759 1760 1761 1762 1763
}

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

1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
	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;
	}

M
Miklos Szeredi 已提交
1776 1777 1778 1779 1780 1781
	err = fuse_writepage_locked(page);
	unlock_page(page);

	return err;
}

1782 1783 1784 1785
struct fuse_fill_wb_data {
	struct fuse_req *req;
	struct fuse_file *ff;
	struct inode *inode;
1786
	struct page **orig_pages;
1787 1788 1789 1790 1791 1792 1793 1794
};

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);
1795 1796
	int num_pages = req->num_pages;
	int i;
1797 1798 1799 1800 1801 1802

	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);
1803 1804 1805

	for (i = 0; i < num_pages; i++)
		end_page_writeback(data->orig_pages[i]);
1806 1807
}

1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
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);
		curr_index = old_req->misc.write.in.offset >> PAGE_CACHE_SHIFT;
		if (curr_index <= page->index &&
		    page->index < curr_index + old_req->num_pages) {
			found = true;
			break;
		}
	}
1831 1832
	if (!found) {
		list_add(&new_req->writepages_entry, &fi->writepages);
1833
		goto out_unlock;
1834
	}
1835

1836
	new_req->num_pages = 1;
1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847
	for (tmp = old_req; tmp != NULL; tmp = tmp->misc.write.next) {
		BUG_ON(tmp->inode != new_req->inode);
		curr_index = tmp->misc.write.in.offset >> PAGE_CACHE_SHIFT;
		if (tmp->num_pages == 1 &&
		    curr_index == page->index) {
			old_req = tmp;
		}
	}

	if (old_req->num_pages == 1 && (old_req->state == FUSE_REQ_INIT ||
					old_req->state == FUSE_REQ_PENDING)) {
1848 1849
		struct backing_dev_info *bdi = page->mapping->backing_dev_info;

1850 1851 1852
		copy_highpage(old_req->pages[0], page);
		spin_unlock(&fc->lock);

1853
		dec_bdi_stat(bdi, BDI_WRITEBACK);
1854
		dec_zone_page_state(page, NR_WRITEBACK_TEMP);
1855
		bdi_writeout_inc(bdi);
1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
		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;
}

1869 1870 1871 1872 1873 1874 1875 1876
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;
1877
	bool is_writeback;
1878 1879 1880 1881 1882 1883 1884 1885 1886
	int err;

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

1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
	/*
	 * 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 ||
	     (req->num_pages + 1) * PAGE_CACHE_SIZE > fc->max_write ||
	     data->orig_pages[req->num_pages - 1]->index + 1 != page->index)) {
		fuse_writepages_send(data);
		data->req = NULL;
1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
	}
	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;
1932
		req->misc.write.next = NULL;
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
		req->in.argpages = 1;
		req->background = 1;
		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;

	inc_bdi_stat(page->mapping->backing_dev_info, BDI_WRITEBACK);
	inc_zone_page_state(tmp_page, NR_WRITEBACK_TEMP);
1954 1955 1956 1957 1958 1959 1960

	err = 0;
	if (is_writeback && fuse_writepage_in_flight(req, page)) {
		end_page_writeback(page);
		data->req = NULL;
		goto out_unlock;
	}
1961
	data->orig_pages[req->num_pages] = page;
1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991

	/*
	 * 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;

1992 1993 1994 1995 1996 1997 1998
	err = -ENOMEM;
	data.orig_pages = kzalloc(sizeof(struct page *) *
				  FUSE_MAX_PAGES_PER_REQ,
				  GFP_NOFS);
	if (!data.orig_pages)
		goto out;

1999 2000 2001 2002 2003 2004 2005 2006 2007
	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);
2008 2009

	kfree(data.orig_pages);
2010 2011 2012 2013
out:
	return err;
}

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 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
/*
 * 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)
{
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	struct fuse_conn *fc = get_fuse_conn(file->f_dentry->d_inode);
	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);

	if (PageUptodate(page) || len == PAGE_CACHE_SIZE)
		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);
	if (fsize <= (pos & PAGE_CACHE_MASK)) {
		size_t off = pos & ~PAGE_CACHE_MASK;
		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);
	page_cache_release(page);
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 */
		size_t endoff = (pos + copied) & ~PAGE_CACHE_MASK;
		if (endoff)
			zero_user_segment(page, endoff, PAGE_CACHE_SIZE);
		SetPageUptodate(page);
	}

	fuse_write_update_size(inode, pos + copied);
	set_page_dirty(page);
	unlock_page(page);
	page_cache_release(page);

	return copied;
}

M
Miklos Szeredi 已提交
2085 2086 2087 2088 2089 2090 2091 2092 2093 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 2119 2120
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
 */
2121
static int fuse_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
M
Miklos Szeredi 已提交
2122
{
2123
	struct page *page = vmf->page;
M
Miklos Szeredi 已提交
2124 2125 2126 2127 2128 2129 2130 2131
	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;
	}
M
Miklos Szeredi 已提交
2132 2133

	fuse_wait_on_page_writeback(inode, page->index);
M
Miklos Szeredi 已提交
2134
	return VM_FAULT_LOCKED;
M
Miklos Szeredi 已提交
2135 2136
}

2137
static const struct vm_operations_struct fuse_file_vm_ops = {
M
Miklos Szeredi 已提交
2138 2139
	.close		= fuse_vma_close,
	.fault		= filemap_fault,
2140
	.map_pages	= filemap_map_pages,
M
Miklos Szeredi 已提交
2141
	.page_mkwrite	= fuse_page_mkwrite,
2142
	.remap_pages	= generic_file_remap_pages,
M
Miklos Szeredi 已提交
2143 2144 2145 2146
};

static int fuse_file_mmap(struct file *file, struct vm_area_struct *vma)
{
2147 2148 2149
	if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE))
		fuse_link_write_file(file);

M
Miklos Szeredi 已提交
2150 2151
	file_accessed(file);
	vma->vm_ops = &fuse_file_vm_ops;
M
Miklos Szeredi 已提交
2152 2153 2154
	return 0;
}

2155 2156 2157 2158 2159 2160
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;

2161 2162
	invalidate_inode_pages2(file->f_mapping);

2163 2164 2165
	return generic_file_mmap(file, vma);
}

2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
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;
}

static void fuse_lk_fill(struct fuse_req *req, struct file *file,
2192 2193
			 const struct file_lock *fl, int opcode, pid_t pid,
			 int flock)
2194
{
A
Al Viro 已提交
2195
	struct inode *inode = file_inode(file);
2196
	struct fuse_conn *fc = get_fuse_conn(inode);
2197 2198 2199 2200
	struct fuse_file *ff = file->private_data;
	struct fuse_lk_in *arg = &req->misc.lk_in;

	arg->fh = ff->fh;
2201
	arg->owner = fuse_lock_owner_id(fc, fl->fl_owner);
2202 2203 2204 2205
	arg->lk.start = fl->fl_start;
	arg->lk.end = fl->fl_end;
	arg->lk.type = fl->fl_type;
	arg->lk.pid = pid;
2206 2207
	if (flock)
		arg->lk_flags |= FUSE_LK_FLOCK;
2208 2209 2210 2211 2212 2213 2214 2215 2216
	req->in.h.opcode = opcode;
	req->in.h.nodeid = get_node_id(inode);
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(*arg);
	req->in.args[0].value = arg;
}

static int fuse_getlk(struct file *file, struct file_lock *fl)
{
A
Al Viro 已提交
2217
	struct inode *inode = file_inode(file);
2218 2219 2220 2221 2222
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_req *req;
	struct fuse_lk_out outarg;
	int err;

M
Maxim Patlasov 已提交
2223
	req = fuse_get_req_nopages(fc);
2224 2225 2226
	if (IS_ERR(req))
		return PTR_ERR(req);

2227
	fuse_lk_fill(req, file, fl, FUSE_GETLK, 0, 0);
2228 2229 2230
	req->out.numargs = 1;
	req->out.args[0].size = sizeof(outarg);
	req->out.args[0].value = &outarg;
2231
	fuse_request_send(fc, req);
2232 2233 2234 2235 2236 2237 2238 2239
	err = req->out.h.error;
	fuse_put_request(fc, req);
	if (!err)
		err = convert_fuse_file_lock(&outarg.lk, fl);

	return err;
}

2240
static int fuse_setlk(struct file *file, struct file_lock *fl, int flock)
2241
{
A
Al Viro 已提交
2242
	struct inode *inode = file_inode(file);
2243 2244 2245 2246 2247 2248
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_req *req;
	int opcode = (fl->fl_flags & FL_SLEEP) ? FUSE_SETLKW : FUSE_SETLK;
	pid_t pid = fl->fl_type != F_UNLCK ? current->tgid : 0;
	int err;

J
J. Bruce Fields 已提交
2249
	if (fl->fl_lmops && fl->fl_lmops->lm_grant) {
M
Miklos Szeredi 已提交
2250 2251 2252 2253
		/* NLM needs asynchronous locks, which we don't support yet */
		return -ENOLCK;
	}

2254 2255 2256 2257
	/* Unlock on close is handled by the flush method */
	if (fl->fl_flags & FL_CLOSE)
		return 0;

M
Maxim Patlasov 已提交
2258
	req = fuse_get_req_nopages(fc);
2259 2260 2261
	if (IS_ERR(req))
		return PTR_ERR(req);

2262
	fuse_lk_fill(req, file, fl, opcode, pid, flock);
2263
	fuse_request_send(fc, req);
2264
	err = req->out.h.error;
2265 2266 2267
	/* locking is restartable */
	if (err == -EINTR)
		err = -ERESTARTSYS;
2268 2269 2270 2271 2272 2273
	fuse_put_request(fc, req);
	return err;
}

static int fuse_file_lock(struct file *file, int cmd, struct file_lock *fl)
{
A
Al Viro 已提交
2274
	struct inode *inode = file_inode(file);
2275 2276 2277
	struct fuse_conn *fc = get_fuse_conn(inode);
	int err;

M
Miklos Szeredi 已提交
2278 2279 2280
	if (cmd == F_CANCELLK) {
		err = 0;
	} else if (cmd == F_GETLK) {
2281
		if (fc->no_lock) {
2282
			posix_test_lock(file, fl);
2283 2284 2285 2286 2287
			err = 0;
		} else
			err = fuse_getlk(file, fl);
	} else {
		if (fc->no_lock)
M
Miklos Szeredi 已提交
2288
			err = posix_lock_file(file, fl, NULL);
2289
		else
2290
			err = fuse_setlk(file, fl, 0);
2291 2292 2293 2294
	}
	return err;
}

2295 2296
static int fuse_file_flock(struct file *file, int cmd, struct file_lock *fl)
{
A
Al Viro 已提交
2297
	struct inode *inode = file_inode(file);
2298 2299 2300
	struct fuse_conn *fc = get_fuse_conn(inode);
	int err;

M
Miklos Szeredi 已提交
2301
	if (fc->no_flock) {
2302 2303
		err = flock_lock_file_wait(file, fl);
	} else {
M
Miklos Szeredi 已提交
2304 2305
		struct fuse_file *ff = file->private_data;

2306 2307
		/* emulate flock with POSIX locks */
		fl->fl_owner = (fl_owner_t) file;
M
Miklos Szeredi 已提交
2308
		ff->flock = true;
2309 2310 2311 2312 2313 2314
		err = fuse_setlk(file, fl, 1);
	}

	return err;
}

M
Miklos Szeredi 已提交
2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326
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);
	struct fuse_req *req;
	struct fuse_bmap_in inarg;
	struct fuse_bmap_out outarg;
	int err;

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

M
Maxim Patlasov 已提交
2327
	req = fuse_get_req_nopages(fc);
M
Miklos Szeredi 已提交
2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341
	if (IS_ERR(req))
		return 0;

	memset(&inarg, 0, sizeof(inarg));
	inarg.block = block;
	inarg.blocksize = inode->i_sb->s_blocksize;
	req->in.h.opcode = FUSE_BMAP;
	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;
	req->out.numargs = 1;
	req->out.args[0].size = sizeof(outarg);
	req->out.args[0].value = &outarg;
2342
	fuse_request_send(fc, req);
M
Miklos Szeredi 已提交
2343 2344 2345 2346 2347 2348 2349 2350
	err = req->out.h.error;
	fuse_put_request(fc, req);
	if (err == -ENOSYS)
		fc->no_bmap = 1;

	return err ? 0 : outarg.block;
}

2351
static loff_t fuse_file_llseek(struct file *file, loff_t offset, int whence)
M
Miklos Szeredi 已提交
2352 2353
{
	loff_t retval;
A
Al Viro 已提交
2354
	struct inode *inode = file_inode(file);
M
Miklos Szeredi 已提交
2355

2356
	/* No i_mutex protection necessary for SEEK_CUR and SEEK_SET */
2357 2358
	if (whence == SEEK_CUR || whence == SEEK_SET)
		return generic_file_llseek(file, offset, whence);
2359

2360 2361 2362
	mutex_lock(&inode->i_mutex);
	retval = fuse_update_attributes(inode, NULL, file, NULL);
	if (!retval)
2363
		retval = generic_file_llseek(file, offset, whence);
M
Miklos Szeredi 已提交
2364
	mutex_unlock(&inode->i_mutex);
2365

M
Miklos Szeredi 已提交
2366 2367 2368
	return retval;
}

T
Tejun Heo 已提交
2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382
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;

	iov_iter_init(&ii, iov, nr_segs, bytes, 0);

	while (iov_iter_count(&ii)) {
		struct page *page = pages[page_idx++];
		size_t todo = min_t(size_t, PAGE_SIZE, iov_iter_count(&ii));
D
Dan Carpenter 已提交
2383
		void *kaddr;
T
Tejun Heo 已提交
2384

D
Dan Carpenter 已提交
2385
		kaddr = kmap(page);
T
Tejun Heo 已提交
2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405

		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;
		}

2406
		kunmap(page);
T
Tejun Heo 已提交
2407 2408 2409 2410 2411
	}

	return 0;
}

2412 2413 2414 2415 2416 2417
/*
 * 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.
 */
M
Miklos Szeredi 已提交
2418 2419 2420
static int fuse_copy_ioctl_iovec_old(struct iovec *dst, void *src,
				     size_t transferred, unsigned count,
				     bool is_compat)
2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449
{
#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;
}

M
Miklos Szeredi 已提交
2450 2451 2452 2453 2454 2455
/* 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;

2456
	for (n = 0; n < count; n++, iov++) {
M
Miklos Szeredi 已提交
2457 2458 2459 2460 2461 2462 2463
		if (iov->iov_len > (size_t) max)
			return -ENOMEM;
		max -= iov->iov_len;
	}
	return 0;
}

M
Miklos Szeredi 已提交
2464 2465 2466 2467 2468 2469 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
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;
}


T
Tejun Heo 已提交
2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545
/*
 * 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.
 */
2546 2547
long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
		   unsigned int flags)
T
Tejun Heo 已提交
2548 2549
{
	struct fuse_file *ff = file->private_data;
2550
	struct fuse_conn *fc = ff->fc;
T
Tejun Heo 已提交
2551 2552 2553 2554 2555 2556 2557 2558 2559
	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;
M
Miklos Szeredi 已提交
2560
	struct iovec *iov_page = NULL;
T
Tejun Heo 已提交
2561 2562 2563 2564 2565
	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;

M
Miklos Szeredi 已提交
2566 2567 2568 2569 2570 2571 2572
#if BITS_PER_LONG == 32
	inarg.flags |= FUSE_IOCTL_32BIT;
#else
	if (flags & FUSE_IOCTL_COMPAT)
		inarg.flags |= FUSE_IOCTL_32BIT;
#endif

T
Tejun Heo 已提交
2573
	/* assume all the iovs returned by client always fits in a page */
M
Miklos Szeredi 已提交
2574
	BUILD_BUG_ON(sizeof(struct fuse_ioctl_iovec) * FUSE_IOCTL_MAX_IOV > PAGE_SIZE);
T
Tejun Heo 已提交
2575 2576

	err = -ENOMEM;
2577
	pages = kcalloc(FUSE_MAX_PAGES_PER_REQ, sizeof(pages[0]), GFP_KERNEL);
M
Miklos Szeredi 已提交
2578
	iov_page = (struct iovec *) __get_free_page(GFP_KERNEL);
T
Tejun Heo 已提交
2579 2580 2581 2582 2583 2584 2585 2586
	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)) {
M
Miklos Szeredi 已提交
2587
		struct iovec *iov = iov_page;
T
Tejun Heo 已提交
2588

M
Miklos Szeredi 已提交
2589
		iov->iov_base = (void __user *)arg;
T
Tejun Heo 已提交
2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624
		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++;
	}

M
Maxim Patlasov 已提交
2625
	req = fuse_get_req(fc, num_pages);
T
Tejun Heo 已提交
2626 2627 2628 2629 2630 2631 2632
	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;
2633
	fuse_page_descs_length_init(req, 0, req->num_pages);
T
Tejun Heo 已提交
2634 2635 2636

	/* okay, let's send it to the client */
	req->in.h.opcode = FUSE_IOCTL;
2637
	req->in.h.nodeid = ff->nodeid;
T
Tejun Heo 已提交
2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
	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;

2659
	fuse_request_send(fc, req);
T
Tejun Heo 已提交
2660 2661 2662 2663 2664 2665 2666 2667 2668
	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) {
M
Miklos Szeredi 已提交
2669
		void *vaddr;
T
Tejun Heo 已提交
2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688

		/* 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;

2689
		vaddr = kmap_atomic(pages[0]);
M
Miklos Szeredi 已提交
2690
		err = fuse_copy_ioctl_iovec(fc, iov_page, vaddr,
2691 2692
					    transferred, in_iovs + out_iovs,
					    (flags & FUSE_IOCTL_COMPAT) != 0);
2693
		kunmap_atomic(vaddr);
2694 2695
		if (err)
			goto out;
T
Tejun Heo 已提交
2696

M
Miklos Szeredi 已提交
2697
		in_iov = iov_page;
T
Tejun Heo 已提交
2698 2699
		out_iov = in_iov + in_iovs;

M
Miklos Szeredi 已提交
2700 2701 2702 2703 2704 2705 2706 2707
		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;

T
Tejun Heo 已提交
2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718
		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);
M
Miklos Szeredi 已提交
2719
	free_page((unsigned long) iov_page);
T
Tejun Heo 已提交
2720 2721 2722 2723 2724 2725
	while (num_pages)
		__free_page(pages[--num_pages]);
	kfree(pages);

	return err ? err : outarg.result;
}
2726
EXPORT_SYMBOL_GPL(fuse_do_ioctl);
T
Tejun Heo 已提交
2727

2728 2729
long fuse_ioctl_common(struct file *file, unsigned int cmd,
		       unsigned long arg, unsigned int flags)
2730
{
A
Al Viro 已提交
2731
	struct inode *inode = file_inode(file);
2732 2733
	struct fuse_conn *fc = get_fuse_conn(inode);

2734
	if (!fuse_allow_current_process(fc))
2735 2736 2737 2738 2739 2740 2741 2742
		return -EACCES;

	if (is_bad_inode(inode))
		return -EIO;

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

T
Tejun Heo 已提交
2743 2744 2745
static long fuse_file_ioctl(struct file *file, unsigned int cmd,
			    unsigned long arg)
{
2746
	return fuse_ioctl_common(file, cmd, arg, 0);
T
Tejun Heo 已提交
2747 2748 2749 2750 2751
}

static long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
				   unsigned long arg)
{
2752
	return fuse_ioctl_common(file, cmd, arg, FUSE_IOCTL_COMPAT);
T
Tejun Heo 已提交
2753 2754
}

T
Tejun Heo 已提交
2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795
/*
 * 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)) {
2796
		struct rb_node **link, *uninitialized_var(parent);
T
Tejun Heo 已提交
2797 2798 2799 2800 2801 2802 2803 2804 2805

		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);
}

2806
unsigned fuse_file_poll(struct file *file, poll_table *wait)
T
Tejun Heo 已提交
2807 2808
{
	struct fuse_file *ff = file->private_data;
2809
	struct fuse_conn *fc = ff->fc;
T
Tejun Heo 已提交
2810 2811 2812 2813 2814 2815 2816 2817 2818
	struct fuse_poll_in inarg = { .fh = ff->fh, .kh = ff->kh };
	struct fuse_poll_out outarg;
	struct fuse_req *req;
	int err;

	if (fc->no_poll)
		return DEFAULT_POLLMASK;

	poll_wait(file, &ff->poll_wait, wait);
E
Enke Chen 已提交
2819
	inarg.events = (__u32)poll_requested_events(wait);
T
Tejun Heo 已提交
2820 2821 2822 2823 2824 2825 2826 2827 2828 2829

	/*
	 * 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);
	}

M
Maxim Patlasov 已提交
2830
	req = fuse_get_req_nopages(fc);
T
Tejun Heo 已提交
2831
	if (IS_ERR(req))
2832
		return POLLERR;
T
Tejun Heo 已提交
2833 2834

	req->in.h.opcode = FUSE_POLL;
2835
	req->in.h.nodeid = ff->nodeid;
T
Tejun Heo 已提交
2836 2837 2838 2839 2840 2841
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(inarg);
	req->in.args[0].value = &inarg;
	req->out.numargs = 1;
	req->out.args[0].size = sizeof(outarg);
	req->out.args[0].value = &outarg;
2842
	fuse_request_send(fc, req);
T
Tejun Heo 已提交
2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853
	err = req->out.h.error;
	fuse_put_request(fc, req);

	if (!err)
		return outarg.revents;
	if (err == -ENOSYS) {
		fc->no_poll = 1;
		return DEFAULT_POLLMASK;
	}
	return POLLERR;
}
2854
EXPORT_SYMBOL_GPL(fuse_file_poll);
T
Tejun Heo 已提交
2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879

/*
 * 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;
}

2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893
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);
}

2894 2895 2896 2897 2898
static inline loff_t fuse_round_up(loff_t off)
{
	return round_up(off, FUSE_MAX_PAGES_PER_REQ << PAGE_SHIFT);
}

A
Anand Avati 已提交
2899 2900 2901 2902 2903
static ssize_t
fuse_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
			loff_t offset, unsigned long nr_segs)
{
	ssize_t ret = 0;
2904 2905
	struct file *file = iocb->ki_filp;
	struct fuse_file *ff = file->private_data;
2906
	bool async_dio = ff->fc->async_dio;
A
Anand Avati 已提交
2907
	loff_t pos = 0;
2908 2909 2910
	struct inode *inode;
	loff_t i_size;
	size_t count = iov_length(iov, nr_segs);
2911
	struct fuse_io_priv *io;
A
Anand Avati 已提交
2912 2913

	pos = offset;
2914 2915
	inode = file->f_mapping->host;
	i_size = i_size_read(inode);
A
Anand Avati 已提交
2916

2917 2918 2919
	if ((rw == READ) && (offset > i_size))
		return 0;

2920
	/* optimization for short read */
2921
	if (async_dio && rw != WRITE && offset + count > i_size) {
2922 2923
		if (offset >= i_size)
			return 0;
2924
		count = min_t(loff_t, count, fuse_round_up(i_size - offset));
2925 2926
	}

2927
	io = kmalloc(sizeof(struct fuse_io_priv), GFP_KERNEL);
2928 2929
	if (!io)
		return -ENOMEM;
2930 2931 2932 2933 2934 2935 2936
	spin_lock_init(&io->lock);
	io->reqs = 1;
	io->bytes = -1;
	io->size = 0;
	io->offset = offset;
	io->write = (rw == WRITE);
	io->err = 0;
2937
	io->file = file;
2938 2939
	/*
	 * By default, we want to optimize all I/Os with async request
2940
	 * submission to the client filesystem if supported.
2941
	 */
2942
	io->async = async_dio;
2943 2944 2945 2946 2947 2948 2949
	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.
	 */
2950
	if (!is_sync_kiocb(iocb) && (offset + count > i_size) && rw == WRITE)
2951
		io->async = false;
A
Anand Avati 已提交
2952

2953
	if (rw == WRITE)
2954
		ret = __fuse_direct_write(io, iov, nr_segs, &pos);
2955
	else
2956
		ret = __fuse_direct_read(io, iov, nr_segs, &pos, count);
2957

2958 2959 2960 2961
	if (io->async) {
		fuse_aio_complete(io, ret < 0 ? ret : 0, -1);

		/* we have a non-extending, async request, so return */
2962
		if (!is_sync_kiocb(iocb))
2963 2964 2965 2966 2967 2968 2969
			return -EIOCBQUEUED;

		ret = wait_on_sync_kiocb(iocb);
	} else {
		kfree(io);
	}

2970 2971 2972 2973 2974 2975
	if (rw == WRITE) {
		if (ret > 0)
			fuse_write_update_size(inode, pos);
		else if (ret < 0 && offset + count > i_size)
			fuse_do_truncate(file);
	}
A
Anand Avati 已提交
2976 2977 2978 2979

	return ret;
}

2980 2981
static long fuse_file_fallocate(struct file *file, int mode, loff_t offset,
				loff_t length)
A
Anatol Pomozov 已提交
2982 2983
{
	struct fuse_file *ff = file->private_data;
2984
	struct inode *inode = file->f_inode;
2985
	struct fuse_inode *fi = get_fuse_inode(inode);
A
Anatol Pomozov 已提交
2986 2987 2988 2989 2990 2991 2992 2993 2994
	struct fuse_conn *fc = ff->fc;
	struct fuse_req *req;
	struct fuse_fallocate_in inarg = {
		.fh = ff->fh,
		.offset = offset,
		.length = length,
		.mode = mode
	};
	int err;
2995 2996
	bool lock_inode = !(mode & FALLOC_FL_KEEP_SIZE) ||
			   (mode & FALLOC_FL_PUNCH_HOLE);
A
Anatol Pomozov 已提交
2997

M
Miklos Szeredi 已提交
2998 2999 3000
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

3001 3002 3003
	if (fc->no_fallocate)
		return -EOPNOTSUPP;

3004
	if (lock_inode) {
3005
		mutex_lock(&inode->i_mutex);
3006 3007 3008 3009 3010 3011 3012 3013 3014
		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);
		}
3015 3016
	}

3017 3018 3019
	if (!(mode & FALLOC_FL_KEEP_SIZE))
		set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);

M
Maxim Patlasov 已提交
3020
	req = fuse_get_req_nopages(fc);
3021 3022 3023 3024
	if (IS_ERR(req)) {
		err = PTR_ERR(req);
		goto out;
	}
A
Anatol Pomozov 已提交
3025 3026 3027 3028 3029 3030 3031 3032

	req->in.h.opcode = FUSE_FALLOCATE;
	req->in.h.nodeid = ff->nodeid;
	req->in.numargs = 1;
	req->in.args[0].size = sizeof(inarg);
	req->in.args[0].value = &inarg;
	fuse_request_send(fc, req);
	err = req->out.h.error;
3033 3034 3035 3036
	if (err == -ENOSYS) {
		fc->no_fallocate = 1;
		err = -EOPNOTSUPP;
	}
A
Anatol Pomozov 已提交
3037 3038
	fuse_put_request(fc, req);

3039 3040 3041 3042
	if (err)
		goto out;

	/* we could have extended the file */
M
Maxim Patlasov 已提交
3043 3044 3045
	if (!(mode & FALLOC_FL_KEEP_SIZE)) {
		bool changed = fuse_write_update_size(inode, offset + length);

3046 3047
		if (changed && fc->writeback_cache)
			file_update_time(file);
M
Maxim Patlasov 已提交
3048
	}
3049 3050 3051 3052 3053 3054

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

	fuse_invalidate_attr(inode);

3055
out:
3056 3057 3058
	if (!(mode & FALLOC_FL_KEEP_SIZE))
		clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);

3059
	if (lock_inode)
3060 3061
		mutex_unlock(&inode->i_mutex);

A
Anatol Pomozov 已提交
3062 3063 3064
	return err;
}

3065
static const struct file_operations fuse_file_operations = {
M
Miklos Szeredi 已提交
3066
	.llseek		= fuse_file_llseek,
3067
	.read		= do_sync_read,
M
Miklos Szeredi 已提交
3068
	.aio_read	= fuse_file_aio_read,
3069
	.write		= do_sync_write,
N
Nick Piggin 已提交
3070
	.aio_write	= fuse_file_aio_write,
M
Miklos Szeredi 已提交
3071 3072 3073 3074 3075
	.mmap		= fuse_file_mmap,
	.open		= fuse_open,
	.flush		= fuse_flush,
	.release	= fuse_release,
	.fsync		= fuse_fsync,
3076
	.lock		= fuse_file_lock,
3077
	.flock		= fuse_file_flock,
3078
	.splice_read	= generic_file_splice_read,
T
Tejun Heo 已提交
3079 3080
	.unlocked_ioctl	= fuse_file_ioctl,
	.compat_ioctl	= fuse_file_compat_ioctl,
T
Tejun Heo 已提交
3081
	.poll		= fuse_file_poll,
A
Anatol Pomozov 已提交
3082
	.fallocate	= fuse_file_fallocate,
M
Miklos Szeredi 已提交
3083 3084
};

3085
static const struct file_operations fuse_direct_io_file_operations = {
M
Miklos Szeredi 已提交
3086
	.llseek		= fuse_file_llseek,
M
Miklos Szeredi 已提交
3087 3088
	.read		= fuse_direct_read,
	.write		= fuse_direct_write,
3089
	.mmap		= fuse_direct_mmap,
M
Miklos Szeredi 已提交
3090 3091 3092 3093
	.open		= fuse_open,
	.flush		= fuse_flush,
	.release	= fuse_release,
	.fsync		= fuse_fsync,
3094
	.lock		= fuse_file_lock,
3095
	.flock		= fuse_file_flock,
T
Tejun Heo 已提交
3096 3097
	.unlocked_ioctl	= fuse_file_ioctl,
	.compat_ioctl	= fuse_file_compat_ioctl,
T
Tejun Heo 已提交
3098
	.poll		= fuse_file_poll,
A
Anatol Pomozov 已提交
3099
	.fallocate	= fuse_file_fallocate,
3100
	/* no splice_read */
M
Miklos Szeredi 已提交
3101 3102
};

3103
static const struct address_space_operations fuse_file_aops  = {
M
Miklos Szeredi 已提交
3104
	.readpage	= fuse_readpage,
M
Miklos Szeredi 已提交
3105
	.writepage	= fuse_writepage,
3106
	.writepages	= fuse_writepages,
M
Miklos Szeredi 已提交
3107
	.launder_page	= fuse_launder_page,
3108
	.readpages	= fuse_readpages,
M
Miklos Szeredi 已提交
3109
	.set_page_dirty	= __set_page_dirty_nobuffers,
M
Miklos Szeredi 已提交
3110
	.bmap		= fuse_bmap,
A
Anand Avati 已提交
3111
	.direct_IO	= fuse_direct_IO,
3112 3113
	.write_begin	= fuse_write_begin,
	.write_end	= fuse_write_end,
M
Miklos Szeredi 已提交
3114 3115 3116 3117
};

void fuse_init_file_inode(struct inode *inode)
{
3118 3119
	inode->i_fop = &fuse_file_operations;
	inode->i_data.a_ops = &fuse_file_aops;
M
Miklos Szeredi 已提交
3120
}