file.c 93.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4
/*
 *   fs/cifs/file.c
 *
 *   vfs operations that deal with files
S
Steve French 已提交
5
 *
6
 *   Copyright (C) International Business Machines  Corp., 2002,2010
L
Linus Torvalds 已提交
7
 *   Author(s): Steve French (sfrench@us.ibm.com)
J
[CIFS]  
Jeremy Allison 已提交
8
 *              Jeremy Allison (jra@samba.org)
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 *
 *   This library is free software; you can redistribute it and/or modify
 *   it under the terms of the GNU Lesser General Public License as published
 *   by the Free Software Foundation; either version 2.1 of the License, or
 *   (at your option) any later version.
 *
 *   This library is distributed in the hope that it will be useful,
 *   but WITHOUT ANY WARRANTY; without even the implied warranty of
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
 *   the GNU Lesser General Public License for more details.
 *
 *   You should have received a copy of the GNU Lesser General Public License
 *   along with this library; if not, write to the Free Software
 *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */
#include <linux/fs.h>
25
#include <linux/backing-dev.h>
L
Linus Torvalds 已提交
26 27 28 29
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/pagemap.h>
#include <linux/pagevec.h>
30
#include <linux/writeback.h>
31
#include <linux/task_io_accounting_ops.h>
32
#include <linux/delay.h>
33
#include <linux/mount.h>
34
#include <linux/slab.h>
35
#include <linux/swap.h>
L
Linus Torvalds 已提交
36 37 38 39 40 41 42 43
#include <asm/div64.h>
#include "cifsfs.h"
#include "cifspdu.h"
#include "cifsglob.h"
#include "cifsproto.h"
#include "cifs_unicode.h"
#include "cifs_debug.h"
#include "cifs_fs_sb.h"
44
#include "fscache.h"
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52 53 54 55 56 57 58

static inline int cifs_convert_flags(unsigned int flags)
{
	if ((flags & O_ACCMODE) == O_RDONLY)
		return GENERIC_READ;
	else if ((flags & O_ACCMODE) == O_WRONLY)
		return GENERIC_WRITE;
	else if ((flags & O_ACCMODE) == O_RDWR) {
		/* GENERIC_ALL is too much permission to request
		   can cause unnecessary access denied on create */
		/* return GENERIC_ALL; */
		return (GENERIC_READ | GENERIC_WRITE);
	}

59 60 61
	return (READ_CONTROL | FILE_WRITE_ATTRIBUTES | FILE_READ_ATTRIBUTES |
		FILE_WRITE_EA | FILE_APPEND_DATA | FILE_WRITE_DATA |
		FILE_READ_DATA);
62
}
63

64
static u32 cifs_posix_convert_flags(unsigned int flags)
65
{
66
	u32 posix_flags = 0;
67

68
	if ((flags & O_ACCMODE) == O_RDONLY)
69
		posix_flags = SMB_O_RDONLY;
70
	else if ((flags & O_ACCMODE) == O_WRONLY)
71 72 73 74 75 76 77 78 79 80 81
		posix_flags = SMB_O_WRONLY;
	else if ((flags & O_ACCMODE) == O_RDWR)
		posix_flags = SMB_O_RDWR;

	if (flags & O_CREAT)
		posix_flags |= SMB_O_CREAT;
	if (flags & O_EXCL)
		posix_flags |= SMB_O_EXCL;
	if (flags & O_TRUNC)
		posix_flags |= SMB_O_TRUNC;
	/* be safe and imply O_SYNC for O_DSYNC */
82
	if (flags & O_DSYNC)
83
		posix_flags |= SMB_O_SYNC;
84
	if (flags & O_DIRECTORY)
85
		posix_flags |= SMB_O_DIRECTORY;
86
	if (flags & O_NOFOLLOW)
87
		posix_flags |= SMB_O_NOFOLLOW;
88
	if (flags & O_DIRECT)
89
		posix_flags |= SMB_O_DIRECT;
90 91

	return posix_flags;
L
Linus Torvalds 已提交
92 93 94 95 96 97 98 99 100 101
}

static inline int cifs_get_disposition(unsigned int flags)
{
	if ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
		return FILE_CREATE;
	else if ((flags & (O_CREAT | O_TRUNC)) == (O_CREAT | O_TRUNC))
		return FILE_OVERWRITE_IF;
	else if ((flags & O_CREAT) == O_CREAT)
		return FILE_OPEN_IF;
102 103
	else if ((flags & O_TRUNC) == O_TRUNC)
		return FILE_OVERWRITE;
L
Linus Torvalds 已提交
104 105 106 107
	else
		return FILE_OPEN;
}

108 109
int cifs_posix_open(char *full_path, struct inode **pinode,
			struct super_block *sb, int mode, unsigned int f_flags,
110
			__u32 *poplock, __u16 *pnetfid, unsigned int xid)
111 112 113 114 115 116 117
{
	int rc;
	FILE_UNIX_BASIC_INFO *presp_data;
	__u32 posix_flags = 0;
	struct cifs_sb_info *cifs_sb = CIFS_SB(sb);
	struct cifs_fattr fattr;
	struct tcon_link *tlink;
118
	struct cifs_tcon *tcon;
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169

	cFYI(1, "posix open %s", full_path);

	presp_data = kzalloc(sizeof(FILE_UNIX_BASIC_INFO), GFP_KERNEL);
	if (presp_data == NULL)
		return -ENOMEM;

	tlink = cifs_sb_tlink(cifs_sb);
	if (IS_ERR(tlink)) {
		rc = PTR_ERR(tlink);
		goto posix_open_ret;
	}

	tcon = tlink_tcon(tlink);
	mode &= ~current_umask();

	posix_flags = cifs_posix_convert_flags(f_flags);
	rc = CIFSPOSIXCreate(xid, tcon, posix_flags, mode, pnetfid, presp_data,
			     poplock, full_path, cifs_sb->local_nls,
			     cifs_sb->mnt_cifs_flags &
					CIFS_MOUNT_MAP_SPECIAL_CHR);
	cifs_put_tlink(tlink);

	if (rc)
		goto posix_open_ret;

	if (presp_data->Type == cpu_to_le32(-1))
		goto posix_open_ret; /* open ok, caller does qpathinfo */

	if (!pinode)
		goto posix_open_ret; /* caller does not need info */

	cifs_unix_basic_to_fattr(&fattr, presp_data, cifs_sb);

	/* get new inode and set it up */
	if (*pinode == NULL) {
		cifs_fill_uniqueid(sb, &fattr);
		*pinode = cifs_iget(sb, &fattr);
		if (!*pinode) {
			rc = -ENOMEM;
			goto posix_open_ret;
		}
	} else {
		cifs_fattr_to_inode(*pinode, &fattr);
	}

posix_open_ret:
	kfree(presp_data);
	return rc;
}

170 171
static int
cifs_nt_open(char *full_path, struct inode *inode, struct cifs_sb_info *cifs_sb,
172 173
	     struct cifs_tcon *tcon, unsigned int f_flags, __u32 *oplock,
	     struct cifs_fid *fid, unsigned int xid)
174 175
{
	int rc;
176
	int desired_access;
177
	int disposition;
178
	int create_options = CREATE_NOT_DIR;
179
	FILE_ALL_INFO *buf;
P
Pavel Shilovsky 已提交
180
	struct TCP_Server_Info *server = tcon->ses->server;
181

P
Pavel Shilovsky 已提交
182
	if (!server->ops->open)
183 184 185
		return -ENOSYS;

	desired_access = cifs_convert_flags(f_flags);
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218

/*********************************************************************
 *  open flag mapping table:
 *
 *	POSIX Flag            CIFS Disposition
 *	----------            ----------------
 *	O_CREAT               FILE_OPEN_IF
 *	O_CREAT | O_EXCL      FILE_CREATE
 *	O_CREAT | O_TRUNC     FILE_OVERWRITE_IF
 *	O_TRUNC               FILE_OVERWRITE
 *	none of the above     FILE_OPEN
 *
 *	Note that there is not a direct match between disposition
 *	FILE_SUPERSEDE (ie create whether or not file exists although
 *	O_CREAT | O_TRUNC is similar but truncates the existing
 *	file rather than creating a new file as FILE_SUPERSEDE does
 *	(which uses the attributes / metadata passed in on open call)
 *?
 *?  O_SYNC is a reasonable match to CIFS writethrough flag
 *?  and the read write flags match reasonably.  O_LARGEFILE
 *?  is irrelevant because largefile support is always used
 *?  by this client. Flags O_APPEND, O_DIRECT, O_DIRECTORY,
 *	 O_FASYNC, O_NOFOLLOW, O_NONBLOCK need further investigation
 *********************************************************************/

	disposition = cifs_get_disposition(f_flags);

	/* BB pass O_SYNC flag through on file attributes .. BB */

	buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
	if (!buf)
		return -ENOMEM;

219 220 221
	if (backup_cred(cifs_sb))
		create_options |= CREATE_OPEN_BACKUP_INTENT;

P
Pavel Shilovsky 已提交
222 223 224
	rc = server->ops->open(xid, tcon, full_path, disposition,
			       desired_access, create_options, fid, oplock, buf,
			       cifs_sb);
225 226 227 228 229 230 231 232 233

	if (rc)
		goto out;

	if (tcon->unix_ext)
		rc = cifs_get_inode_info_unix(&inode, full_path, inode->i_sb,
					      xid);
	else
		rc = cifs_get_inode_info(&inode, full_path, buf, inode->i_sb,
234
					 xid, &fid->netfid);
235 236 237 238 239 240

out:
	kfree(buf);
	return rc;
}

241
struct cifsFileInfo *
242
cifs_new_fileinfo(struct cifs_fid *fid, struct file *file,
243 244 245 246
		  struct tcon_link *tlink, __u32 oplock)
{
	struct dentry *dentry = file->f_path.dentry;
	struct inode *inode = dentry->d_inode;
247 248
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	struct cifsFileInfo *cfile;
249
	struct cifs_fid_locks *fdlocks;
250
	struct cifs_tcon *tcon = tlink_tcon(tlink);
251 252 253 254 255

	cfile = kzalloc(sizeof(struct cifsFileInfo), GFP_KERNEL);
	if (cfile == NULL)
		return cfile;

256 257 258 259 260 261 262 263 264
	fdlocks = kzalloc(sizeof(struct cifs_fid_locks), GFP_KERNEL);
	if (!fdlocks) {
		kfree(cfile);
		return NULL;
	}

	INIT_LIST_HEAD(&fdlocks->locks);
	fdlocks->cfile = cfile;
	cfile->llist = fdlocks;
265
	down_write(&cinode->lock_sem);
266
	list_add(&fdlocks->llist, &cinode->llist);
267
	up_write(&cinode->lock_sem);
268

269 270 271 272 273 274 275 276
	cfile->count = 1;
	cfile->pid = current->tgid;
	cfile->uid = current_fsuid();
	cfile->dentry = dget(dentry);
	cfile->f_flags = file->f_flags;
	cfile->invalidHandle = false;
	cfile->tlink = cifs_get_tlink(tlink);
	INIT_WORK(&cfile->oplock_break, cifs_oplock_break);
277
	mutex_init(&cfile->fh_mutex);
278

279
	spin_lock(&cifs_file_list_lock);
280 281 282 283 284 285 286
	if (fid->pending_open->oplock != CIFS_OPLOCK_NO_CHANGE)
		oplock = fid->pending_open->oplock;
	list_del(&fid->pending_open->olist);

	tlink_tcon(tlink)->ses->server->ops->set_fid(cfile, fid, oplock);

	list_add(&cfile->tlist, &tcon->openFileList);
287 288
	/* if readable file instance put first in list*/
	if (file->f_mode & FMODE_READ)
289
		list_add(&cfile->flist, &cinode->openFileList);
290
	else
291
		list_add_tail(&cfile->flist, &cinode->openFileList);
292
	spin_unlock(&cifs_file_list_lock);
293

294 295
	file->private_data = cfile;
	return cfile;
296 297
}

298 299 300 301 302 303 304 305 306
struct cifsFileInfo *
cifsFileInfo_get(struct cifsFileInfo *cifs_file)
{
	spin_lock(&cifs_file_list_lock);
	cifsFileInfo_get_locked(cifs_file);
	spin_unlock(&cifs_file_list_lock);
	return cifs_file;
}

307 308
/*
 * Release a reference on the file private data. This may involve closing
309 310
 * the filehandle out on the server. Must be called without holding
 * cifs_file_list_lock.
311
 */
312 313
void cifsFileInfo_put(struct cifsFileInfo *cifs_file)
{
P
Pavel Shilovsky 已提交
314
	struct inode *inode = cifs_file->dentry->d_inode;
315
	struct cifs_tcon *tcon = tlink_tcon(cifs_file->tlink);
316
	struct TCP_Server_Info *server = tcon->ses->server;
P
Pavel Shilovsky 已提交
317
	struct cifsInodeInfo *cifsi = CIFS_I(inode);
318
	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
319
	struct cifsLockInfo *li, *tmp;
320 321
	struct cifs_fid fid;
	struct cifs_pending_open open;
322 323

	spin_lock(&cifs_file_list_lock);
324
	if (--cifs_file->count > 0) {
325 326 327 328
		spin_unlock(&cifs_file_list_lock);
		return;
	}

329 330 331 332 333 334
	if (server->ops->get_lease_key)
		server->ops->get_lease_key(inode, &fid);

	/* store open in pending opens to make sure we don't miss lease break */
	cifs_add_pending_open_locked(&fid, cifs_file->tlink, &open);

335 336 337 338 339 340 341
	/* remove it from the lists */
	list_del(&cifs_file->flist);
	list_del(&cifs_file->tlist);

	if (list_empty(&cifsi->openFileList)) {
		cFYI(1, "closing last open instance for inode %p",
			cifs_file->dentry->d_inode);
342 343 344 345 346
		/*
		 * In strict cache mode we need invalidate mapping on the last
		 * close  because it may cause a error when we open this file
		 * again and get at least level II oplock.
		 */
347 348
		if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO)
			CIFS_I(inode)->invalid_mapping = true;
349
		cifs_set_oplock_level(cifsi, 0);
350 351 352
	}
	spin_unlock(&cifs_file_list_lock);

353 354
	cancel_work_sync(&cifs_file->oplock_break);

355
	if (!tcon->need_reconnect && !cifs_file->invalidHandle) {
356
		struct TCP_Server_Info *server = tcon->ses->server;
357
		unsigned int xid;
358

359
		xid = get_xid();
360
		if (server->ops->close)
361 362
			server->ops->close(xid, tcon, &cifs_file->fid);
		_free_xid(xid);
363 364
	}

365 366
	cifs_del_pending_open(&open);

367 368
	/*
	 * Delete any outstanding lock records. We'll lose them when the file
369 370
	 * is closed anyway.
	 */
371
	down_write(&cifsi->lock_sem);
372
	list_for_each_entry_safe(li, tmp, &cifs_file->llist->locks, llist) {
373
		list_del(&li->llist);
374
		cifs_del_lock_waiters(li);
375
		kfree(li);
376
	}
377 378
	list_del(&cifs_file->llist->llist);
	kfree(cifs_file->llist);
379
	up_write(&cifsi->lock_sem);
380 381 382 383

	cifs_put_tlink(cifs_file->tlink);
	dput(cifs_file->dentry);
	kfree(cifs_file);
384 385
}

L
Linus Torvalds 已提交
386
int cifs_open(struct inode *inode, struct file *file)
387

L
Linus Torvalds 已提交
388 389
{
	int rc = -EACCES;
390
	unsigned int xid;
391
	__u32 oplock;
L
Linus Torvalds 已提交
392
	struct cifs_sb_info *cifs_sb;
P
Pavel Shilovsky 已提交
393
	struct TCP_Server_Info *server;
394
	struct cifs_tcon *tcon;
395
	struct tcon_link *tlink;
396
	struct cifsFileInfo *cfile = NULL;
L
Linus Torvalds 已提交
397
	char *full_path = NULL;
P
Pavel Shilovsky 已提交
398
	bool posix_open_ok = false;
399
	struct cifs_fid fid;
400
	struct cifs_pending_open open;
L
Linus Torvalds 已提交
401

402
	xid = get_xid();
L
Linus Torvalds 已提交
403 404

	cifs_sb = CIFS_SB(inode->i_sb);
405 406
	tlink = cifs_sb_tlink(cifs_sb);
	if (IS_ERR(tlink)) {
407
		free_xid(xid);
408 409 410
		return PTR_ERR(tlink);
	}
	tcon = tlink_tcon(tlink);
P
Pavel Shilovsky 已提交
411
	server = tcon->ses->server;
L
Linus Torvalds 已提交
412

413
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
414
	if (full_path == NULL) {
415
		rc = -ENOMEM;
416
		goto out;
L
Linus Torvalds 已提交
417 418
	}

419 420
	cFYI(1, "inode = 0x%p file flags are 0x%x for %s",
		 inode, file->f_flags, full_path);
421

422
	if (server->oplocks)
423 424 425 426
		oplock = REQ_OPLOCK;
	else
		oplock = 0;

427
	if (!tcon->broken_posix_open && tcon->unix_ext &&
428 429
	    cap_unix(tcon->ses) && (CIFS_UNIX_POSIX_PATH_OPS_CAP &
				le64_to_cpu(tcon->fsUnixInfo.Capability))) {
430
		/* can not refresh inode info since size could be stale */
431
		rc = cifs_posix_open(full_path, &inode, inode->i_sb,
432
				cifs_sb->mnt_file_mode /* ignored */,
433
				file->f_flags, &oplock, &fid.netfid, xid);
434
		if (rc == 0) {
435
			cFYI(1, "posix open succeeded");
P
Pavel Shilovsky 已提交
436
			posix_open_ok = true;
437 438
		} else if ((rc == -EINVAL) || (rc == -EOPNOTSUPP)) {
			if (tcon->ses->serverNOS)
439
				cERROR(1, "server %s of type %s returned"
440 441 442 443
					   " unexpected error on SMB posix open"
					   ", disabling posix open support."
					   " Check if server update available.",
					   tcon->ses->serverName,
444
					   tcon->ses->serverNOS);
445
			tcon->broken_posix_open = true;
446 447 448
		} else if ((rc != -EIO) && (rc != -EREMOTE) &&
			 (rc != -EOPNOTSUPP)) /* path not found or net err */
			goto out;
449 450 451 452
		/*
		 * Else fallthrough to retry open the old way on network i/o
		 * or DFS errors.
		 */
453 454
	}

455 456 457 458 459
	if (server->ops->get_lease_key)
		server->ops->get_lease_key(inode, &fid);

	cifs_add_pending_open(&fid, tlink, &open);

P
Pavel Shilovsky 已提交
460
	if (!posix_open_ok) {
P
Pavel Shilovsky 已提交
461 462 463
		if (server->ops->get_lease_key)
			server->ops->get_lease_key(inode, &fid);

P
Pavel Shilovsky 已提交
464
		rc = cifs_nt_open(full_path, inode, cifs_sb, tcon,
465
				  file->f_flags, &oplock, &fid, xid);
466 467
		if (rc) {
			cifs_del_pending_open(&open);
P
Pavel Shilovsky 已提交
468
			goto out;
469
		}
P
Pavel Shilovsky 已提交
470
	}
471

472 473
	cfile = cifs_new_fileinfo(&fid, file, tlink, oplock);
	if (cfile == NULL) {
P
Pavel Shilovsky 已提交
474 475
		if (server->ops->close)
			server->ops->close(xid, tcon, &fid);
476
		cifs_del_pending_open(&open);
L
Linus Torvalds 已提交
477 478 479 480
		rc = -ENOMEM;
		goto out;
	}

481 482
	cifs_fscache_set_inode_cookie(inode, file);

P
Pavel Shilovsky 已提交
483
	if ((oplock & CIFS_CREATE_ACTION) && !posix_open_ok && tcon->unix_ext) {
484 485 486 487
		/*
		 * Time to set mode which we can not set earlier due to
		 * problems creating new read-only files.
		 */
P
Pavel Shilovsky 已提交
488 489 490 491 492 493 494 495 496
		struct cifs_unix_set_info_args args = {
			.mode	= inode->i_mode,
			.uid	= NO_CHANGE_64,
			.gid	= NO_CHANGE_64,
			.ctime	= NO_CHANGE_64,
			.atime	= NO_CHANGE_64,
			.mtime	= NO_CHANGE_64,
			.device	= 0,
		};
497 498
		CIFSSMBUnixSetFileInfo(xid, tcon, &args, fid.netfid,
				       cfile->pid);
L
Linus Torvalds 已提交
499 500 501 502
	}

out:
	kfree(full_path);
503
	free_xid(xid);
504
	cifs_put_tlink(tlink);
L
Linus Torvalds 已提交
505 506 507
	return rc;
}

P
Pavel Shilovsky 已提交
508 509
static int cifs_push_posix_locks(struct cifsFileInfo *cfile);

510 511
/*
 * Try to reacquire byte range locks that were released when session
P
Pavel Shilovsky 已提交
512
 * to server was lost.
513
 */
P
Pavel Shilovsky 已提交
514 515
static int
cifs_relock_file(struct cifsFileInfo *cfile)
L
Linus Torvalds 已提交
516
{
P
Pavel Shilovsky 已提交
517 518 519
	struct cifs_sb_info *cifs_sb = CIFS_SB(cfile->dentry->d_sb);
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
L
Linus Torvalds 已提交
520 521
	int rc = 0;

P
Pavel Shilovsky 已提交
522 523 524 525 526 527 528 529 530 531 532 533 534 535
	/* we are going to update can_cache_brlcks here - need a write access */
	down_write(&cinode->lock_sem);
	if (cinode->can_cache_brlcks) {
		/* can cache locks - no need to push them */
		up_write(&cinode->lock_sem);
		return rc;
	}

	if (cap_unix(tcon->ses) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
		rc = cifs_push_posix_locks(cfile);
	else
		rc = tcon->ses->server->ops->push_mand_locks(cfile);
L
Linus Torvalds 已提交
536

P
Pavel Shilovsky 已提交
537
	up_write(&cinode->lock_sem);
L
Linus Torvalds 已提交
538 539 540
	return rc;
}

541 542
static int
cifs_reopen_file(struct cifsFileInfo *cfile, bool can_flush)
L
Linus Torvalds 已提交
543 544
{
	int rc = -EACCES;
545
	unsigned int xid;
546
	__u32 oplock;
L
Linus Torvalds 已提交
547
	struct cifs_sb_info *cifs_sb;
548
	struct cifs_tcon *tcon;
549 550
	struct TCP_Server_Info *server;
	struct cifsInodeInfo *cinode;
S
Steve French 已提交
551
	struct inode *inode;
L
Linus Torvalds 已提交
552
	char *full_path = NULL;
553
	int desired_access;
L
Linus Torvalds 已提交
554
	int disposition = FILE_OPEN;
555
	int create_options = CREATE_NOT_DIR;
556
	struct cifs_fid fid;
L
Linus Torvalds 已提交
557

558
	xid = get_xid();
559 560 561
	mutex_lock(&cfile->fh_mutex);
	if (!cfile->invalidHandle) {
		mutex_unlock(&cfile->fh_mutex);
562
		rc = 0;
563
		free_xid(xid);
564
		return rc;
L
Linus Torvalds 已提交
565 566
	}

567
	inode = cfile->dentry->d_inode;
L
Linus Torvalds 已提交
568
	cifs_sb = CIFS_SB(inode->i_sb);
569 570 571 572 573 574 575 576 577 578
	tcon = tlink_tcon(cfile->tlink);
	server = tcon->ses->server;

	/*
	 * Can not grab rename sem here because various ops, including those
	 * that already have the rename sem can end up causing writepage to get
	 * called and if the server was down that means we end up here, and we
	 * can never tell if the caller already has the rename_sem.
	 */
	full_path = build_path_from_dentry(cfile->dentry);
L
Linus Torvalds 已提交
579
	if (full_path == NULL) {
580
		rc = -ENOMEM;
581
		mutex_unlock(&cfile->fh_mutex);
582
		free_xid(xid);
583
		return rc;
L
Linus Torvalds 已提交
584 585
	}

586 587
	cFYI(1, "inode = 0x%p file flags 0x%x for %s", inode, cfile->f_flags,
	     full_path);
L
Linus Torvalds 已提交
588

589
	if (tcon->ses->server->oplocks)
L
Linus Torvalds 已提交
590 591
		oplock = REQ_OPLOCK;
	else
592
		oplock = 0;
L
Linus Torvalds 已提交
593

594
	if (tcon->unix_ext && cap_unix(tcon->ses) &&
595
	    (CIFS_UNIX_POSIX_PATH_OPS_CAP &
596
				le64_to_cpu(tcon->fsUnixInfo.Capability))) {
597 598 599 600
		/*
		 * O_CREAT, O_EXCL and O_TRUNC already had their effect on the
		 * original open. Must mask them off for a reopen.
		 */
601
		unsigned int oflags = cfile->f_flags &
J
Jeff Layton 已提交
602
						~(O_CREAT | O_EXCL | O_TRUNC);
603

604
		rc = cifs_posix_open(full_path, NULL, inode->i_sb,
605 606
				     cifs_sb->mnt_file_mode /* ignored */,
				     oflags, &oplock, &fid.netfid, xid);
607
		if (rc == 0) {
608
			cFYI(1, "posix reopen succeeded");
609 610
			goto reopen_success;
		}
611 612 613 614
		/*
		 * fallthrough to retry open the old way on errors, especially
		 * in the reconnect path it is important to retry hard
		 */
615 616
	}

617
	desired_access = cifs_convert_flags(cfile->f_flags);
618

619 620 621
	if (backup_cred(cifs_sb))
		create_options |= CREATE_OPEN_BACKUP_INTENT;

P
Pavel Shilovsky 已提交
622 623 624
	if (server->ops->get_lease_key)
		server->ops->get_lease_key(inode, &fid);

625 626 627 628 629 630 631 632 633 634
	/*
	 * Can not refresh inode by passing in file_info buf to be returned by
	 * CIFSSMBOpen and then calling get_inode_info with returned buf since
	 * file might have write behind data that needs to be flushed and server
	 * version of file size can be stale. If we knew for sure that inode was
	 * not dirty locally we could do this.
	 */
	rc = server->ops->open(xid, tcon, full_path, disposition,
			       desired_access, create_options, &fid, &oplock,
			       NULL, cifs_sb);
L
Linus Torvalds 已提交
635
	if (rc) {
636 637
		mutex_unlock(&cfile->fh_mutex);
		cFYI(1, "cifs_reopen returned 0x%x", rc);
638
		cFYI(1, "oplock: %d", oplock);
J
Jeff Layton 已提交
639 640 641
		goto reopen_error_exit;
	}

642
reopen_success:
643 644 645
	cfile->invalidHandle = false;
	mutex_unlock(&cfile->fh_mutex);
	cinode = CIFS_I(inode);
J
Jeff Layton 已提交
646 647 648

	if (can_flush) {
		rc = filemap_write_and_wait(inode->i_mapping);
649
		mapping_set_error(inode->i_mapping, rc);
J
Jeff Layton 已提交
650 651

		if (tcon->unix_ext)
652 653
			rc = cifs_get_inode_info_unix(&inode, full_path,
						      inode->i_sb, xid);
J
Jeff Layton 已提交
654
		else
655 656 657 658 659 660 661 662 663 664 665 666
			rc = cifs_get_inode_info(&inode, full_path, NULL,
						 inode->i_sb, xid, NULL);
	}
	/*
	 * Else we are writing out data to server already and could deadlock if
	 * we tried to flush data, and since we do not know if we have data that
	 * would invalidate the current end of file on the server we can not go
	 * to the server to get the new inode info.
	 */

	server->ops->set_fid(cfile, &fid, oplock);
	cifs_relock_file(cfile);
J
Jeff Layton 已提交
667 668

reopen_error_exit:
L
Linus Torvalds 已提交
669
	kfree(full_path);
670
	free_xid(xid);
L
Linus Torvalds 已提交
671 672 673 674 675
	return rc;
}

int cifs_close(struct inode *inode, struct file *file)
{
676 677 678 679
	if (file->private_data != NULL) {
		cifsFileInfo_put(file->private_data);
		file->private_data = NULL;
	}
J
[CIFS]  
Jeremy Allison 已提交
680

681 682
	/* return code from the ->release op is always ignored */
	return 0;
L
Linus Torvalds 已提交
683 684 685 686 687
}

int cifs_closedir(struct inode *inode, struct file *file)
{
	int rc = 0;
688
	unsigned int xid;
689
	struct cifsFileInfo *cfile = file->private_data;
690 691 692
	struct cifs_tcon *tcon;
	struct TCP_Server_Info *server;
	char *buf;
L
Linus Torvalds 已提交
693

694
	cFYI(1, "Closedir inode = 0x%p", inode);
L
Linus Torvalds 已提交
695

696 697 698
	if (cfile == NULL)
		return rc;

699
	xid = get_xid();
700 701
	tcon = tlink_tcon(cfile->tlink);
	server = tcon->ses->server;
L
Linus Torvalds 已提交
702

703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
	cFYI(1, "Freeing private data in close dir");
	spin_lock(&cifs_file_list_lock);
	if (!cfile->srch_inf.endOfSearch && !cfile->invalidHandle) {
		cfile->invalidHandle = true;
		spin_unlock(&cifs_file_list_lock);
		if (server->ops->close_dir)
			rc = server->ops->close_dir(xid, tcon, &cfile->fid);
		else
			rc = -ENOSYS;
		cFYI(1, "Closing uncompleted readdir with rc %d", rc);
		/* not much we can do if it fails anyway, ignore rc */
		rc = 0;
	} else
		spin_unlock(&cifs_file_list_lock);

	buf = cfile->srch_inf.ntwrk_buf_start;
	if (buf) {
		cFYI(1, "closedir free smb buf in srch struct");
		cfile->srch_inf.ntwrk_buf_start = NULL;
		if (cfile->srch_inf.smallBuf)
			cifs_small_buf_release(buf);
		else
			cifs_buf_release(buf);
L
Linus Torvalds 已提交
726
	}
727 728 729 730

	cifs_put_tlink(cfile->tlink);
	kfree(file->private_data);
	file->private_data = NULL;
L
Linus Torvalds 已提交
731
	/* BB can we lock the filestruct while this is going on? */
732
	free_xid(xid);
L
Linus Torvalds 已提交
733 734 735
	return rc;
}

736
static struct cifsLockInfo *
737
cifs_lock_init(__u64 offset, __u64 length, __u8 type)
J
[CIFS]  
Jeremy Allison 已提交
738
{
739
	struct cifsLockInfo *lock =
S
Steve French 已提交
740
		kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
741 742 743 744 745 746 747 748 749
	if (!lock)
		return lock;
	lock->offset = offset;
	lock->length = length;
	lock->type = type;
	lock->pid = current->tgid;
	INIT_LIST_HEAD(&lock->blist);
	init_waitqueue_head(&lock->block_q);
	return lock;
750 751
}

752
void
753 754 755 756 757 758 759 760 761
cifs_del_lock_waiters(struct cifsLockInfo *lock)
{
	struct cifsLockInfo *li, *tmp;
	list_for_each_entry_safe(li, tmp, &lock->blist, blist) {
		list_del_init(&li->blist);
		wake_up(&li->block_q);
	}
}

762 763 764 765 766
#define CIFS_LOCK_OP	0
#define CIFS_READ_OP	1
#define CIFS_WRITE_OP	2

/* @rw_check : 0 - no op, 1 - read, 2 - write */
767
static bool
768 769
cifs_find_fid_lock_conflict(struct cifs_fid_locks *fdlocks, __u64 offset,
			    __u64 length, __u8 type, struct cifsFileInfo *cfile,
770
			    struct cifsLockInfo **conf_lock, int rw_check)
771
{
772
	struct cifsLockInfo *li;
773
	struct cifsFileInfo *cur_cfile = fdlocks->cfile;
774
	struct TCP_Server_Info *server = tlink_tcon(cfile->tlink)->ses->server;
775

776
	list_for_each_entry(li, &fdlocks->locks, llist) {
777 778 779
		if (offset + length <= li->offset ||
		    offset >= li->offset + li->length)
			continue;
780 781 782 783 784 785 786
		if (rw_check != CIFS_LOCK_OP && current->tgid == li->pid &&
		    server->ops->compare_fids(cfile, cur_cfile)) {
			/* shared lock prevents write op through the same fid */
			if (!(li->type & server->vals->shared_lock_type) ||
			    rw_check != CIFS_WRITE_OP)
				continue;
		}
787 788 789
		if ((type & server->vals->shared_lock_type) &&
		    ((server->ops->compare_fids(cfile, cur_cfile) &&
		     current->tgid == li->pid) || type == li->type))
790
			continue;
791 792
		if (conf_lock)
			*conf_lock = li;
793
		return true;
794 795 796 797
	}
	return false;
}

798
bool
799
cifs_find_lock_conflict(struct cifsFileInfo *cfile, __u64 offset, __u64 length,
800
			__u8 type, struct cifsLockInfo **conf_lock,
801
			int rw_check)
802
{
803
	bool rc = false;
804
	struct cifs_fid_locks *cur;
805
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
806

807 808
	list_for_each_entry(cur, &cinode->llist, llist) {
		rc = cifs_find_fid_lock_conflict(cur, offset, length, type,
809
						 cfile, conf_lock, rw_check);
810 811 812 813 814
		if (rc)
			break;
	}

	return rc;
815 816
}

817 818 819 820 821 822 823
/*
 * Check if there is another lock that prevents us to set the lock (mandatory
 * style). If such a lock exists, update the flock structure with its
 * properties. Otherwise, set the flock type to F_UNLCK if we can cache brlocks
 * or leave it the same if we can't. Returns 0 if we don't need to request to
 * the server or 1 otherwise.
 */
824
static int
825 826
cifs_lock_test(struct cifsFileInfo *cfile, __u64 offset, __u64 length,
	       __u8 type, struct file_lock *flock)
827 828 829
{
	int rc = 0;
	struct cifsLockInfo *conf_lock;
830
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
831
	struct TCP_Server_Info *server = tlink_tcon(cfile->tlink)->ses->server;
832 833
	bool exist;

834
	down_read(&cinode->lock_sem);
835

836
	exist = cifs_find_lock_conflict(cfile, offset, length, type,
837
					&conf_lock, CIFS_LOCK_OP);
838 839 840 841
	if (exist) {
		flock->fl_start = conf_lock->offset;
		flock->fl_end = conf_lock->offset + conf_lock->length - 1;
		flock->fl_pid = conf_lock->pid;
842
		if (conf_lock->type & server->vals->shared_lock_type)
843 844 845 846 847 848 849 850
			flock->fl_type = F_RDLCK;
		else
			flock->fl_type = F_WRLCK;
	} else if (!cinode->can_cache_brlcks)
		rc = 1;
	else
		flock->fl_type = F_UNLCK;

851
	up_read(&cinode->lock_sem);
852 853 854
	return rc;
}

855
static void
856
cifs_lock_add(struct cifsFileInfo *cfile, struct cifsLockInfo *lock)
857
{
858
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
859
	down_write(&cinode->lock_sem);
860
	list_add_tail(&lock->llist, &cfile->llist->locks);
861
	up_write(&cinode->lock_sem);
J
[CIFS]  
Jeremy Allison 已提交
862 863
}

864 865 866 867 868 869
/*
 * Set the byte-range lock (mandatory style). Returns:
 * 1) 0, if we set the lock and don't need to request to the server;
 * 2) 1, if no locks prevent us but we need to request to the server;
 * 3) -EACCESS, if there is a lock that prevents us and wait is false.
 */
870
static int
871
cifs_lock_add_if(struct cifsFileInfo *cfile, struct cifsLockInfo *lock,
872
		 bool wait)
873
{
874
	struct cifsLockInfo *conf_lock;
875
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
876 877 878 879 880
	bool exist;
	int rc = 0;

try_again:
	exist = false;
881
	down_write(&cinode->lock_sem);
882

883
	exist = cifs_find_lock_conflict(cfile, lock->offset, lock->length,
884
					lock->type, &conf_lock, CIFS_LOCK_OP);
885
	if (!exist && cinode->can_cache_brlcks) {
886
		list_add_tail(&lock->llist, &cfile->llist->locks);
887
		up_write(&cinode->lock_sem);
888 889 890 891 892 893 894 895 896
		return rc;
	}

	if (!exist)
		rc = 1;
	else if (!wait)
		rc = -EACCES;
	else {
		list_add_tail(&lock->blist, &conf_lock->blist);
897
		up_write(&cinode->lock_sem);
898 899 900 901 902
		rc = wait_event_interruptible(lock->block_q,
					(lock->blist.prev == &lock->blist) &&
					(lock->blist.next == &lock->blist));
		if (!rc)
			goto try_again;
903
		down_write(&cinode->lock_sem);
904
		list_del_init(&lock->blist);
905 906
	}

907
	up_write(&cinode->lock_sem);
908 909 910
	return rc;
}

911 912 913 914 915 916 917
/*
 * Check if there is another lock that prevents us to set the lock (posix
 * style). If such a lock exists, update the flock structure with its
 * properties. Otherwise, set the flock type to F_UNLCK if we can cache brlocks
 * or leave it the same if we can't. Returns 0 if we don't need to request to
 * the server or 1 otherwise.
 */
918
static int
919 920 921 922 923 924
cifs_posix_lock_test(struct file *file, struct file_lock *flock)
{
	int rc = 0;
	struct cifsInodeInfo *cinode = CIFS_I(file->f_path.dentry->d_inode);
	unsigned char saved_type = flock->fl_type;

925 926 927
	if ((flock->fl_flags & FL_POSIX) == 0)
		return 1;

928
	down_read(&cinode->lock_sem);
929 930 931 932 933 934 935
	posix_test_lock(file, flock);

	if (flock->fl_type == F_UNLCK && !cinode->can_cache_brlcks) {
		flock->fl_type = saved_type;
		rc = 1;
	}

936
	up_read(&cinode->lock_sem);
937 938 939
	return rc;
}

940 941 942 943 944 945
/*
 * Set the byte-range lock (posix style). Returns:
 * 1) 0, if we set the lock and don't need to request to the server;
 * 2) 1, if we need to request to the server;
 * 3) <0, if the error occurs while setting the lock.
 */
946 947 948 949
static int
cifs_posix_lock_set(struct file *file, struct file_lock *flock)
{
	struct cifsInodeInfo *cinode = CIFS_I(file->f_path.dentry->d_inode);
950 951 952 953
	int rc = 1;

	if ((flock->fl_flags & FL_POSIX) == 0)
		return rc;
954

955
try_again:
956
	down_write(&cinode->lock_sem);
957
	if (!cinode->can_cache_brlcks) {
958
		up_write(&cinode->lock_sem);
959
		return rc;
960
	}
961 962

	rc = posix_lock_file(file, flock, NULL);
963
	up_write(&cinode->lock_sem);
964 965 966 967 968 969
	if (rc == FILE_LOCK_DEFERRED) {
		rc = wait_event_interruptible(flock->fl_wait, !flock->fl_next);
		if (!rc)
			goto try_again;
		locks_delete_block(flock);
	}
970
	return rc;
971 972
}

973
int
974
cifs_push_mandatory_locks(struct cifsFileInfo *cfile)
975
{
976 977
	unsigned int xid;
	int rc = 0, stored_rc;
978 979
	struct cifsLockInfo *li, *tmp;
	struct cifs_tcon *tcon;
980
	unsigned int num, max_num, max_buf;
981 982 983 984
	LOCKING_ANDX_RANGE *buf, *cur;
	int types[] = {LOCKING_ANDX_LARGE_FILES,
		       LOCKING_ANDX_SHARED_LOCK | LOCKING_ANDX_LARGE_FILES};
	int i;
985

986
	xid = get_xid();
987 988
	tcon = tlink_tcon(cfile->tlink);

989 990 991 992 993 994
	/*
	 * Accessing maxBuf is racy with cifs_reconnect - need to store value
	 * and check it for zero before using.
	 */
	max_buf = tcon->ses->server->maxBuf;
	if (!max_buf) {
995
		free_xid(xid);
996 997 998 999 1000
		return -EINVAL;
	}

	max_num = (max_buf - sizeof(struct smb_hdr)) /
						sizeof(LOCKING_ANDX_RANGE);
1001 1002
	buf = kzalloc(max_num * sizeof(LOCKING_ANDX_RANGE), GFP_KERNEL);
	if (!buf) {
1003
		free_xid(xid);
1004
		return -ENOMEM;
1005 1006 1007 1008 1009
	}

	for (i = 0; i < 2; i++) {
		cur = buf;
		num = 0;
1010
		list_for_each_entry_safe(li, tmp, &cfile->llist->locks, llist) {
1011 1012 1013 1014 1015 1016 1017 1018
			if (li->type != types[i])
				continue;
			cur->Pid = cpu_to_le16(li->pid);
			cur->LengthLow = cpu_to_le32((u32)li->length);
			cur->LengthHigh = cpu_to_le32((u32)(li->length>>32));
			cur->OffsetLow = cpu_to_le32((u32)li->offset);
			cur->OffsetHigh = cpu_to_le32((u32)(li->offset>>32));
			if (++num == max_num) {
1019 1020
				stored_rc = cifs_lockv(xid, tcon,
						       cfile->fid.netfid,
1021 1022
						       (__u8)li->type, 0, num,
						       buf);
1023 1024 1025 1026 1027 1028 1029 1030 1031
				if (stored_rc)
					rc = stored_rc;
				cur = buf;
				num = 0;
			} else
				cur++;
		}

		if (num) {
1032
			stored_rc = cifs_lockv(xid, tcon, cfile->fid.netfid,
1033
					       (__u8)types[i], 0, num, buf);
1034 1035 1036
			if (stored_rc)
				rc = stored_rc;
		}
1037 1038
	}

1039
	kfree(buf);
1040
	free_xid(xid);
1041 1042 1043
	return rc;
}

1044 1045 1046 1047 1048
/* copied from fs/locks.c with a name change */
#define cifs_for_each_lock(inode, lockp) \
	for (lockp = &inode->i_flock; *lockp != NULL; \
	     lockp = &(*lockp)->fl_next)

1049 1050 1051 1052 1053 1054 1055 1056 1057
struct lock_to_push {
	struct list_head llist;
	__u64 offset;
	__u64 length;
	__u32 pid;
	__u16 netfid;
	__u8 type;
};

1058
static int
1059
cifs_push_posix_locks(struct cifsFileInfo *cfile)
1060 1061 1062
{
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
	struct file_lock *flock, **before;
1063
	unsigned int count = 0, i = 0;
1064
	int rc = 0, xid, type;
1065 1066
	struct list_head locks_to_send, *el;
	struct lock_to_push *lck, *tmp;
1067 1068
	__u64 length;

1069
	xid = get_xid();
1070

1071 1072 1073 1074 1075 1076 1077
	lock_flocks();
	cifs_for_each_lock(cfile->dentry->d_inode, before) {
		if ((*before)->fl_flags & FL_POSIX)
			count++;
	}
	unlock_flocks();

1078 1079
	INIT_LIST_HEAD(&locks_to_send);

1080
	/*
1081
	 * Allocating count locks is enough because no FL_POSIX locks can be
1082
	 * added to the list while we are holding cinode->lock_sem that
1083
	 * protects locking operations of this inode.
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
	 */
	for (; i < count; i++) {
		lck = kmalloc(sizeof(struct lock_to_push), GFP_KERNEL);
		if (!lck) {
			rc = -ENOMEM;
			goto err_out;
		}
		list_add_tail(&lck->llist, &locks_to_send);
	}

	el = locks_to_send.next;
1095 1096
	lock_flocks();
	cifs_for_each_lock(cfile->dentry->d_inode, before) {
1097 1098 1099
		flock = *before;
		if ((flock->fl_flags & FL_POSIX) == 0)
			continue;
1100
		if (el == &locks_to_send) {
1101 1102 1103 1104
			/*
			 * The list ended. We don't have enough allocated
			 * structures - something is really wrong.
			 */
1105 1106 1107
			cERROR(1, "Can't push all brlocks!");
			break;
		}
1108 1109 1110 1111 1112
		length = 1 + flock->fl_end - flock->fl_start;
		if (flock->fl_type == F_RDLCK || flock->fl_type == F_SHLCK)
			type = CIFS_RDLCK;
		else
			type = CIFS_WRLCK;
1113
		lck = list_entry(el, struct lock_to_push, llist);
1114
		lck->pid = flock->fl_pid;
1115
		lck->netfid = cfile->fid.netfid;
1116 1117 1118 1119
		lck->length = length;
		lck->type = type;
		lck->offset = flock->fl_start;
		el = el->next;
1120 1121 1122 1123 1124 1125 1126
	}
	unlock_flocks();

	list_for_each_entry_safe(lck, tmp, &locks_to_send, llist) {
		int stored_rc;

		stored_rc = CIFSSMBPosixLock(xid, tcon, lck->netfid, lck->pid,
1127
					     lck->offset, lck->length, NULL,
1128 1129 1130 1131 1132 1133 1134
					     lck->type, 0);
		if (stored_rc)
			rc = stored_rc;
		list_del(&lck->llist);
		kfree(lck);
	}

1135
out:
1136
	free_xid(xid);
1137
	return rc;
1138 1139 1140 1141 1142 1143
err_out:
	list_for_each_entry_safe(lck, tmp, &locks_to_send, llist) {
		list_del(&lck->llist);
		kfree(lck);
	}
	goto out;
1144 1145
}

1146
static int
1147
cifs_push_locks(struct cifsFileInfo *cfile)
1148
{
1149
	struct cifs_sb_info *cifs_sb = CIFS_SB(cfile->dentry->d_sb);
1150
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
1151
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
1152 1153 1154 1155 1156 1157 1158 1159
	int rc = 0;

	/* we are going to update can_cache_brlcks here - need a write access */
	down_write(&cinode->lock_sem);
	if (!cinode->can_cache_brlcks) {
		up_write(&cinode->lock_sem);
		return rc;
	}
1160

1161
	if (cap_unix(tcon->ses) &&
1162 1163
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
1164 1165 1166
		rc = cifs_push_posix_locks(cfile);
	else
		rc = tcon->ses->server->ops->push_mand_locks(cfile);
1167

1168 1169 1170
	cinode->can_cache_brlcks = false;
	up_write(&cinode->lock_sem);
	return rc;
1171 1172
}

1173
static void
1174
cifs_read_flock(struct file_lock *flock, __u32 *type, int *lock, int *unlock,
1175
		bool *wait_flag, struct TCP_Server_Info *server)
L
Linus Torvalds 已提交
1176
{
1177
	if (flock->fl_flags & FL_POSIX)
1178
		cFYI(1, "Posix");
1179
	if (flock->fl_flags & FL_FLOCK)
1180
		cFYI(1, "Flock");
1181
	if (flock->fl_flags & FL_SLEEP) {
1182
		cFYI(1, "Blocking lock");
1183
		*wait_flag = true;
L
Linus Torvalds 已提交
1184
	}
1185
	if (flock->fl_flags & FL_ACCESS)
1186
		cFYI(1, "Process suspended by mandatory locking - "
1187 1188
			"not implemented yet");
	if (flock->fl_flags & FL_LEASE)
1189
		cFYI(1, "Lease on file - not implemented yet");
1190
	if (flock->fl_flags &
1191 1192
	    (~(FL_POSIX | FL_FLOCK | FL_SLEEP |
	       FL_ACCESS | FL_LEASE | FL_CLOSE)))
1193
		cFYI(1, "Unknown lock flags 0x%x", flock->fl_flags);
L
Linus Torvalds 已提交
1194

1195
	*type = server->vals->large_lock_type;
1196
	if (flock->fl_type == F_WRLCK) {
1197
		cFYI(1, "F_WRLCK ");
1198
		*type |= server->vals->exclusive_lock_type;
1199 1200
		*lock = 1;
	} else if (flock->fl_type == F_UNLCK) {
1201
		cFYI(1, "F_UNLCK");
1202
		*type |= server->vals->unlock_lock_type;
1203 1204 1205
		*unlock = 1;
		/* Check if unlock includes more than one lock range */
	} else if (flock->fl_type == F_RDLCK) {
1206
		cFYI(1, "F_RDLCK");
1207
		*type |= server->vals->shared_lock_type;
1208 1209
		*lock = 1;
	} else if (flock->fl_type == F_EXLCK) {
1210
		cFYI(1, "F_EXLCK");
1211
		*type |= server->vals->exclusive_lock_type;
1212 1213
		*lock = 1;
	} else if (flock->fl_type == F_SHLCK) {
1214
		cFYI(1, "F_SHLCK");
1215
		*type |= server->vals->shared_lock_type;
1216
		*lock = 1;
L
Linus Torvalds 已提交
1217
	} else
1218
		cFYI(1, "Unknown type of lock");
1219
}
L
Linus Torvalds 已提交
1220

1221
static int
1222
cifs_getlk(struct file *file, struct file_lock *flock, __u32 type,
1223
	   bool wait_flag, bool posix_lck, unsigned int xid)
1224 1225 1226
{
	int rc = 0;
	__u64 length = 1 + flock->fl_end - flock->fl_start;
1227 1228
	struct cifsFileInfo *cfile = (struct cifsFileInfo *)file->private_data;
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
1229
	struct TCP_Server_Info *server = tcon->ses->server;
1230
	__u16 netfid = cfile->fid.netfid;
1231

1232 1233
	if (posix_lck) {
		int posix_lock_type;
1234 1235 1236 1237 1238

		rc = cifs_posix_lock_test(file, flock);
		if (!rc)
			return rc;

1239
		if (type & server->vals->shared_lock_type)
1240 1241 1242
			posix_lock_type = CIFS_RDLCK;
		else
			posix_lock_type = CIFS_WRLCK;
1243
		rc = CIFSSMBPosixLock(xid, tcon, netfid, current->tgid,
1244
				      flock->fl_start, length, flock,
1245
				      posix_lock_type, wait_flag);
1246 1247
		return rc;
	}
L
Linus Torvalds 已提交
1248

1249
	rc = cifs_lock_test(cfile, flock->fl_start, length, type, flock);
1250 1251 1252
	if (!rc)
		return rc;

1253
	/* BB we could chain these into one lock request BB */
1254 1255
	rc = server->ops->mand_lock(xid, cfile, flock->fl_start, length, type,
				    1, 0, false);
1256
	if (rc == 0) {
1257 1258
		rc = server->ops->mand_lock(xid, cfile, flock->fl_start, length,
					    type, 0, 1, false);
1259 1260 1261
		flock->fl_type = F_UNLCK;
		if (rc != 0)
			cERROR(1, "Error unlocking previously locked "
1262
				  "range %d during test of lock", rc);
1263
		return 0;
L
Linus Torvalds 已提交
1264
	}
J
[CIFS]  
Jeremy Allison 已提交
1265

1266
	if (type & server->vals->shared_lock_type) {
1267
		flock->fl_type = F_WRLCK;
1268
		return 0;
J
[CIFS]  
Jeremy Allison 已提交
1269 1270
	}

1271 1272 1273 1274 1275
	type &= ~server->vals->exclusive_lock_type;

	rc = server->ops->mand_lock(xid, cfile, flock->fl_start, length,
				    type | server->vals->shared_lock_type,
				    1, 0, false);
1276
	if (rc == 0) {
1277 1278
		rc = server->ops->mand_lock(xid, cfile, flock->fl_start, length,
			type | server->vals->shared_lock_type, 0, 1, false);
1279 1280 1281 1282 1283 1284 1285
		flock->fl_type = F_RDLCK;
		if (rc != 0)
			cERROR(1, "Error unlocking previously locked "
				  "range %d during test of lock", rc);
	} else
		flock->fl_type = F_WRLCK;

1286
	return 0;
1287 1288
}

1289
void
1290 1291 1292 1293 1294 1295 1296
cifs_move_llist(struct list_head *source, struct list_head *dest)
{
	struct list_head *li, *tmp;
	list_for_each_safe(li, tmp, source)
		list_move(li, dest);
}

1297
void
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
cifs_free_llist(struct list_head *llist)
{
	struct cifsLockInfo *li, *tmp;
	list_for_each_entry_safe(li, tmp, llist, llist) {
		cifs_del_lock_waiters(li);
		list_del(&li->llist);
		kfree(li);
	}
}

1308
int
1309 1310
cifs_unlock_range(struct cifsFileInfo *cfile, struct file_lock *flock,
		  unsigned int xid)
1311 1312 1313 1314 1315
{
	int rc = 0, stored_rc;
	int types[] = {LOCKING_ANDX_LARGE_FILES,
		       LOCKING_ANDX_SHARED_LOCK | LOCKING_ANDX_LARGE_FILES};
	unsigned int i;
1316
	unsigned int max_num, num, max_buf;
1317 1318 1319 1320 1321 1322 1323 1324 1325
	LOCKING_ANDX_RANGE *buf, *cur;
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
	struct cifsInodeInfo *cinode = CIFS_I(cfile->dentry->d_inode);
	struct cifsLockInfo *li, *tmp;
	__u64 length = 1 + flock->fl_end - flock->fl_start;
	struct list_head tmp_llist;

	INIT_LIST_HEAD(&tmp_llist);

1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
	/*
	 * Accessing maxBuf is racy with cifs_reconnect - need to store value
	 * and check it for zero before using.
	 */
	max_buf = tcon->ses->server->maxBuf;
	if (!max_buf)
		return -EINVAL;

	max_num = (max_buf - sizeof(struct smb_hdr)) /
						sizeof(LOCKING_ANDX_RANGE);
1336 1337 1338 1339
	buf = kzalloc(max_num * sizeof(LOCKING_ANDX_RANGE), GFP_KERNEL);
	if (!buf)
		return -ENOMEM;

1340
	down_write(&cinode->lock_sem);
1341 1342 1343
	for (i = 0; i < 2; i++) {
		cur = buf;
		num = 0;
1344
		list_for_each_entry_safe(li, tmp, &cfile->llist->locks, llist) {
1345 1346 1347 1348 1349 1350 1351 1352
			if (flock->fl_start > li->offset ||
			    (flock->fl_start + length) <
			    (li->offset + li->length))
				continue;
			if (current->tgid != li->pid)
				continue;
			if (types[i] != li->type)
				continue;
1353
			if (cinode->can_cache_brlcks) {
1354 1355
				/*
				 * We can cache brlock requests - simply remove
1356
				 * a lock from the file's list.
1357 1358 1359 1360
				 */
				list_del(&li->llist);
				cifs_del_lock_waiters(li);
				kfree(li);
1361
				continue;
1362
			}
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
			cur->Pid = cpu_to_le16(li->pid);
			cur->LengthLow = cpu_to_le32((u32)li->length);
			cur->LengthHigh = cpu_to_le32((u32)(li->length>>32));
			cur->OffsetLow = cpu_to_le32((u32)li->offset);
			cur->OffsetHigh = cpu_to_le32((u32)(li->offset>>32));
			/*
			 * We need to save a lock here to let us add it again to
			 * the file's list if the unlock range request fails on
			 * the server.
			 */
			list_move(&li->llist, &tmp_llist);
			if (++num == max_num) {
1375 1376
				stored_rc = cifs_lockv(xid, tcon,
						       cfile->fid.netfid,
1377 1378 1379 1380 1381 1382 1383 1384
						       li->type, num, 0, buf);
				if (stored_rc) {
					/*
					 * We failed on the unlock range
					 * request - add all locks from the tmp
					 * list to the head of the file's list.
					 */
					cifs_move_llist(&tmp_llist,
1385
							&cfile->llist->locks);
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396
					rc = stored_rc;
				} else
					/*
					 * The unlock range request succeed -
					 * free the tmp list.
					 */
					cifs_free_llist(&tmp_llist);
				cur = buf;
				num = 0;
			} else
				cur++;
1397 1398
		}
		if (num) {
1399
			stored_rc = cifs_lockv(xid, tcon, cfile->fid.netfid,
1400 1401
					       types[i], num, 0, buf);
			if (stored_rc) {
1402 1403
				cifs_move_llist(&tmp_llist,
						&cfile->llist->locks);
1404 1405 1406 1407 1408 1409
				rc = stored_rc;
			} else
				cifs_free_llist(&tmp_llist);
		}
	}

1410
	up_write(&cinode->lock_sem);
1411 1412 1413 1414
	kfree(buf);
	return rc;
}

1415
static int
1416
cifs_setlk(struct file *file, struct file_lock *flock, __u32 type,
1417 1418
	   bool wait_flag, bool posix_lck, int lock, int unlock,
	   unsigned int xid)
1419 1420 1421 1422 1423
{
	int rc = 0;
	__u64 length = 1 + flock->fl_end - flock->fl_start;
	struct cifsFileInfo *cfile = (struct cifsFileInfo *)file->private_data;
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
1424
	struct TCP_Server_Info *server = tcon->ses->server;
1425 1426

	if (posix_lck) {
1427
		int posix_lock_type;
1428 1429 1430 1431 1432

		rc = cifs_posix_lock_set(file, flock);
		if (!rc || rc < 0)
			return rc;

1433
		if (type & server->vals->shared_lock_type)
1434 1435 1436
			posix_lock_type = CIFS_RDLCK;
		else
			posix_lock_type = CIFS_WRLCK;
1437

1438
		if (unlock == 1)
1439
			posix_lock_type = CIFS_UNLCK;
J
[CIFS]  
Jeremy Allison 已提交
1440

1441 1442 1443
		rc = CIFSSMBPosixLock(xid, tcon, cfile->fid.netfid,
				      current->tgid, flock->fl_start, length,
				      NULL, posix_lock_type, wait_flag);
1444 1445
		goto out;
	}
J
[CIFS]  
Jeremy Allison 已提交
1446

1447
	if (lock) {
1448 1449
		struct cifsLockInfo *lock;

1450
		lock = cifs_lock_init(flock->fl_start, length, type);
1451 1452 1453
		if (!lock)
			return -ENOMEM;

1454
		rc = cifs_lock_add_if(cfile, lock, wait_flag);
1455
		if (rc < 0) {
1456
			kfree(lock);
1457 1458 1459
			return rc;
		}
		if (!rc)
1460 1461
			goto out;

1462 1463
		rc = server->ops->mand_lock(xid, cfile, flock->fl_start, length,
					    type, 1, 0, wait_flag);
1464 1465
		if (rc) {
			kfree(lock);
1466
			return rc;
1467
		}
1468

1469
		cifs_lock_add(cfile, lock);
1470
	} else if (unlock)
1471
		rc = server->ops->mand_unlock_range(cfile, flock, xid);
1472 1473 1474

out:
	if (flock->fl_flags & FL_POSIX)
1475
		posix_lock_file_wait(file, flock);
1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489
	return rc;
}

int cifs_lock(struct file *file, int cmd, struct file_lock *flock)
{
	int rc, xid;
	int lock = 0, unlock = 0;
	bool wait_flag = false;
	bool posix_lck = false;
	struct cifs_sb_info *cifs_sb;
	struct cifs_tcon *tcon;
	struct cifsInodeInfo *cinode;
	struct cifsFileInfo *cfile;
	__u16 netfid;
1490
	__u32 type;
1491 1492

	rc = -EACCES;
1493
	xid = get_xid();
1494 1495 1496 1497 1498 1499 1500

	cFYI(1, "Lock parm: 0x%x flockflags: 0x%x flocktype: 0x%x start: %lld "
		"end: %lld", cmd, flock->fl_flags, flock->fl_type,
		flock->fl_start, flock->fl_end);

	cfile = (struct cifsFileInfo *)file->private_data;
	tcon = tlink_tcon(cfile->tlink);
1501 1502 1503 1504 1505

	cifs_read_flock(flock, &type, &lock, &unlock, &wait_flag,
			tcon->ses->server);

	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
1506
	netfid = cfile->fid.netfid;
1507 1508
	cinode = CIFS_I(file->f_path.dentry->d_inode);

1509
	if (cap_unix(tcon->ses) &&
1510 1511 1512 1513 1514 1515 1516 1517
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
		posix_lck = true;
	/*
	 * BB add code here to normalize offset and length to account for
	 * negative length which we can not accept over the wire.
	 */
	if (IS_GETLK(cmd)) {
1518
		rc = cifs_getlk(file, flock, type, wait_flag, posix_lck, xid);
1519
		free_xid(xid);
1520 1521 1522 1523 1524 1525 1526 1527
		return rc;
	}

	if (!lock && !unlock) {
		/*
		 * if no lock or unlock then nothing to do since we do not
		 * know what it is
		 */
1528
		free_xid(xid);
1529
		return -EOPNOTSUPP;
J
[CIFS]  
Jeremy Allison 已提交
1530 1531
	}

1532 1533
	rc = cifs_setlk(file, flock, type, wait_flag, posix_lck, lock, unlock,
			xid);
1534
	free_xid(xid);
L
Linus Torvalds 已提交
1535 1536 1537
	return rc;
}

1538 1539 1540 1541
/*
 * update the file size (if needed) after a write. Should be called with
 * the inode->i_lock held
 */
1542
void
1543 1544 1545 1546 1547 1548 1549 1550 1551
cifs_update_eof(struct cifsInodeInfo *cifsi, loff_t offset,
		      unsigned int bytes_written)
{
	loff_t end_of_write = offset + bytes_written;

	if (end_of_write > cifsi->server_eof)
		cifsi->server_eof = end_of_write;
}

1552 1553 1554
static ssize_t
cifs_write(struct cifsFileInfo *open_file, __u32 pid, const char *write_data,
	   size_t write_size, loff_t *offset)
L
Linus Torvalds 已提交
1555 1556 1557 1558 1559
{
	int rc = 0;
	unsigned int bytes_written = 0;
	unsigned int total_written;
	struct cifs_sb_info *cifs_sb;
1560 1561
	struct cifs_tcon *tcon;
	struct TCP_Server_Info *server;
1562
	unsigned int xid;
1563 1564
	struct dentry *dentry = open_file->dentry;
	struct cifsInodeInfo *cifsi = CIFS_I(dentry->d_inode);
1565
	struct cifs_io_parms io_parms;
L
Linus Torvalds 已提交
1566

1567
	cifs_sb = CIFS_SB(dentry->d_sb);
L
Linus Torvalds 已提交
1568

1569
	cFYI(1, "write %zd bytes to offset %lld of %s", write_size,
1570
	     *offset, dentry->d_name.name);
L
Linus Torvalds 已提交
1571

1572 1573 1574 1575 1576
	tcon = tlink_tcon(open_file->tlink);
	server = tcon->ses->server;

	if (!server->ops->sync_write)
		return -ENOSYS;
1577

1578
	xid = get_xid();
L
Linus Torvalds 已提交
1579 1580 1581 1582 1583

	for (total_written = 0; write_size > total_written;
	     total_written += bytes_written) {
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
1584 1585 1586
			struct kvec iov[2];
			unsigned int len;

L
Linus Torvalds 已提交
1587 1588 1589
			if (open_file->invalidHandle) {
				/* we could deadlock if we called
				   filemap_fdatawait from here so tell
S
Steve French 已提交
1590
				   reopen_file not to flush data to
L
Linus Torvalds 已提交
1591
				   server now */
J
Jeff Layton 已提交
1592
				rc = cifs_reopen_file(open_file, false);
L
Linus Torvalds 已提交
1593 1594 1595
				if (rc != 0)
					break;
			}
1596 1597 1598 1599 1600 1601

			len = min((size_t)cifs_sb->wsize,
				  write_size - total_written);
			/* iov[0] is reserved for smb header */
			iov[1].iov_base = (char *)write_data + total_written;
			iov[1].iov_len = len;
1602
			io_parms.pid = pid;
1603 1604
			io_parms.tcon = tcon;
			io_parms.offset = *offset;
1605
			io_parms.length = len;
1606 1607
			rc = server->ops->sync_write(xid, open_file, &io_parms,
						     &bytes_written, iov, 1);
L
Linus Torvalds 已提交
1608 1609 1610 1611 1612
		}
		if (rc || (bytes_written == 0)) {
			if (total_written)
				break;
			else {
1613
				free_xid(xid);
L
Linus Torvalds 已提交
1614 1615
				return rc;
			}
1616
		} else {
1617
			spin_lock(&dentry->d_inode->i_lock);
1618
			cifs_update_eof(cifsi, *offset, bytes_written);
1619
			spin_unlock(&dentry->d_inode->i_lock);
1620
			*offset += bytes_written;
1621
		}
L
Linus Torvalds 已提交
1622 1623
	}

1624
	cifs_stats_bytes_written(tcon, total_written);
L
Linus Torvalds 已提交
1625

1626 1627
	if (total_written > 0) {
		spin_lock(&dentry->d_inode->i_lock);
1628 1629
		if (*offset > dentry->d_inode->i_size)
			i_size_write(dentry->d_inode, *offset);
1630
		spin_unlock(&dentry->d_inode->i_lock);
L
Linus Torvalds 已提交
1631
	}
1632
	mark_inode_dirty_sync(dentry->d_inode);
1633
	free_xid(xid);
L
Linus Torvalds 已提交
1634 1635 1636
	return total_written;
}

1637 1638
struct cifsFileInfo *find_readable_file(struct cifsInodeInfo *cifs_inode,
					bool fsuid_only)
S
Steve French 已提交
1639 1640
{
	struct cifsFileInfo *open_file = NULL;
1641 1642 1643 1644 1645
	struct cifs_sb_info *cifs_sb = CIFS_SB(cifs_inode->vfs_inode.i_sb);

	/* only filter by fsuid on multiuser mounts */
	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MULTIUSER))
		fsuid_only = false;
S
Steve French 已提交
1646

1647
	spin_lock(&cifs_file_list_lock);
S
Steve French 已提交
1648 1649 1650 1651
	/* we could simply get the first_list_entry since write-only entries
	   are always at the end of the list but since the first entry might
	   have a close pending, we go through the whole list */
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1652 1653
		if (fsuid_only && open_file->uid != current_fsuid())
			continue;
1654
		if (OPEN_FMODE(open_file->f_flags) & FMODE_READ) {
S
Steve French 已提交
1655 1656 1657
			if (!open_file->invalidHandle) {
				/* found a good file */
				/* lock it so it will not be closed on us */
1658
				cifsFileInfo_get_locked(open_file);
1659
				spin_unlock(&cifs_file_list_lock);
S
Steve French 已提交
1660 1661 1662 1663 1664 1665 1666
				return open_file;
			} /* else might as well continue, and look for
			     another, or simply have the caller reopen it
			     again rather than trying to fix this handle */
		} else /* write only file */
			break; /* write only files are last so must be done */
	}
1667
	spin_unlock(&cifs_file_list_lock);
S
Steve French 已提交
1668 1669 1670
	return NULL;
}

1671 1672
struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode,
					bool fsuid_only)
1673
{
1674
	struct cifsFileInfo *open_file, *inv_file = NULL;
1675
	struct cifs_sb_info *cifs_sb;
1676
	bool any_available = false;
1677
	int rc;
1678
	unsigned int refind = 0;
1679

1680 1681 1682 1683
	/* Having a null inode here (because mapping->host was set to zero by
	the VFS or MM) should not happen but we had reports of on oops (due to
	it being zero) during stress testcases so we need to check for it */

S
Steve French 已提交
1684
	if (cifs_inode == NULL) {
1685
		cERROR(1, "Null inode passed to cifs_writeable_file");
1686 1687 1688 1689
		dump_stack();
		return NULL;
	}

1690 1691
	cifs_sb = CIFS_SB(cifs_inode->vfs_inode.i_sb);

1692 1693 1694 1695
	/* only filter by fsuid on multiuser mounts */
	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MULTIUSER))
		fsuid_only = false;

1696
	spin_lock(&cifs_file_list_lock);
1697
refind_writable:
1698 1699 1700 1701
	if (refind > MAX_REOPEN_ATT) {
		spin_unlock(&cifs_file_list_lock);
		return NULL;
	}
1702
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1703 1704 1705
		if (!any_available && open_file->pid != current->tgid)
			continue;
		if (fsuid_only && open_file->uid != current_fsuid())
1706
			continue;
1707
		if (OPEN_FMODE(open_file->f_flags) & FMODE_WRITE) {
1708 1709
			if (!open_file->invalidHandle) {
				/* found a good writable file */
1710
				cifsFileInfo_get_locked(open_file);
1711
				spin_unlock(&cifs_file_list_lock);
1712
				return open_file;
1713 1714 1715
			} else {
				if (!inv_file)
					inv_file = open_file;
1716
			}
1717 1718
		}
	}
1719 1720 1721 1722 1723
	/* couldn't find useable FH with same pid, try any available */
	if (!any_available) {
		any_available = true;
		goto refind_writable;
	}
1724 1725 1726

	if (inv_file) {
		any_available = false;
1727
		cifsFileInfo_get_locked(inv_file);
1728 1729
	}

1730
	spin_unlock(&cifs_file_list_lock);
1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747

	if (inv_file) {
		rc = cifs_reopen_file(inv_file, false);
		if (!rc)
			return inv_file;
		else {
			spin_lock(&cifs_file_list_lock);
			list_move_tail(&inv_file->flist,
					&cifs_inode->openFileList);
			spin_unlock(&cifs_file_list_lock);
			cifsFileInfo_put(inv_file);
			spin_lock(&cifs_file_list_lock);
			++refind;
			goto refind_writable;
		}
	}

1748 1749 1750
	return NULL;
}

L
Linus Torvalds 已提交
1751 1752 1753 1754 1755 1756 1757 1758
static int cifs_partialpagewrite(struct page *page, unsigned from, unsigned to)
{
	struct address_space *mapping = page->mapping;
	loff_t offset = (loff_t)page->index << PAGE_CACHE_SHIFT;
	char *write_data;
	int rc = -EFAULT;
	int bytes_written = 0;
	struct inode *inode;
1759
	struct cifsFileInfo *open_file;
L
Linus Torvalds 已提交
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782

	if (!mapping || !mapping->host)
		return -EFAULT;

	inode = page->mapping->host;

	offset += (loff_t)from;
	write_data = kmap(page);
	write_data += from;

	if ((to > PAGE_CACHE_SIZE) || (from > to)) {
		kunmap(page);
		return -EIO;
	}

	/* racing with truncate? */
	if (offset > mapping->host->i_size) {
		kunmap(page);
		return 0; /* don't care */
	}

	/* check to make sure that we are not extending the file */
	if (mapping->host->i_size - offset < (loff_t)to)
S
Steve French 已提交
1783
		to = (unsigned)(mapping->host->i_size - offset);
L
Linus Torvalds 已提交
1784

1785
	open_file = find_writable_file(CIFS_I(mapping->host), false);
1786
	if (open_file) {
1787 1788
		bytes_written = cifs_write(open_file, open_file->pid,
					   write_data, to - from, &offset);
1789
		cifsFileInfo_put(open_file);
L
Linus Torvalds 已提交
1790
		/* Does mm or vfs already set times? */
1791
		inode->i_atime = inode->i_mtime = current_fs_time(inode->i_sb);
1792
		if ((bytes_written > 0) && (offset))
1793
			rc = 0;
1794 1795
		else if (bytes_written < 0)
			rc = bytes_written;
1796
	} else {
1797
		cFYI(1, "No writeable filehandles for inode");
L
Linus Torvalds 已提交
1798 1799 1800 1801 1802 1803 1804 1805
		rc = -EIO;
	}

	kunmap(page);
	return rc;
}

static int cifs_writepages(struct address_space *mapping,
1806
			   struct writeback_control *wbc)
L
Linus Torvalds 已提交
1807
{
1808 1809 1810 1811
	struct cifs_sb_info *cifs_sb = CIFS_SB(mapping->host->i_sb);
	bool done = false, scanned = false, range_whole = false;
	pgoff_t end, index;
	struct cifs_writedata *wdata;
1812
	struct TCP_Server_Info *server;
1813 1814
	struct page *page;
	int rc = 0;
1815

1816
	/*
1817
	 * If wsize is smaller than the page cache size, default to writing
1818 1819 1820 1821 1822
	 * one page at a time via cifs_writepage
	 */
	if (cifs_sb->wsize < PAGE_CACHE_SIZE)
		return generic_writepages(mapping, wbc);

1823
	if (wbc->range_cyclic) {
1824
		index = mapping->writeback_index; /* Start from prev offset */
1825 1826 1827 1828 1829
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
1830 1831
			range_whole = true;
		scanned = true;
1832 1833
	}
retry:
1834 1835 1836 1837 1838 1839 1840 1841
	while (!done && index <= end) {
		unsigned int i, nr_pages, found_pages;
		pgoff_t next = 0, tofind;
		struct page **pages;

		tofind = min((cifs_sb->wsize / PAGE_CACHE_SIZE) - 1,
				end - index) + 1;

1842 1843
		wdata = cifs_writedata_alloc((unsigned int)tofind,
					     cifs_writev_complete);
1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
		if (!wdata) {
			rc = -ENOMEM;
			break;
		}

		/*
		 * find_get_pages_tag seems to return a max of 256 on each
		 * iteration, so we must call it several times in order to
		 * fill the array or the wsize is effectively limited to
		 * 256 * PAGE_CACHE_SIZE.
		 */
		found_pages = 0;
		pages = wdata->pages;
		do {
			nr_pages = find_get_pages_tag(mapping, &index,
							PAGECACHE_TAG_DIRTY,
							tofind, pages);
			found_pages += nr_pages;
			tofind -= nr_pages;
			pages += nr_pages;
		} while (nr_pages && tofind && index <= end);

		if (found_pages == 0) {
			kref_put(&wdata->refcount, cifs_writedata_release);
			break;
		}

		nr_pages = 0;
		for (i = 0; i < found_pages; i++) {
			page = wdata->pages[i];
1874 1875 1876 1877 1878 1879 1880 1881
			/*
			 * At this point we hold neither mapping->tree_lock nor
			 * lock on the page itself: the page may be truncated or
			 * invalidated (changing page->mapping to NULL), or even
			 * swizzled back from swapper_space to tmpfs file
			 * mapping
			 */

1882
			if (nr_pages == 0)
1883
				lock_page(page);
N
Nick Piggin 已提交
1884
			else if (!trylock_page(page))
1885 1886 1887 1888 1889 1890 1891
				break;

			if (unlikely(page->mapping != mapping)) {
				unlock_page(page);
				break;
			}

1892
			if (!wbc->range_cyclic && page->index > end) {
1893
				done = true;
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
				unlock_page(page);
				break;
			}

			if (next && (page->index != next)) {
				/* Not next consecutive page */
				unlock_page(page);
				break;
			}

			if (wbc->sync_mode != WB_SYNC_NONE)
				wait_on_page_writeback(page);

			if (PageWriteback(page) ||
1908
					!clear_page_dirty_for_io(page)) {
1909 1910 1911
				unlock_page(page);
				break;
			}
1912

1913 1914 1915 1916 1917 1918
			/*
			 * This actually clears the dirty bit in the radix tree.
			 * See cifs_writepage() for more commentary.
			 */
			set_page_writeback(page);

1919
			if (page_offset(page) >= i_size_read(mapping->host)) {
1920
				done = true;
1921
				unlock_page(page);
1922
				end_page_writeback(page);
1923 1924 1925
				break;
			}

1926 1927 1928 1929
			wdata->pages[i] = page;
			next = page->index + 1;
			++nr_pages;
		}
1930

1931 1932 1933
		/* reset index to refind any pages skipped */
		if (nr_pages == 0)
			index = wdata->pages[0]->index + 1;
1934

1935 1936 1937 1938 1939
		/* put any pages we aren't going to use */
		for (i = nr_pages; i < found_pages; i++) {
			page_cache_release(wdata->pages[i]);
			wdata->pages[i] = NULL;
		}
1940

1941 1942 1943 1944
		/* nothing to write? */
		if (nr_pages == 0) {
			kref_put(&wdata->refcount, cifs_writedata_release);
			continue;
1945
		}
1946

1947 1948 1949
		wdata->sync_mode = wbc->sync_mode;
		wdata->nr_pages = nr_pages;
		wdata->offset = page_offset(wdata->pages[0]);
1950 1951
		wdata->pagesz = PAGE_CACHE_SIZE;
		wdata->tailsz =
1952 1953
			min(i_size_read(mapping->host) -
			    page_offset(wdata->pages[nr_pages - 1]),
1954 1955 1956
			    (loff_t)PAGE_CACHE_SIZE);
		wdata->bytes = ((nr_pages - 1) * PAGE_CACHE_SIZE) +
					wdata->tailsz;
1957

1958 1959 1960 1961 1962 1963 1964 1965 1966
		do {
			if (wdata->cfile != NULL)
				cifsFileInfo_put(wdata->cfile);
			wdata->cfile = find_writable_file(CIFS_I(mapping->host),
							  false);
			if (!wdata->cfile) {
				cERROR(1, "No writable handles for inode");
				rc = -EBADF;
				break;
1967
			}
1968
			wdata->pid = wdata->cfile->pid;
1969 1970
			server = tlink_tcon(wdata->cfile->tlink)->ses->server;
			rc = server->ops->async_writev(wdata);
1971
		} while (wbc->sync_mode == WB_SYNC_ALL && rc == -EAGAIN);
1972

1973 1974
		for (i = 0; i < nr_pages; ++i)
			unlock_page(wdata->pages[i]);
1975

1976 1977 1978
		/* send failure -- clean up the mess */
		if (rc != 0) {
			for (i = 0; i < nr_pages; ++i) {
1979
				if (rc == -EAGAIN)
1980 1981 1982 1983 1984 1985
					redirty_page_for_writepage(wbc,
							   wdata->pages[i]);
				else
					SetPageError(wdata->pages[i]);
				end_page_writeback(wdata->pages[i]);
				page_cache_release(wdata->pages[i]);
1986
			}
1987 1988
			if (rc != -EAGAIN)
				mapping_set_error(mapping, rc);
1989 1990
		}
		kref_put(&wdata->refcount, cifs_writedata_release);
1991

1992 1993 1994
		wbc->nr_to_write -= nr_pages;
		if (wbc->nr_to_write <= 0)
			done = true;
1995

1996
		index = next;
1997
	}
1998

1999 2000 2001 2002 2003
	if (!scanned && !done) {
		/*
		 * We hit the last page and there is more work to be done: wrap
		 * back to the start of the file
		 */
2004
		scanned = true;
2005 2006 2007
		index = 0;
		goto retry;
	}
2008

2009
	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
2010 2011
		mapping->writeback_index = index;

L
Linus Torvalds 已提交
2012 2013 2014
	return rc;
}

2015 2016
static int
cifs_writepage_locked(struct page *page, struct writeback_control *wbc)
L
Linus Torvalds 已提交
2017
{
2018
	int rc;
2019
	unsigned int xid;
L
Linus Torvalds 已提交
2020

2021
	xid = get_xid();
L
Linus Torvalds 已提交
2022 2023
/* BB add check for wbc flags */
	page_cache_get(page);
S
Steve French 已提交
2024
	if (!PageUptodate(page))
2025
		cFYI(1, "ppw - page not up to date");
2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036

	/*
	 * Set the "writeback" flag, and clear "dirty" in the radix tree.
	 *
	 * A writepage() implementation always needs to do either this,
	 * or re-dirty the page with "redirty_page_for_writepage()" in
	 * the case of a failure.
	 *
	 * Just unlocking the page will cause the radix tree tag-bits
	 * to fail to update with the state of the page correctly.
	 */
S
Steve French 已提交
2037
	set_page_writeback(page);
2038
retry_write:
L
Linus Torvalds 已提交
2039
	rc = cifs_partialpagewrite(page, 0, PAGE_CACHE_SIZE);
2040 2041 2042 2043 2044 2045 2046 2047
	if (rc == -EAGAIN && wbc->sync_mode == WB_SYNC_ALL)
		goto retry_write;
	else if (rc == -EAGAIN)
		redirty_page_for_writepage(wbc, page);
	else if (rc != 0)
		SetPageError(page);
	else
		SetPageUptodate(page);
2048 2049
	end_page_writeback(page);
	page_cache_release(page);
2050
	free_xid(xid);
L
Linus Torvalds 已提交
2051 2052 2053
	return rc;
}

2054 2055 2056 2057 2058 2059 2060
static int cifs_writepage(struct page *page, struct writeback_control *wbc)
{
	int rc = cifs_writepage_locked(page, wbc);
	unlock_page(page);
	return rc;
}

N
Nick Piggin 已提交
2061 2062 2063
static int cifs_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
L
Linus Torvalds 已提交
2064
{
N
Nick Piggin 已提交
2065 2066
	int rc;
	struct inode *inode = mapping->host;
2067 2068 2069 2070 2071 2072 2073 2074
	struct cifsFileInfo *cfile = file->private_data;
	struct cifs_sb_info *cifs_sb = CIFS_SB(cfile->dentry->d_sb);
	__u32 pid;

	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
		pid = cfile->pid;
	else
		pid = current->tgid;
L
Linus Torvalds 已提交
2075

2076 2077
	cFYI(1, "write_end for page %p from pos %lld with %d bytes",
		 page, pos, copied);
N
Nick Piggin 已提交
2078

2079 2080 2081 2082 2083
	if (PageChecked(page)) {
		if (copied == len)
			SetPageUptodate(page);
		ClearPageChecked(page);
	} else if (!PageUptodate(page) && copied == PAGE_CACHE_SIZE)
N
Nick Piggin 已提交
2084
		SetPageUptodate(page);
S
Steve French 已提交
2085

L
Linus Torvalds 已提交
2086
	if (!PageUptodate(page)) {
N
Nick Piggin 已提交
2087 2088
		char *page_data;
		unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
2089
		unsigned int xid;
N
Nick Piggin 已提交
2090

2091
		xid = get_xid();
L
Linus Torvalds 已提交
2092 2093 2094 2095 2096 2097
		/* this is probably better than directly calling
		   partialpage_write since in this function the file handle is
		   known which we might as well	leverage */
		/* BB check if anything else missing out of ppw
		   such as updating last write time */
		page_data = kmap(page);
2098
		rc = cifs_write(cfile, pid, page_data + offset, copied, &pos);
N
Nick Piggin 已提交
2099
		/* if (rc < 0) should we set writebehind rc? */
L
Linus Torvalds 已提交
2100
		kunmap(page);
N
Nick Piggin 已提交
2101

2102
		free_xid(xid);
S
Steve French 已提交
2103
	} else {
N
Nick Piggin 已提交
2104 2105
		rc = copied;
		pos += copied;
2106 2107 2108 2109 2110 2111 2112 2113 2114
		/*
		 * When we use strict cache mode and cifs_strict_writev was run
		 * with level II oplock (indicated by leave_pages_clean field of
		 * CIFS_I(inode)), we can leave pages clean - cifs_strict_writev
		 * sent the data to the server itself.
		 */
		if (!CIFS_I(inode)->leave_pages_clean ||
		    !(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO))
			set_page_dirty(page);
L
Linus Torvalds 已提交
2115 2116
	}

N
Nick Piggin 已提交
2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
	if (rc > 0) {
		spin_lock(&inode->i_lock);
		if (pos > inode->i_size)
			i_size_write(inode, pos);
		spin_unlock(&inode->i_lock);
	}

	unlock_page(page);
	page_cache_release(page);

L
Linus Torvalds 已提交
2127 2128 2129
	return rc;
}

2130 2131
int cifs_strict_fsync(struct file *file, loff_t start, loff_t end,
		      int datasync)
L
Linus Torvalds 已提交
2132
{
2133
	unsigned int xid;
L
Linus Torvalds 已提交
2134
	int rc = 0;
2135
	struct cifs_tcon *tcon;
2136
	struct TCP_Server_Info *server;
2137
	struct cifsFileInfo *smbfile = file->private_data;
2138
	struct inode *inode = file->f_path.dentry->d_inode;
2139
	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
L
Linus Torvalds 已提交
2140

2141 2142 2143 2144 2145
	rc = filemap_write_and_wait_range(inode->i_mapping, start, end);
	if (rc)
		return rc;
	mutex_lock(&inode->i_mutex);

2146
	xid = get_xid();
L
Linus Torvalds 已提交
2147

2148
	cFYI(1, "Sync file - name: %s datasync: 0x%x",
2149
		file->f_path.dentry->d_name.name, datasync);
2150

2151 2152 2153 2154 2155 2156 2157
	if (!CIFS_I(inode)->clientCanCacheRead) {
		rc = cifs_invalidate_mapping(inode);
		if (rc) {
			cFYI(1, "rc: %d during invalidate phase", rc);
			rc = 0; /* don't care about it in fsync */
		}
	}
2158

2159
	tcon = tlink_tcon(smbfile->tlink);
2160 2161 2162 2163 2164 2165 2166
	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC)) {
		server = tcon->ses->server;
		if (server->ops->flush)
			rc = server->ops->flush(xid, tcon, &smbfile->fid);
		else
			rc = -ENOSYS;
	}
2167

2168
	free_xid(xid);
2169
	mutex_unlock(&inode->i_mutex);
2170 2171 2172
	return rc;
}

2173
int cifs_fsync(struct file *file, loff_t start, loff_t end, int datasync)
2174
{
2175
	unsigned int xid;
2176
	int rc = 0;
2177
	struct cifs_tcon *tcon;
2178
	struct TCP_Server_Info *server;
2179 2180
	struct cifsFileInfo *smbfile = file->private_data;
	struct cifs_sb_info *cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
2181 2182 2183 2184 2185 2186
	struct inode *inode = file->f_mapping->host;

	rc = filemap_write_and_wait_range(inode->i_mapping, start, end);
	if (rc)
		return rc;
	mutex_lock(&inode->i_mutex);
2187

2188
	xid = get_xid();
2189 2190 2191 2192 2193

	cFYI(1, "Sync file - name: %s datasync: 0x%x",
		file->f_path.dentry->d_name.name, datasync);

	tcon = tlink_tcon(smbfile->tlink);
2194 2195 2196 2197 2198 2199 2200
	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC)) {
		server = tcon->ses->server;
		if (server->ops->flush)
			rc = server->ops->flush(xid, tcon, &smbfile->fid);
		else
			rc = -ENOSYS;
	}
2201

2202
	free_xid(xid);
2203
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
2204 2205 2206 2207 2208 2209 2210
	return rc;
}

/*
 * As file closes, flush all cached write data for this inode checking
 * for write behind errors.
 */
2211
int cifs_flush(struct file *file, fl_owner_t id)
L
Linus Torvalds 已提交
2212
{
S
Steve French 已提交
2213
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
2214 2215
	int rc = 0;

2216
	if (file->f_mode & FMODE_WRITE)
2217
		rc = filemap_write_and_wait(inode->i_mapping);
2218

2219
	cFYI(1, "Flush inode %p file %p rc %d", inode, file, rc);
L
Linus Torvalds 已提交
2220 2221 2222 2223

	return rc;
}

2224 2225 2226 2227 2228 2229 2230
static int
cifs_write_allocate_pages(struct page **pages, unsigned long num_pages)
{
	int rc = 0;
	unsigned long i;

	for (i = 0; i < num_pages; i++) {
2231
		pages[i] = alloc_page(GFP_KERNEL|__GFP_HIGHMEM);
2232 2233 2234 2235 2236 2237 2238
		if (!pages[i]) {
			/*
			 * save number of pages we have already allocated and
			 * return with ENOMEM error
			 */
			num_pages = i;
			rc = -ENOMEM;
2239
			break;
2240 2241 2242
		}
	}

2243 2244 2245 2246
	if (rc) {
		for (i = 0; i < num_pages; i++)
			put_page(pages[i]);
	}
2247 2248 2249 2250 2251 2252 2253 2254 2255 2256
	return rc;
}

static inline
size_t get_numpages(const size_t wsize, const size_t len, size_t *cur_len)
{
	size_t num_pages;
	size_t clen;

	clen = min_t(const size_t, len, wsize);
J
Jeff Layton 已提交
2257
	num_pages = DIV_ROUND_UP(clen, PAGE_SIZE);
2258 2259 2260 2261 2262 2263 2264

	if (cur_len)
		*cur_len = clen;

	return num_pages;
}

2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294
static void
cifs_uncached_writev_complete(struct work_struct *work)
{
	int i;
	struct cifs_writedata *wdata = container_of(work,
					struct cifs_writedata, work);
	struct inode *inode = wdata->cfile->dentry->d_inode;
	struct cifsInodeInfo *cifsi = CIFS_I(inode);

	spin_lock(&inode->i_lock);
	cifs_update_eof(cifsi, wdata->offset, wdata->bytes);
	if (cifsi->server_eof > inode->i_size)
		i_size_write(inode, cifsi->server_eof);
	spin_unlock(&inode->i_lock);

	complete(&wdata->done);

	if (wdata->result != -EAGAIN) {
		for (i = 0; i < wdata->nr_pages; i++)
			put_page(wdata->pages[i]);
	}

	kref_put(&wdata->refcount, cifs_writedata_release);
}

/* attempt to send write to server, retry on any -EAGAIN errors */
static int
cifs_uncached_retry_writev(struct cifs_writedata *wdata)
{
	int rc;
2295 2296 2297
	struct TCP_Server_Info *server;

	server = tlink_tcon(wdata->cfile->tlink)->ses->server;
2298 2299 2300 2301 2302 2303 2304

	do {
		if (wdata->cfile->invalidHandle) {
			rc = cifs_reopen_file(wdata->cfile, false);
			if (rc != 0)
				continue;
		}
2305
		rc = server->ops->async_writev(wdata);
2306 2307 2308 2309 2310
	} while (rc == -EAGAIN);

	return rc;
}

2311 2312 2313 2314
static ssize_t
cifs_iovec_write(struct file *file, const struct iovec *iov,
		 unsigned long nr_segs, loff_t *poffset)
{
2315
	unsigned long nr_pages, i;
2316 2317
	size_t copied, len, cur_len;
	ssize_t total_written = 0;
2318
	loff_t offset;
2319 2320
	struct iov_iter it;
	struct cifsFileInfo *open_file;
2321
	struct cifs_tcon *tcon;
2322
	struct cifs_sb_info *cifs_sb;
2323 2324 2325 2326
	struct cifs_writedata *wdata, *tmp;
	struct list_head wdata_list;
	int rc;
	pid_t pid;
2327 2328 2329 2330 2331 2332 2333 2334 2335

	len = iov_length(iov, nr_segs);
	if (!len)
		return 0;

	rc = generic_write_checks(file, poffset, &len, 0);
	if (rc)
		return rc;

2336
	INIT_LIST_HEAD(&wdata_list);
2337 2338
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
	open_file = file->private_data;
2339
	tcon = tlink_tcon(open_file->tlink);
2340 2341 2342 2343

	if (!tcon->ses->server->ops->async_writev)
		return -ENOSYS;

2344
	offset = *poffset;
2345 2346 2347 2348 2349 2350

	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
		pid = open_file->pid;
	else
		pid = current->tgid;

2351 2352
	iov_iter_init(&it, iov, nr_segs, len, 0);
	do {
2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373
		size_t save_len;

		nr_pages = get_numpages(cifs_sb->wsize, len, &cur_len);
		wdata = cifs_writedata_alloc(nr_pages,
					     cifs_uncached_writev_complete);
		if (!wdata) {
			rc = -ENOMEM;
			break;
		}

		rc = cifs_write_allocate_pages(wdata->pages, nr_pages);
		if (rc) {
			kfree(wdata);
			break;
		}

		save_len = cur_len;
		for (i = 0; i < nr_pages; i++) {
			copied = min_t(const size_t, cur_len, PAGE_SIZE);
			copied = iov_iter_copy_from_user(wdata->pages[i], &it,
							 0, copied);
2374 2375 2376 2377 2378
			cur_len -= copied;
			iov_iter_advance(&it, copied);
		}
		cur_len = save_len - cur_len;

2379 2380 2381 2382 2383 2384
		wdata->sync_mode = WB_SYNC_ALL;
		wdata->nr_pages = nr_pages;
		wdata->offset = (__u64)offset;
		wdata->cfile = cifsFileInfo_get(open_file);
		wdata->pid = pid;
		wdata->bytes = cur_len;
2385 2386
		wdata->pagesz = PAGE_SIZE;
		wdata->tailsz = cur_len - ((nr_pages - 1) * PAGE_SIZE);
2387 2388 2389
		rc = cifs_uncached_retry_writev(wdata);
		if (rc) {
			kref_put(&wdata->refcount, cifs_writedata_release);
2390 2391 2392
			break;
		}

2393 2394 2395
		list_add_tail(&wdata->list, &wdata_list);
		offset += cur_len;
		len -= cur_len;
2396 2397
	} while (len > 0);

2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431
	/*
	 * If at least one write was successfully sent, then discard any rc
	 * value from the later writes. If the other write succeeds, then
	 * we'll end up returning whatever was written. If it fails, then
	 * we'll get a new rc value from that.
	 */
	if (!list_empty(&wdata_list))
		rc = 0;

	/*
	 * Wait for and collect replies for any successful sends in order of
	 * increasing offset. Once an error is hit or we get a fatal signal
	 * while waiting, then return without waiting for any more replies.
	 */
restart_loop:
	list_for_each_entry_safe(wdata, tmp, &wdata_list, list) {
		if (!rc) {
			/* FIXME: freezable too? */
			rc = wait_for_completion_killable(&wdata->done);
			if (rc)
				rc = -EINTR;
			else if (wdata->result)
				rc = wdata->result;
			else
				total_written += wdata->bytes;

			/* resend call if it's a retryable error */
			if (rc == -EAGAIN) {
				rc = cifs_uncached_retry_writev(wdata);
				goto restart_loop;
			}
		}
		list_del_init(&wdata->list);
		kref_put(&wdata->refcount, cifs_writedata_release);
2432 2433
	}

2434 2435
	if (total_written > 0)
		*poffset += total_written;
2436

2437 2438
	cifs_stats_bytes_written(tcon, total_written);
	return total_written ? total_written : (ssize_t)rc;
2439 2440
}

2441
ssize_t cifs_user_writev(struct kiocb *iocb, const struct iovec *iov,
2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463
				unsigned long nr_segs, loff_t pos)
{
	ssize_t written;
	struct inode *inode;

	inode = iocb->ki_filp->f_path.dentry->d_inode;

	/*
	 * BB - optimize the way when signing is disabled. We can drop this
	 * extra memory-to-memory copying and use iovec buffers for constructing
	 * write request.
	 */

	written = cifs_iovec_write(iocb->ki_filp, iov, nr_segs, &pos);
	if (written > 0) {
		CIFS_I(inode)->invalid_mapping = true;
		iocb->ki_pos = pos;
	}

	return written;
}

2464
static ssize_t
2465 2466
cifs_pagecache_writev(struct kiocb *iocb, const struct iovec *iov,
		      unsigned long nr_segs, loff_t pos, bool cache_ex)
2467
{
2468 2469 2470 2471 2472 2473
	struct file *file = iocb->ki_filp;
	struct cifsFileInfo *cfile = (struct cifsFileInfo *)file->private_data;
	struct inode *inode = file->f_mapping->host;
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	struct TCP_Server_Info *server = tlink_tcon(cfile->tlink)->ses->server;
	ssize_t rc = -EACCES;
2474

2475
	BUG_ON(iocb->ki_pos != pos);
2476

2477 2478 2479 2480 2481 2482 2483 2484 2485
	sb_start_write(inode->i_sb);

	/*
	 * We need to hold the sem to be sure nobody modifies lock list
	 * with a brlock that prevents writing.
	 */
	down_read(&cinode->lock_sem);
	if (!cifs_find_lock_conflict(cfile, pos, iov_length(iov, nr_segs),
				     server->vals->exclusive_lock_type, NULL,
2486
				     CIFS_WRITE_OP)) {
2487
		mutex_lock(&inode->i_mutex);
2488 2489
		if (!cache_ex)
			cinode->leave_pages_clean = true;
2490
		rc = __generic_file_aio_write(iocb, iov, nr_segs,
2491 2492 2493
					      &iocb->ki_pos);
		if (!cache_ex)
			cinode->leave_pages_clean = false;
2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
		mutex_unlock(&inode->i_mutex);
	}

	if (rc > 0 || rc == -EIOCBQUEUED) {
		ssize_t err;

		err = generic_write_sync(file, pos, rc);
		if (err < 0 && rc > 0)
			rc = err;
	}

	up_read(&cinode->lock_sem);
	sb_end_write(inode->i_sb);
	return rc;
}

ssize_t
cifs_strict_writev(struct kiocb *iocb, const struct iovec *iov,
		   unsigned long nr_segs, loff_t pos)
{
	struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
	struct cifsFileInfo *cfile = (struct cifsFileInfo *)
						iocb->ki_filp->private_data;
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
2520
	ssize_t written, written2;
2521
	/*
2522 2523 2524 2525 2526 2527 2528 2529
	 * We need to store clientCanCacheAll here to prevent race
	 * conditions - this value can be changed during an execution
	 * of generic_file_aio_write. For CIFS it can be changed from
	 * true to false only, but for SMB2 it can be changed both from
	 * true to false and vice versa. So, we can end up with a data
	 * stored in the cache, not marked dirty and not sent to the
	 * server if this value changes its state from false to true
	 * after cifs_write_end.
2530
	 */
2531 2532 2533 2534
	bool cache_ex = cinode->clientCanCacheAll;
	bool cache_read = cinode->clientCanCacheRead;
	int rc;
	loff_t saved_pos;
2535

2536 2537 2538 2539 2540 2541 2542
	if (cache_ex) {
		if (cap_unix(tcon->ses) &&
		    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0) &&
		    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(
						tcon->fsUnixInfo.Capability)))
			return generic_file_aio_write(iocb, iov, nr_segs, pos);
		return cifs_pagecache_writev(iocb, iov, nr_segs, pos, cache_ex);
2543 2544 2545
	}

	/*
2546 2547 2548 2549 2550
	 * For files without exclusive oplock in strict cache mode we need to
	 * write the data to the server exactly from the pos to pos+len-1 rather
	 * than flush all affected pages because it may cause a error with
	 * mandatory locks on these pages but not on the region from pos to
	 * ppos+len-1.
2551
	 */
2552 2553 2554
	written = cifs_user_writev(iocb, iov, nr_segs, pos);
	if (!cache_read || written <= 0)
		return written;
2555

2556 2557 2558
	saved_pos = iocb->ki_pos;
	iocb->ki_pos = pos;
	/* we have a read oplock - need to store a data in the page cache */
2559
	if (cap_unix(tcon->ses) &&
2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(
					tcon->fsUnixInfo.Capability)))
		written2 = generic_file_aio_write(iocb, iov, nr_segs, pos);
	else
		written2 = cifs_pagecache_writev(iocb, iov, nr_segs, pos,
						 cache_ex);
	/* errors occured during writing - invalidate the page cache */
	if (written2 < 0) {
		rc = cifs_invalidate_mapping(inode);
		if (rc)
			written = (ssize_t)rc;
		else
			iocb->ki_pos = saved_pos;
	}
	return written;
2576 2577
}

2578
static struct cifs_readdata *
2579
cifs_readdata_alloc(unsigned int nr_pages, work_func_t complete)
2580 2581
{
	struct cifs_readdata *rdata;
2582

2583 2584
	rdata = kzalloc(sizeof(*rdata) + (sizeof(struct page *) * nr_pages),
			GFP_KERNEL);
2585
	if (rdata != NULL) {
2586
		kref_init(&rdata->refcount);
2587 2588
		INIT_LIST_HEAD(&rdata->list);
		init_completion(&rdata->done);
2589 2590
		INIT_WORK(&rdata->work, complete);
	}
2591

2592 2593 2594
	return rdata;
}

2595 2596
void
cifs_readdata_release(struct kref *refcount)
2597
{
2598 2599 2600 2601 2602 2603
	struct cifs_readdata *rdata = container_of(refcount,
					struct cifs_readdata, refcount);

	if (rdata->cfile)
		cifsFileInfo_put(rdata->cfile);

2604 2605 2606
	kfree(rdata);
}

2607
static int
2608
cifs_read_allocate_pages(struct cifs_readdata *rdata, unsigned int nr_pages)
2609 2610
{
	int rc = 0;
2611
	struct page *page;
2612 2613
	unsigned int i;

2614
	for (i = 0; i < nr_pages; i++) {
2615 2616 2617 2618 2619
		page = alloc_page(GFP_KERNEL|__GFP_HIGHMEM);
		if (!page) {
			rc = -ENOMEM;
			break;
		}
2620
		rdata->pages[i] = page;
2621 2622 2623
	}

	if (rc) {
2624 2625 2626
		for (i = 0; i < nr_pages; i++) {
			put_page(rdata->pages[i]);
			rdata->pages[i] = NULL;
2627 2628 2629 2630 2631 2632 2633 2634 2635 2636
		}
	}
	return rc;
}

static void
cifs_uncached_readdata_release(struct kref *refcount)
{
	struct cifs_readdata *rdata = container_of(refcount,
					struct cifs_readdata, refcount);
2637
	unsigned int i;
2638

2639 2640 2641
	for (i = 0; i < rdata->nr_pages; i++) {
		put_page(rdata->pages[i]);
		rdata->pages[i] = NULL;
2642 2643 2644 2645
	}
	cifs_readdata_release(refcount);
}

2646 2647 2648 2649
static int
cifs_retry_async_readv(struct cifs_readdata *rdata)
{
	int rc;
2650 2651 2652
	struct TCP_Server_Info *server;

	server = tlink_tcon(rdata->cfile->tlink)->ses->server;
2653 2654 2655 2656 2657 2658 2659

	do {
		if (rdata->cfile->invalidHandle) {
			rc = cifs_reopen_file(rdata->cfile, true);
			if (rc != 0)
				continue;
		}
2660
		rc = server->ops->async_readv(rdata);
2661 2662 2663 2664 2665
	} while (rc == -EAGAIN);

	return rc;
}

2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686
/**
 * cifs_readdata_to_iov - copy data from pages in response to an iovec
 * @rdata:	the readdata response with list of pages holding data
 * @iov:	vector in which we should copy the data
 * @nr_segs:	number of segments in vector
 * @offset:	offset into file of the first iovec
 * @copied:	used to return the amount of data copied to the iov
 *
 * This function copies data from a list of pages in a readdata response into
 * an array of iovecs. It will first calculate where the data should go
 * based on the info in the readdata and then copy the data into that spot.
 */
static ssize_t
cifs_readdata_to_iov(struct cifs_readdata *rdata, const struct iovec *iov,
			unsigned long nr_segs, loff_t offset, ssize_t *copied)
{
	int rc = 0;
	struct iov_iter ii;
	size_t pos = rdata->offset - offset;
	ssize_t remaining = rdata->bytes;
	unsigned char *pdata;
2687
	unsigned int i;
2688 2689 2690 2691 2692 2693

	/* set up iov_iter and advance to the correct offset */
	iov_iter_init(&ii, iov, nr_segs, iov_length(iov, nr_segs), 0);
	iov_iter_advance(&ii, pos);

	*copied = 0;
2694
	for (i = 0; i < rdata->nr_pages; i++) {
2695
		ssize_t copy;
2696
		struct page *page = rdata->pages[i];
2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731

		/* copy a whole page or whatever's left */
		copy = min_t(ssize_t, remaining, PAGE_SIZE);

		/* ...but limit it to whatever space is left in the iov */
		copy = min_t(ssize_t, copy, iov_iter_count(&ii));

		/* go while there's data to be copied and no errors */
		if (copy && !rc) {
			pdata = kmap(page);
			rc = memcpy_toiovecend(ii.iov, pdata, ii.iov_offset,
						(int)copy);
			kunmap(page);
			if (!rc) {
				*copied += copy;
				remaining -= copy;
				iov_iter_advance(&ii, copy);
			}
		}
	}

	return rc;
}

static void
cifs_uncached_readv_complete(struct work_struct *work)
{
	struct cifs_readdata *rdata = container_of(work,
						struct cifs_readdata, work);

	complete(&rdata->done);
	kref_put(&rdata->refcount, cifs_uncached_readdata_release);
}

static int
2732 2733
cifs_uncached_read_into_pages(struct TCP_Server_Info *server,
			struct cifs_readdata *rdata, unsigned int len)
2734
{
2735
	int total_read = 0, result = 0;
2736 2737
	unsigned int i;
	unsigned int nr_pages = rdata->nr_pages;
2738
	struct kvec iov;
2739

2740
	rdata->tailsz = PAGE_SIZE;
2741 2742 2743
	for (i = 0; i < nr_pages; i++) {
		struct page *page = rdata->pages[i];

2744
		if (len >= PAGE_SIZE) {
2745
			/* enough data to fill the page */
2746 2747 2748 2749 2750 2751
			iov.iov_base = kmap(page);
			iov.iov_len = PAGE_SIZE;
			cFYI(1, "%u: iov_base=%p iov_len=%zu",
				i, iov.iov_base, iov.iov_len);
			len -= PAGE_SIZE;
		} else if (len > 0) {
2752
			/* enough for partial page, fill and zero the rest */
2753 2754 2755 2756 2757 2758 2759
			iov.iov_base = kmap(page);
			iov.iov_len = len;
			cFYI(1, "%u: iov_base=%p iov_len=%zu",
				i, iov.iov_base, iov.iov_len);
			memset(iov.iov_base + len, '\0', PAGE_SIZE - len);
			rdata->tailsz = len;
			len = 0;
2760 2761
		} else {
			/* no need to hold page hostage */
2762 2763
			rdata->pages[i] = NULL;
			rdata->nr_pages--;
2764
			put_page(page);
2765
			continue;
2766
		}
2767 2768 2769 2770 2771 2772 2773

		result = cifs_readv_from_socket(server, &iov, 1, iov.iov_len);
		kunmap(page);
		if (result < 0)
			break;

		total_read += result;
2774 2775
	}

2776
	return total_read > 0 ? total_read : result;
2777 2778
}

2779 2780 2781
static ssize_t
cifs_iovec_read(struct file *file, const struct iovec *iov,
		 unsigned long nr_segs, loff_t *poffset)
L
Linus Torvalds 已提交
2782
{
2783
	ssize_t rc;
2784
	size_t len, cur_len;
2785 2786 2787
	ssize_t total_read = 0;
	loff_t offset = *poffset;
	unsigned int npages;
L
Linus Torvalds 已提交
2788
	struct cifs_sb_info *cifs_sb;
2789
	struct cifs_tcon *tcon;
L
Linus Torvalds 已提交
2790
	struct cifsFileInfo *open_file;
2791 2792 2793
	struct cifs_readdata *rdata, *tmp;
	struct list_head rdata_list;
	pid_t pid;
2794 2795 2796 2797 2798 2799 2800

	if (!nr_segs)
		return 0;

	len = iov_length(iov, nr_segs);
	if (!len)
		return 0;
L
Linus Torvalds 已提交
2801

2802
	INIT_LIST_HEAD(&rdata_list);
2803
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
2804
	open_file = file->private_data;
2805
	tcon = tlink_tcon(open_file->tlink);
L
Linus Torvalds 已提交
2806

2807 2808 2809
	if (!tcon->ses->server->ops->async_readv)
		return -ENOSYS;

2810 2811 2812 2813 2814
	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
		pid = open_file->pid;
	else
		pid = current->tgid;

S
Steve French 已提交
2815
	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
2816
		cFYI(1, "attempting read on write only file instance");
S
Steve French 已提交
2817

2818 2819 2820
	do {
		cur_len = min_t(const size_t, len - total_read, cifs_sb->rsize);
		npages = DIV_ROUND_UP(cur_len, PAGE_SIZE);
2821

2822 2823 2824 2825 2826 2827
		/* allocate a readdata struct */
		rdata = cifs_readdata_alloc(npages,
					    cifs_uncached_readv_complete);
		if (!rdata) {
			rc = -ENOMEM;
			goto error;
L
Linus Torvalds 已提交
2828
		}
2829

2830
		rc = cifs_read_allocate_pages(rdata, npages);
2831 2832 2833 2834
		if (rc)
			goto error;

		rdata->cfile = cifsFileInfo_get(open_file);
2835
		rdata->nr_pages = npages;
2836 2837 2838
		rdata->offset = offset;
		rdata->bytes = cur_len;
		rdata->pid = pid;
2839 2840
		rdata->pagesz = PAGE_SIZE;
		rdata->read_into_pages = cifs_uncached_read_into_pages;
2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 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 2880 2881

		rc = cifs_retry_async_readv(rdata);
error:
		if (rc) {
			kref_put(&rdata->refcount,
				 cifs_uncached_readdata_release);
			break;
		}

		list_add_tail(&rdata->list, &rdata_list);
		offset += cur_len;
		len -= cur_len;
	} while (len > 0);

	/* if at least one read request send succeeded, then reset rc */
	if (!list_empty(&rdata_list))
		rc = 0;

	/* the loop below should proceed in the order of increasing offsets */
restart_loop:
	list_for_each_entry_safe(rdata, tmp, &rdata_list, list) {
		if (!rc) {
			ssize_t copied;

			/* FIXME: freezable sleep too? */
			rc = wait_for_completion_killable(&rdata->done);
			if (rc)
				rc = -EINTR;
			else if (rdata->result)
				rc = rdata->result;
			else {
				rc = cifs_readdata_to_iov(rdata, iov,
							nr_segs, *poffset,
							&copied);
				total_read += copied;
			}

			/* resend call if it's a retryable error */
			if (rc == -EAGAIN) {
				rc = cifs_retry_async_readv(rdata);
				goto restart_loop;
L
Linus Torvalds 已提交
2882 2883
			}
		}
2884 2885
		list_del_init(&rdata->list);
		kref_put(&rdata->refcount, cifs_uncached_readdata_release);
L
Linus Torvalds 已提交
2886
	}
2887

2888 2889 2890
	cifs_stats_bytes_read(tcon, total_read);
	*poffset += total_read;

2891 2892 2893 2894
	/* mask nodata case */
	if (rc == -ENODATA)
		rc = 0;

2895
	return total_read ? total_read : rc;
L
Linus Torvalds 已提交
2896 2897
}

2898
ssize_t cifs_user_readv(struct kiocb *iocb, const struct iovec *iov,
2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909
			       unsigned long nr_segs, loff_t pos)
{
	ssize_t read;

	read = cifs_iovec_read(iocb->ki_filp, iov, nr_segs, &pos);
	if (read > 0)
		iocb->ki_pos = pos;

	return read;
}

2910 2911 2912
ssize_t
cifs_strict_readv(struct kiocb *iocb, const struct iovec *iov,
		  unsigned long nr_segs, loff_t pos)
2913
{
2914 2915 2916 2917 2918 2919 2920
	struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
	struct cifsFileInfo *cfile = (struct cifsFileInfo *)
						iocb->ki_filp->private_data;
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
	int rc = -EACCES;
2921 2922 2923 2924 2925 2926 2927 2928 2929

	/*
	 * In strict cache mode we need to read from the server all the time
	 * if we don't have level II oplock because the server can delay mtime
	 * change - so we can't make a decision about inode invalidating.
	 * And we can also fail with pagereading if there are mandatory locks
	 * on pages affected by this read but not on the region from pos to
	 * pos+len-1.
	 */
2930 2931
	if (!cinode->clientCanCacheRead)
		return cifs_user_readv(iocb, iov, nr_segs, pos);
2932

2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944
	if (cap_unix(tcon->ses) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
		return generic_file_aio_read(iocb, iov, nr_segs, pos);

	/*
	 * We need to hold the sem to be sure nobody modifies lock list
	 * with a brlock that prevents reading.
	 */
	down_read(&cinode->lock_sem);
	if (!cifs_find_lock_conflict(cfile, pos, iov_length(iov, nr_segs),
				     tcon->ses->server->vals->shared_lock_type,
2945
				     NULL, CIFS_READ_OP))
2946 2947 2948
		rc = generic_file_aio_read(iocb, iov, nr_segs, pos);
	up_read(&cinode->lock_sem);
	return rc;
2949
}
L
Linus Torvalds 已提交
2950

2951 2952
static ssize_t
cifs_read(struct file *file, char *read_data, size_t read_size, loff_t *offset)
L
Linus Torvalds 已提交
2953 2954 2955 2956 2957
{
	int rc = -EACCES;
	unsigned int bytes_read = 0;
	unsigned int total_read;
	unsigned int current_read_size;
2958
	unsigned int rsize;
L
Linus Torvalds 已提交
2959
	struct cifs_sb_info *cifs_sb;
2960
	struct cifs_tcon *tcon;
2961
	struct TCP_Server_Info *server;
2962
	unsigned int xid;
2963
	char *cur_offset;
L
Linus Torvalds 已提交
2964
	struct cifsFileInfo *open_file;
2965
	struct cifs_io_parms io_parms;
2966
	int buf_type = CIFS_NO_BUFFER;
2967
	__u32 pid;
L
Linus Torvalds 已提交
2968

2969
	xid = get_xid();
2970
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
2971

2972 2973 2974
	/* FIXME: set up handlers for larger reads and/or convert to async */
	rsize = min_t(unsigned int, cifs_sb->rsize, CIFSMaxBufSize);

L
Linus Torvalds 已提交
2975
	if (file->private_data == NULL) {
2976
		rc = -EBADF;
2977
		free_xid(xid);
2978
		return rc;
L
Linus Torvalds 已提交
2979
	}
2980
	open_file = file->private_data;
2981
	tcon = tlink_tcon(open_file->tlink);
2982 2983 2984 2985 2986 2987
	server = tcon->ses->server;

	if (!server->ops->sync_read) {
		free_xid(xid);
		return -ENOSYS;
	}
L
Linus Torvalds 已提交
2988

2989 2990 2991 2992 2993
	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
		pid = open_file->pid;
	else
		pid = current->tgid;

L
Linus Torvalds 已提交
2994
	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
2995
		cFYI(1, "attempting read on write only file instance");
L
Linus Torvalds 已提交
2996

2997 2998
	for (total_read = 0, cur_offset = read_data; read_size > total_read;
	     total_read += bytes_read, cur_offset += bytes_read) {
2999
		current_read_size = min_t(uint, read_size - total_read, rsize);
3000 3001 3002 3003 3004 3005
		/*
		 * For windows me and 9x we do not want to request more than it
		 * negotiated since it will refuse the read then.
		 */
		if ((tcon->ses) && !(tcon->ses->capabilities &
				tcon->ses->server->vals->cap_large_files)) {
3006
			current_read_size = min_t(uint, current_read_size,
3007
					CIFSMaxBufSize);
3008
		}
L
Linus Torvalds 已提交
3009 3010
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
3011
			if (open_file->invalidHandle) {
J
Jeff Layton 已提交
3012
				rc = cifs_reopen_file(open_file, true);
L
Linus Torvalds 已提交
3013 3014 3015
				if (rc != 0)
					break;
			}
3016
			io_parms.pid = pid;
3017
			io_parms.tcon = tcon;
3018
			io_parms.offset = *offset;
3019
			io_parms.length = current_read_size;
3020 3021 3022
			rc = server->ops->sync_read(xid, open_file, &io_parms,
						    &bytes_read, &cur_offset,
						    &buf_type);
L
Linus Torvalds 已提交
3023 3024 3025 3026 3027
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
3028
				free_xid(xid);
L
Linus Torvalds 已提交
3029 3030 3031
				return rc;
			}
		} else {
3032
			cifs_stats_bytes_read(tcon, total_read);
3033
			*offset += bytes_read;
L
Linus Torvalds 已提交
3034 3035
		}
	}
3036
	free_xid(xid);
L
Linus Torvalds 已提交
3037 3038 3039
	return total_read;
}

3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055
/*
 * If the page is mmap'ed into a process' page tables, then we need to make
 * sure that it doesn't change while being written back.
 */
static int
cifs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
{
	struct page *page = vmf->page;

	lock_page(page);
	return VM_FAULT_LOCKED;
}

static struct vm_operations_struct cifs_file_vm_ops = {
	.fault = filemap_fault,
	.page_mkwrite = cifs_page_mkwrite,
3056
	.remap_pages = generic_file_remap_pages,
3057 3058
};

3059 3060 3061 3062 3063
int cifs_file_strict_mmap(struct file *file, struct vm_area_struct *vma)
{
	int rc, xid;
	struct inode *inode = file->f_path.dentry->d_inode;

3064
	xid = get_xid();
3065

3066 3067 3068 3069 3070
	if (!CIFS_I(inode)->clientCanCacheRead) {
		rc = cifs_invalidate_mapping(inode);
		if (rc)
			return rc;
	}
3071 3072

	rc = generic_file_mmap(file, vma);
3073 3074
	if (rc == 0)
		vma->vm_ops = &cifs_file_vm_ops;
3075
	free_xid(xid);
3076 3077 3078
	return rc;
}

L
Linus Torvalds 已提交
3079 3080 3081 3082
int cifs_file_mmap(struct file *file, struct vm_area_struct *vma)
{
	int rc, xid;

3083
	xid = get_xid();
J
Jeff Layton 已提交
3084
	rc = cifs_revalidate_file(file);
L
Linus Torvalds 已提交
3085
	if (rc) {
3086
		cFYI(1, "Validation prior to mmap failed, error=%d", rc);
3087
		free_xid(xid);
L
Linus Torvalds 已提交
3088 3089 3090
		return rc;
	}
	rc = generic_file_mmap(file, vma);
3091 3092
	if (rc == 0)
		vma->vm_ops = &cifs_file_vm_ops;
3093
	free_xid(xid);
L
Linus Torvalds 已提交
3094 3095 3096
	return rc;
}

3097 3098 3099
static void
cifs_readv_complete(struct work_struct *work)
{
3100
	unsigned int i;
3101 3102 3103
	struct cifs_readdata *rdata = container_of(work,
						struct cifs_readdata, work);

3104 3105 3106
	for (i = 0; i < rdata->nr_pages; i++) {
		struct page *page = rdata->pages[i];

3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119
		lru_cache_add_file(page);

		if (rdata->result == 0) {
			flush_dcache_page(page);
			SetPageUptodate(page);
		}

		unlock_page(page);

		if (rdata->result == 0)
			cifs_readpage_to_fscache(rdata->mapping->host, page);

		page_cache_release(page);
3120
		rdata->pages[i] = NULL;
3121
	}
3122
	kref_put(&rdata->refcount, cifs_readdata_release);
3123 3124
}

3125
static int
3126 3127
cifs_readpages_read_into_pages(struct TCP_Server_Info *server,
			struct cifs_readdata *rdata, unsigned int len)
3128
{
3129
	int total_read = 0, result = 0;
3130
	unsigned int i;
3131 3132
	u64 eof;
	pgoff_t eof_index;
3133
	unsigned int nr_pages = rdata->nr_pages;
3134
	struct kvec iov;
3135 3136 3137 3138 3139 3140

	/* determine the eof that the server (probably) has */
	eof = CIFS_I(rdata->mapping->host)->server_eof;
	eof_index = eof ? (eof - 1) >> PAGE_CACHE_SHIFT : 0;
	cFYI(1, "eof=%llu eof_index=%lu", eof, eof_index);

3141
	rdata->tailsz = PAGE_CACHE_SIZE;
3142 3143 3144
	for (i = 0; i < nr_pages; i++) {
		struct page *page = rdata->pages[i];

3145
		if (len >= PAGE_CACHE_SIZE) {
3146
			/* enough data to fill the page */
3147 3148
			iov.iov_base = kmap(page);
			iov.iov_len = PAGE_CACHE_SIZE;
3149
			cFYI(1, "%u: idx=%lu iov_base=%p iov_len=%zu",
3150 3151 3152
				i, page->index, iov.iov_base, iov.iov_len);
			len -= PAGE_CACHE_SIZE;
		} else if (len > 0) {
3153
			/* enough for partial page, fill and zero the rest */
3154 3155
			iov.iov_base = kmap(page);
			iov.iov_len = len;
3156
			cFYI(1, "%u: idx=%lu iov_base=%p iov_len=%zu",
3157 3158 3159 3160 3161
				i, page->index, iov.iov_base, iov.iov_len);
			memset(iov.iov_base + len,
				'\0', PAGE_CACHE_SIZE - len);
			rdata->tailsz = len;
			len = 0;
3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176
		} else if (page->index > eof_index) {
			/*
			 * The VFS will not try to do readahead past the
			 * i_size, but it's possible that we have outstanding
			 * writes with gaps in the middle and the i_size hasn't
			 * caught up yet. Populate those with zeroed out pages
			 * to prevent the VFS from repeatedly attempting to
			 * fill them until the writes are flushed.
			 */
			zero_user(page, 0, PAGE_CACHE_SIZE);
			lru_cache_add_file(page);
			flush_dcache_page(page);
			SetPageUptodate(page);
			unlock_page(page);
			page_cache_release(page);
3177 3178
			rdata->pages[i] = NULL;
			rdata->nr_pages--;
3179
			continue;
3180 3181 3182 3183 3184
		} else {
			/* no need to hold page hostage */
			lru_cache_add_file(page);
			unlock_page(page);
			page_cache_release(page);
3185 3186
			rdata->pages[i] = NULL;
			rdata->nr_pages--;
3187
			continue;
3188
		}
3189 3190 3191 3192 3193 3194 3195

		result = cifs_readv_from_socket(server, &iov, 1, iov.iov_len);
		kunmap(page);
		if (result < 0)
			break;

		total_read += result;
3196 3197
	}

3198
	return total_read > 0 ? total_read : result;
3199 3200
}

L
Linus Torvalds 已提交
3201 3202 3203
static int cifs_readpages(struct file *file, struct address_space *mapping,
	struct list_head *page_list, unsigned num_pages)
{
3204 3205 3206 3207 3208 3209
	int rc;
	struct list_head tmplist;
	struct cifsFileInfo *open_file = file->private_data;
	struct cifs_sb_info *cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
	unsigned int rsize = cifs_sb->rsize;
	pid_t pid;
L
Linus Torvalds 已提交
3210

3211 3212 3213 3214 3215 3216 3217 3218
	/*
	 * Give up immediately if rsize is too small to read an entire page.
	 * The VFS will fall back to readpage. We should never reach this
	 * point however since we set ra_pages to 0 when the rsize is smaller
	 * than a cache page.
	 */
	if (unlikely(rsize < PAGE_CACHE_SIZE))
		return 0;
3219

3220 3221 3222 3223 3224 3225 3226
	/*
	 * Reads as many pages as possible from fscache. Returns -ENOBUFS
	 * immediately if the cookie is negative
	 */
	rc = cifs_readpages_from_fscache(mapping->host, mapping, page_list,
					 &num_pages);
	if (rc == 0)
3227
		return rc;
3228

3229 3230 3231 3232 3233
	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_RWPIDFORWARD)
		pid = open_file->pid;
	else
		pid = current->tgid;

3234 3235
	rc = 0;
	INIT_LIST_HEAD(&tmplist);
L
Linus Torvalds 已提交
3236

3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251
	cFYI(1, "%s: file=%p mapping=%p num_pages=%u", __func__, file,
		mapping, num_pages);

	/*
	 * Start with the page at end of list and move it to private
	 * list. Do the same with any following pages until we hit
	 * the rsize limit, hit an index discontinuity, or run out of
	 * pages. Issue the async read and then start the loop again
	 * until the list is empty.
	 *
	 * Note that list order is important. The page_list is in
	 * the order of declining indexes. When we put the pages in
	 * the rdata->pages, then we want them in increasing order.
	 */
	while (!list_empty(page_list)) {
3252
		unsigned int i;
3253 3254 3255 3256 3257 3258
		unsigned int bytes = PAGE_CACHE_SIZE;
		unsigned int expected_index;
		unsigned int nr_pages = 1;
		loff_t offset;
		struct page *page, *tpage;
		struct cifs_readdata *rdata;
L
Linus Torvalds 已提交
3259 3260

		page = list_entry(page_list->prev, struct page, lru);
3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277

		/*
		 * Lock the page and put it in the cache. Since no one else
		 * should have access to this page, we're safe to simply set
		 * PG_locked without checking it first.
		 */
		__set_page_locked(page);
		rc = add_to_page_cache_locked(page, mapping,
					      page->index, GFP_KERNEL);

		/* give up if we can't stick it in the cache */
		if (rc) {
			__clear_page_locked(page);
			break;
		}

		/* move first page to the tmplist */
L
Linus Torvalds 已提交
3278
		offset = (loff_t)page->index << PAGE_CACHE_SHIFT;
3279
		list_move_tail(&page->lru, &tmplist);
L
Linus Torvalds 已提交
3280

3281 3282 3283 3284 3285
		/* now try and add more pages onto the request */
		expected_index = page->index + 1;
		list_for_each_entry_safe_reverse(page, tpage, page_list, lru) {
			/* discontinuity ? */
			if (page->index != expected_index)
S
Steve French 已提交
3286
				break;
3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301

			/* would this page push the read over the rsize? */
			if (bytes + PAGE_CACHE_SIZE > rsize)
				break;

			__set_page_locked(page);
			if (add_to_page_cache_locked(page, mapping,
						page->index, GFP_KERNEL)) {
				__clear_page_locked(page);
				break;
			}
			list_move_tail(&page->lru, &tmplist);
			bytes += PAGE_CACHE_SIZE;
			expected_index++;
			nr_pages++;
L
Linus Torvalds 已提交
3302
		}
3303

3304
		rdata = cifs_readdata_alloc(nr_pages, cifs_readv_complete);
3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316
		if (!rdata) {
			/* best to give up if we're out of mem */
			list_for_each_entry_safe(page, tpage, &tmplist, lru) {
				list_del(&page->lru);
				lru_cache_add_file(page);
				unlock_page(page);
				page_cache_release(page);
			}
			rc = -ENOMEM;
			break;
		}

3317
		rdata->cfile = cifsFileInfo_get(open_file);
3318 3319 3320 3321
		rdata->mapping = mapping;
		rdata->offset = offset;
		rdata->bytes = bytes;
		rdata->pid = pid;
3322 3323
		rdata->pagesz = PAGE_CACHE_SIZE;
		rdata->read_into_pages = cifs_readpages_read_into_pages;
3324 3325 3326 3327 3328

		list_for_each_entry_safe(page, tpage, &tmplist, lru) {
			list_del(&page->lru);
			rdata->pages[rdata->nr_pages++] = page;
		}
3329

3330
		rc = cifs_retry_async_readv(rdata);
3331
		if (rc != 0) {
3332 3333
			for (i = 0; i < rdata->nr_pages; i++) {
				page = rdata->pages[i];
3334 3335 3336
				lru_cache_add_file(page);
				unlock_page(page);
				page_cache_release(page);
L
Linus Torvalds 已提交
3337
			}
3338
			kref_put(&rdata->refcount, cifs_readdata_release);
L
Linus Torvalds 已提交
3339 3340
			break;
		}
3341 3342

		kref_put(&rdata->refcount, cifs_readdata_release);
L
Linus Torvalds 已提交
3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353
	}

	return rc;
}

static int cifs_readpage_worker(struct file *file, struct page *page,
	loff_t *poffset)
{
	char *read_data;
	int rc;

3354 3355 3356 3357 3358
	/* Is the page cached? */
	rc = cifs_readpage_from_fscache(file->f_path.dentry->d_inode, page);
	if (rc == 0)
		goto read_complete;

L
Linus Torvalds 已提交
3359 3360 3361
	page_cache_get(page);
	read_data = kmap(page);
	/* for reads over a certain size could initiate async read ahead */
S
Steve French 已提交
3362

L
Linus Torvalds 已提交
3363
	rc = cifs_read(file, read_data, PAGE_CACHE_SIZE, poffset);
S
Steve French 已提交
3364

L
Linus Torvalds 已提交
3365 3366 3367
	if (rc < 0)
		goto io_error;
	else
3368
		cFYI(1, "Bytes read %d", rc);
S
Steve French 已提交
3369

3370 3371
	file->f_path.dentry->d_inode->i_atime =
		current_fs_time(file->f_path.dentry->d_inode->i_sb);
S
Steve French 已提交
3372

L
Linus Torvalds 已提交
3373 3374 3375 3376 3377
	if (PAGE_CACHE_SIZE > rc)
		memset(read_data + rc, 0, PAGE_CACHE_SIZE - rc);

	flush_dcache_page(page);
	SetPageUptodate(page);
3378 3379 3380 3381

	/* send this page to the cache */
	cifs_readpage_to_fscache(file->f_path.dentry->d_inode, page);

L
Linus Torvalds 已提交
3382
	rc = 0;
S
Steve French 已提交
3383

L
Linus Torvalds 已提交
3384
io_error:
S
Steve French 已提交
3385
	kunmap(page);
L
Linus Torvalds 已提交
3386
	page_cache_release(page);
3387 3388

read_complete:
L
Linus Torvalds 已提交
3389 3390 3391 3392 3393 3394 3395
	return rc;
}

static int cifs_readpage(struct file *file, struct page *page)
{
	loff_t offset = (loff_t)page->index << PAGE_CACHE_SHIFT;
	int rc = -EACCES;
3396
	unsigned int xid;
L
Linus Torvalds 已提交
3397

3398
	xid = get_xid();
L
Linus Torvalds 已提交
3399 3400

	if (file->private_data == NULL) {
3401
		rc = -EBADF;
3402
		free_xid(xid);
3403
		return rc;
L
Linus Torvalds 已提交
3404 3405
	}

3406
	cFYI(1, "readpage %p at offset %d 0x%x",
3407
		 page, (int)offset, (int)offset);
L
Linus Torvalds 已提交
3408 3409 3410 3411 3412

	rc = cifs_readpage_worker(file, page, &offset);

	unlock_page(page);

3413
	free_xid(xid);
L
Linus Torvalds 已提交
3414 3415 3416
	return rc;
}

3417 3418 3419 3420
static int is_inode_writable(struct cifsInodeInfo *cifs_inode)
{
	struct cifsFileInfo *open_file;

3421
	spin_lock(&cifs_file_list_lock);
3422
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
3423
		if (OPEN_FMODE(open_file->f_flags) & FMODE_WRITE) {
3424
			spin_unlock(&cifs_file_list_lock);
3425 3426 3427
			return 1;
		}
	}
3428
	spin_unlock(&cifs_file_list_lock);
3429 3430 3431
	return 0;
}

L
Linus Torvalds 已提交
3432 3433 3434
/* We do not want to update the file size from server for inodes
   open for write - to avoid races with writepage extending
   the file - in the future we could consider allowing
S
Steve French 已提交
3435
   refreshing the inode only on increases in the file size
L
Linus Torvalds 已提交
3436 3437
   but this is tricky to do without racing with writebehind
   page caching in the current Linux kernel design */
3438
bool is_size_safe_to_change(struct cifsInodeInfo *cifsInode, __u64 end_of_file)
L
Linus Torvalds 已提交
3439
{
3440
	if (!cifsInode)
3441
		return true;
3442

3443 3444
	if (is_inode_writable(cifsInode)) {
		/* This inode is open for write at least once */
3445 3446 3447
		struct cifs_sb_info *cifs_sb;

		cifs_sb = CIFS_SB(cifsInode->vfs_inode.i_sb);
S
Steve French 已提交
3448
		if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DIRECT_IO) {
S
Steve French 已提交
3449
			/* since no page cache to corrupt on directio
3450
			we can change size safely */
3451
			return true;
3452 3453
		}

S
Steve French 已提交
3454
		if (i_size_read(&cifsInode->vfs_inode) < end_of_file)
3455
			return true;
3456

3457
		return false;
3458
	} else
3459
		return true;
L
Linus Torvalds 已提交
3460 3461
}

N
Nick Piggin 已提交
3462 3463 3464
static int cifs_write_begin(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned flags,
			struct page **pagep, void **fsdata)
L
Linus Torvalds 已提交
3465
{
N
Nick Piggin 已提交
3466 3467
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	loff_t offset = pos & (PAGE_CACHE_SIZE - 1);
3468 3469 3470 3471
	loff_t page_start = pos & PAGE_MASK;
	loff_t i_size;
	struct page *page;
	int rc = 0;
N
Nick Piggin 已提交
3472

3473
	cFYI(1, "write_begin from %lld len %d", (long long)pos, len);
N
Nick Piggin 已提交
3474

3475
	page = grab_cache_page_write_begin(mapping, index, flags);
3476 3477 3478 3479
	if (!page) {
		rc = -ENOMEM;
		goto out;
	}
3480

3481 3482
	if (PageUptodate(page))
		goto out;
3483

3484 3485 3486 3487 3488 3489 3490
	/*
	 * If we write a full page it will be up to date, no need to read from
	 * the server. If the write is short, we'll end up doing a sync write
	 * instead.
	 */
	if (len == PAGE_CACHE_SIZE)
		goto out;
3491

3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514
	/*
	 * optimize away the read when we have an oplock, and we're not
	 * expecting to use any of the data we'd be reading in. That
	 * is, when the page lies beyond the EOF, or straddles the EOF
	 * and the write will cover all of the existing data.
	 */
	if (CIFS_I(mapping->host)->clientCanCacheRead) {
		i_size = i_size_read(mapping->host);
		if (page_start >= i_size ||
		    (offset == 0 && (pos + len) >= i_size)) {
			zero_user_segments(page, 0, offset,
					   offset + len,
					   PAGE_CACHE_SIZE);
			/*
			 * PageChecked means that the parts of the page
			 * to which we're not writing are considered up
			 * to date. Once the data is copied to the
			 * page, it can be set uptodate.
			 */
			SetPageChecked(page);
			goto out;
		}
	}
N
Nick Piggin 已提交
3515

3516 3517 3518 3519 3520 3521 3522
	if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
		/*
		 * might as well read a page, it is fast enough. If we get
		 * an error, we don't need to return it. cifs_write_end will
		 * do a sync write instead since PG_uptodate isn't set.
		 */
		cifs_readpage_worker(file, page, &page_start);
3523 3524 3525 3526
	} else {
		/* we could try using another file handle if there is one -
		   but how would we lock it to prevent close of that handle
		   racing with this read? In any case
N
Nick Piggin 已提交
3527
		   this will be written out by write_end so is fine */
L
Linus Torvalds 已提交
3528
	}
3529 3530 3531
out:
	*pagep = page;
	return rc;
L
Linus Torvalds 已提交
3532 3533
}

3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549
static int cifs_release_page(struct page *page, gfp_t gfp)
{
	if (PagePrivate(page))
		return 0;

	return cifs_fscache_release_page(page, gfp);
}

static void cifs_invalidate_page(struct page *page, unsigned long offset)
{
	struct cifsInodeInfo *cifsi = CIFS_I(page->mapping->host);

	if (offset == 0)
		cifs_fscache_invalidate_page(page, &cifsi->vfs_inode);
}

3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
static int cifs_launder_page(struct page *page)
{
	int rc = 0;
	loff_t range_start = page_offset(page);
	loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = 0,
		.range_start = range_start,
		.range_end = range_end,
	};

	cFYI(1, "Launder page: %p", page);

	if (clear_page_dirty_for_io(page))
		rc = cifs_writepage_locked(page, &wbc);

	cifs_fscache_invalidate_page(page, page->mapping->host);
	return rc;
}

3571
void cifs_oplock_break(struct work_struct *work)
3572 3573 3574
{
	struct cifsFileInfo *cfile = container_of(work, struct cifsFileInfo,
						  oplock_break);
3575
	struct inode *inode = cfile->dentry->d_inode;
3576
	struct cifsInodeInfo *cinode = CIFS_I(inode);
3577
	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
3578
	int rc = 0;
3579 3580

	if (inode && S_ISREG(inode->i_mode)) {
S
Steve French 已提交
3581
		if (cinode->clientCanCacheRead)
3582
			break_lease(inode, O_RDONLY);
S
Steve French 已提交
3583
		else
3584
			break_lease(inode, O_WRONLY);
3585 3586
		rc = filemap_fdatawrite(inode->i_mapping);
		if (cinode->clientCanCacheRead == 0) {
3587 3588
			rc = filemap_fdatawait(inode->i_mapping);
			mapping_set_error(inode->i_mapping, rc);
3589
			cifs_invalidate_mapping(inode);
3590
		}
3591
		cFYI(1, "Oplock flush inode %p rc %d", inode, rc);
3592 3593
	}

3594 3595 3596 3597
	rc = cifs_push_locks(cfile);
	if (rc)
		cERROR(1, "Push locks rc = %d", rc);

3598 3599 3600 3601 3602 3603
	/*
	 * releasing stale oplock after recent reconnect of smb session using
	 * a now incorrect file handle is not a data integrity issue but do
	 * not bother sending an oplock release if session to server still is
	 * disconnected since oplock already released by the server
	 */
3604
	if (!cfile->oplock_break_cancelled) {
3605 3606
		rc = tcon->ses->server->ops->oplock_response(tcon, &cfile->fid,
							     cinode);
3607
		cFYI(1, "Oplock release rc = %d", rc);
3608 3609 3610
	}
}

3611
const struct address_space_operations cifs_addr_ops = {
L
Linus Torvalds 已提交
3612 3613 3614
	.readpage = cifs_readpage,
	.readpages = cifs_readpages,
	.writepage = cifs_writepage,
3615
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
3616 3617
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
L
Linus Torvalds 已提交
3618
	.set_page_dirty = __set_page_dirty_nobuffers,
3619 3620
	.releasepage = cifs_release_page,
	.invalidatepage = cifs_invalidate_page,
3621
	.launder_page = cifs_launder_page,
L
Linus Torvalds 已提交
3622
};
D
Dave Kleikamp 已提交
3623 3624 3625 3626 3627 3628

/*
 * cifs_readpages requires the server to support a buffer large enough to
 * contain the header plus one complete page of data.  Otherwise, we need
 * to leave cifs_readpages out of the address space operations.
 */
3629
const struct address_space_operations cifs_addr_ops_smallbuf = {
D
Dave Kleikamp 已提交
3630 3631 3632
	.readpage = cifs_readpage,
	.writepage = cifs_writepage,
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
3633 3634
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
D
Dave Kleikamp 已提交
3635
	.set_page_dirty = __set_page_dirty_nobuffers,
3636 3637
	.releasepage = cifs_release_page,
	.invalidatepage = cifs_invalidate_page,
3638
	.launder_page = cifs_launder_page,
D
Dave Kleikamp 已提交
3639
};