file.c 65.3 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>
L
Linus Torvalds 已提交
35 36 37 38 39 40 41 42
#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"
43
#include "fscache.h"
L
Linus Torvalds 已提交
44 45 46 47 48 49 50 51 52 53 54 55 56 57

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);
	}

58 59 60
	return (READ_CONTROL | FILE_WRITE_ATTRIBUTES | FILE_READ_ATTRIBUTES |
		FILE_WRITE_EA | FILE_APPEND_DATA | FILE_WRITE_DATA |
		FILE_READ_DATA);
61
}
62

63 64 65
static inline fmode_t cifs_posix_convert_flags(unsigned int flags)
{
	fmode_t posix_flags = 0;
66

67 68 69 70 71 72 73 74 75 76 77 78 79 80
	if ((flags & O_ACCMODE) == O_RDONLY)
		posix_flags = FMODE_READ;
	else if ((flags & O_ACCMODE) == O_WRONLY)
		posix_flags = FMODE_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; */
		posix_flags = FMODE_READ | FMODE_WRITE;
	}
	/* can not map O_CREAT or O_EXCL or O_TRUNC flags when
	   reopening a file.  They had their effect on the original open */
	if (flags & O_APPEND)
		posix_flags |= (fmode_t)O_APPEND;
81 82 83 84
	if (flags & O_DSYNC)
		posix_flags |= (fmode_t)O_DSYNC;
	if (flags & __O_SYNC)
		posix_flags |= (fmode_t)__O_SYNC;
85 86 87 88 89 90 91 92
	if (flags & O_DIRECTORY)
		posix_flags |= (fmode_t)O_DIRECTORY;
	if (flags & O_NOFOLLOW)
		posix_flags |= (fmode_t)O_NOFOLLOW;
	if (flags & O_DIRECT)
		posix_flags |= (fmode_t)O_DIRECT;

	return posix_flags;
L
Linus Torvalds 已提交
93 94 95 96 97 98 99 100 101 102
}

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;
103 104
	else if ((flags & O_TRUNC) == O_TRUNC)
		return FILE_OVERWRITE;
L
Linus Torvalds 已提交
105 106 107 108
	else
		return FILE_OPEN;
}

109
/* all arguments to this function must be checked for validity in caller */
110 111
static inline int
cifs_posix_open_inode_helper(struct inode *inode, struct file *file,
112
			     struct cifsInodeInfo *pCifsInode, __u32 oplock,
113
			     u16 netfid)
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
{

	write_lock(&GlobalSMBSeslock);

	pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
	if (pCifsInode == NULL) {
		write_unlock(&GlobalSMBSeslock);
		return -EINVAL;
	}

	if (pCifsInode->clientCanCacheRead) {
		/* we have the inode open somewhere else
		   no need to discard cache data */
		goto psx_client_can_cache;
	}

	/* BB FIXME need to fix this check to move it earlier into posix_open
	   BB  fIX following section BB FIXME */

	/* if not oplocked, invalidate inode pages if mtime or file
	   size changed */
/*	temp = cifs_NTtimeToUnix(le64_to_cpu(buf->LastWriteTime));
	if (timespec_equal(&file->f_path.dentry->d_inode->i_mtime, &temp) &&
			   (file->f_path.dentry->d_inode->i_size ==
			    (loff_t)le64_to_cpu(buf->EndOfFile))) {
139
		cFYI(1, "inode unchanged on server");
140 141 142 143 144 145
	} else {
		if (file->f_path.dentry->d_inode->i_mapping) {
			rc = filemap_write_and_wait(file->f_path.dentry->d_inode->i_mapping);
			if (rc != 0)
				CIFS_I(file->f_path.dentry->d_inode)->write_behind_rc = rc;
		}
146 147
		cFYI(1, "invalidating remote inode since open detected it "
			 "changed");
148 149 150 151 152 153 154
		invalidate_remote_inode(file->f_path.dentry->d_inode);
	} */

psx_client_can_cache:
	if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) {
		pCifsInode->clientCanCacheAll = true;
		pCifsInode->clientCanCacheRead = true;
155 156
		cFYI(1, "Exclusive Oplock granted on inode %p",
			 file->f_path.dentry->d_inode);
157 158 159 160 161 162 163 164 165
	} else if ((oplock & 0xF) == OPLOCK_READ)
		pCifsInode->clientCanCacheRead = true;

	/* will have to change the unlock if we reenable the
	   filemap_fdatawrite (which does not seem necessary */
	write_unlock(&GlobalSMBSeslock);
	return 0;
}

L
Linus Torvalds 已提交
166
/* all arguments to this function must be checked for validity in caller */
167
static inline int cifs_open_inode_helper(struct inode *inode,
168
	struct cifsTconInfo *pTcon, __u32 oplock, FILE_ALL_INFO *buf,
L
Linus Torvalds 已提交
169 170
	char *full_path, int xid)
{
171
	struct cifsInodeInfo *pCifsInode = CIFS_I(inode);
L
Linus Torvalds 已提交
172 173 174 175 176 177 178 179 180 181 182 183
	struct timespec temp;
	int rc;

	if (pCifsInode->clientCanCacheRead) {
		/* we have the inode open somewhere else
		   no need to discard cache data */
		goto client_can_cache;
	}

	/* BB need same check in cifs_create too? */
	/* if not oplocked, invalidate inode pages if mtime or file
	   size changed */
184
	temp = cifs_NTtimeToUnix(buf->LastWriteTime);
185 186
	if (timespec_equal(&inode->i_mtime, &temp) &&
			   (inode->i_size ==
L
Linus Torvalds 已提交
187
			    (loff_t)le64_to_cpu(buf->EndOfFile))) {
188
		cFYI(1, "inode unchanged on server");
L
Linus Torvalds 已提交
189
	} else {
190
		if (inode->i_mapping) {
S
Steve French 已提交
191 192
			/* BB no need to lock inode until after invalidate
			since namei code should already have it locked? */
193
			rc = filemap_write_and_wait(inode->i_mapping);
194
			if (rc != 0)
195
				pCifsInode->write_behind_rc = rc;
L
Linus Torvalds 已提交
196
		}
197 198
		cFYI(1, "invalidating remote inode since open detected it "
			 "changed");
199
		invalidate_remote_inode(inode);
L
Linus Torvalds 已提交
200 201 202
	}

client_can_cache:
203
	if (pTcon->unix_ext)
204 205
		rc = cifs_get_inode_info_unix(&inode, full_path, inode->i_sb,
					      xid);
L
Linus Torvalds 已提交
206
	else
207 208
		rc = cifs_get_inode_info(&inode, full_path, buf, inode->i_sb,
					 xid, NULL);
L
Linus Torvalds 已提交
209

210
	if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) {
211 212
		pCifsInode->clientCanCacheAll = true;
		pCifsInode->clientCanCacheRead = true;
213
		cFYI(1, "Exclusive Oplock granted on inode %p", inode);
214
	} else if ((oplock & 0xF) == OPLOCK_READ)
215
		pCifsInode->clientCanCacheRead = true;
L
Linus Torvalds 已提交
216 217 218 219 220 221 222

	return rc;
}

int cifs_open(struct inode *inode, struct file *file)
{
	int rc = -EACCES;
223 224
	int xid;
	__u32 oplock;
L
Linus Torvalds 已提交
225
	struct cifs_sb_info *cifs_sb;
226
	struct cifsTconInfo *tcon;
227
	struct tcon_link *tlink;
228
	struct cifsFileInfo *pCifsFile = NULL;
L
Linus Torvalds 已提交
229 230 231 232 233 234 235 236 237 238
	struct cifsInodeInfo *pCifsInode;
	char *full_path = NULL;
	int desiredAccess;
	int disposition;
	__u16 netfid;
	FILE_ALL_INFO *buf = NULL;

	xid = GetXid();

	cifs_sb = CIFS_SB(inode->i_sb);
239 240 241 242 243 244
	tlink = cifs_sb_tlink(cifs_sb);
	if (IS_ERR(tlink)) {
		FreeXid(xid);
		return PTR_ERR(tlink);
	}
	tcon = tlink_tcon(tlink);
L
Linus Torvalds 已提交
245

246
	pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
247

248
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
249
	if (full_path == NULL) {
250
		rc = -ENOMEM;
251
		goto out;
L
Linus Torvalds 已提交
252 253
	}

254 255
	cFYI(1, "inode = 0x%p file flags are 0x%x for %s",
		 inode, file->f_flags, full_path);
256 257 258 259 260 261

	if (oplockEnabled)
		oplock = REQ_OPLOCK;
	else
		oplock = 0;

262 263
	if (!tcon->broken_posix_open && tcon->unix_ext &&
	    (tcon->ses->capabilities & CAP_UNIX) &&
264 265 266
	    (CIFS_UNIX_POSIX_PATH_OPS_CAP &
			le64_to_cpu(tcon->fsUnixInfo.Capability))) {
		int oflags = (int) cifs_posix_convert_flags(file->f_flags);
267
		oflags |= SMB_O_CREAT;
268
		/* can not refresh inode info since size could be stale */
269
		rc = cifs_posix_open(full_path, &inode, inode->i_sb,
270 271
				cifs_sb->mnt_file_mode /* ignored */,
				oflags, &oplock, &netfid, xid);
272
		if (rc == 0) {
273
			cFYI(1, "posix open succeeded");
274 275 276
			/* no need for special case handling of setting mode
			   on read only files needed here */

277 278 279 280 281 282 283
			rc = cifs_posix_open_inode_helper(inode, file,
					pCifsInode, oplock, netfid);
			if (rc != 0) {
				CIFSSMBClose(xid, tcon, netfid);
				goto out;
			}

284 285
			pCifsFile = cifs_new_fileinfo(inode, netfid, file,
							file->f_path.mnt,
286
							tlink, oflags, oplock);
287 288 289 290
			if (pCifsFile == NULL) {
				CIFSSMBClose(xid, tcon, netfid);
				rc = -ENOMEM;
			}
291 292 293

			cifs_fscache_set_inode_cookie(inode, file);

294
			goto out;
295 296
		} else if ((rc == -EINVAL) || (rc == -EOPNOTSUPP)) {
			if (tcon->ses->serverNOS)
297
				cERROR(1, "server %s of type %s returned"
298 299 300 301
					   " unexpected error on SMB posix open"
					   ", disabling posix open support."
					   " Check if server update available.",
					   tcon->ses->serverName,
302
					   tcon->ses->serverNOS);
303
			tcon->broken_posix_open = true;
304 305 306
		} else if ((rc != -EIO) && (rc != -EREMOTE) &&
			 (rc != -EOPNOTSUPP)) /* path not found or net err */
			goto out;
307 308
		/* else fallthrough to retry open the old way on network i/o
		   or DFS errors */
309 310
	}

L
Linus Torvalds 已提交
311 312 313 314
	desiredAccess = cifs_convert_flags(file->f_flags);

/*********************************************************************
 *  open flag mapping table:
S
Steve French 已提交
315
 *
L
Linus Torvalds 已提交
316
 *	POSIX Flag            CIFS Disposition
S
Steve French 已提交
317
 *	----------            ----------------
L
Linus Torvalds 已提交
318 319 320 321 322 323 324
 *	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
S
Steve French 已提交
325
 *	FILE_SUPERSEDE (ie create whether or not file exists although
L
Linus Torvalds 已提交
326 327 328 329
 *	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)
 *?
S
Steve French 已提交
330
 *?  O_SYNC is a reasonable match to CIFS writethrough flag
L
Linus Torvalds 已提交
331 332 333 334 335 336 337 338 339 340 341 342 343 344
 *?  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(file->f_flags);

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

	/* Also refresh inode by passing in file_info buf returned by SMBOpen
	   and calling get_inode_info with returned buf (at least helps
	   non-Unix server case) */

S
Steve French 已提交
345 346
	/* BB we can not do this if this is the second open of a file
	   and the first handle has writebehind data, we might be
L
Linus Torvalds 已提交
347 348 349 350 351 352
	   able to simply do a filemap_fdatawrite/filemap_fdatawait first */
	buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
	if (!buf) {
		rc = -ENOMEM;
		goto out;
	}
353

354
	if (tcon->ses->capabilities & CAP_NT_SMBS)
355
		rc = CIFSSMBOpen(xid, tcon, full_path, disposition,
356
			 desiredAccess, CREATE_NOT_DIR, &netfid, &oplock, buf,
357 358
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags
				 & CIFS_MOUNT_MAP_SPECIAL_CHR);
359 360 361
	else
		rc = -EIO; /* no NT SMB support fall into legacy open below */

362 363
	if (rc == -EIO) {
		/* Old server, try legacy style OpenX */
364
		rc = SMBLegacyOpen(xid, tcon, full_path, disposition,
365 366 367 368
			desiredAccess, CREATE_NOT_DIR, &netfid, &oplock, buf,
			cifs_sb->local_nls, cifs_sb->mnt_cifs_flags
				& CIFS_MOUNT_MAP_SPECIAL_CHR);
	}
L
Linus Torvalds 已提交
369
	if (rc) {
370
		cFYI(1, "cifs_open returned 0x%x", rc);
L
Linus Torvalds 已提交
371 372
		goto out;
	}
373

374
	rc = cifs_open_inode_helper(inode, tcon, oplock, buf, full_path, xid);
375 376 377
	if (rc != 0)
		goto out;

J
Jeff Layton 已提交
378
	pCifsFile = cifs_new_fileinfo(inode, netfid, file, file->f_path.mnt,
379
					tlink, file->f_flags, oplock);
380
	if (pCifsFile == NULL) {
L
Linus Torvalds 已提交
381 382 383 384
		rc = -ENOMEM;
		goto out;
	}

385 386
	cifs_fscache_set_inode_cookie(inode, file);

S
Steve French 已提交
387
	if (oplock & CIFS_CREATE_ACTION) {
L
Linus Torvalds 已提交
388 389
		/* time to set mode which we can not set earlier due to
		   problems creating new read-only files */
390
		if (tcon->unix_ext) {
391 392 393 394 395 396 397 398 399
			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,
			};
400 401 402
			CIFSSMBUnixSetPathInfo(xid, tcon, full_path, &args,
					       cifs_sb->local_nls,
					       cifs_sb->mnt_cifs_flags &
403
						CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
404 405 406 407 408 409 410
		}
	}

out:
	kfree(buf);
	kfree(full_path);
	FreeXid(xid);
411
	cifs_put_tlink(tlink);
L
Linus Torvalds 已提交
412 413 414
	return rc;
}

A
Adrian Bunk 已提交
415
/* Try to reacquire byte range locks that were released when session */
L
Linus Torvalds 已提交
416 417 418 419 420 421 422 423 424 425
/* to server was lost */
static int cifs_relock_file(struct cifsFileInfo *cifsFile)
{
	int rc = 0;

/* BB list all locks open on this file and relock */

	return rc;
}

426
static int cifs_reopen_file(struct file *file, bool can_flush)
L
Linus Torvalds 已提交
427 428
{
	int rc = -EACCES;
429 430
	int xid;
	__u32 oplock;
L
Linus Torvalds 已提交
431
	struct cifs_sb_info *cifs_sb;
432
	struct cifsTconInfo *tcon;
L
Linus Torvalds 已提交
433 434
	struct cifsFileInfo *pCifsFile;
	struct cifsInodeInfo *pCifsInode;
S
Steve French 已提交
435
	struct inode *inode;
L
Linus Torvalds 已提交
436 437 438 439 440
	char *full_path = NULL;
	int desiredAccess;
	int disposition = FILE_OPEN;
	__u16 netfid;

S
Steve French 已提交
441
	if (file->private_data)
442
		pCifsFile = file->private_data;
S
Steve French 已提交
443
	else
L
Linus Torvalds 已提交
444 445 446
		return -EBADF;

	xid = GetXid();
447
	mutex_lock(&pCifsFile->fh_mutex);
448
	if (!pCifsFile->invalidHandle) {
449
		mutex_unlock(&pCifsFile->fh_mutex);
450
		rc = 0;
L
Linus Torvalds 已提交
451
		FreeXid(xid);
452
		return rc;
L
Linus Torvalds 已提交
453 454
	}

455
	if (file->f_path.dentry == NULL) {
456
		cERROR(1, "no valid name if dentry freed");
457 458 459 460 461 462
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
	}

	inode = file->f_path.dentry->d_inode;
S
Steve French 已提交
463
	if (inode == NULL) {
464
		cERROR(1, "inode not valid");
465 466 467
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
L
Linus Torvalds 已提交
468
	}
469

L
Linus Torvalds 已提交
470
	cifs_sb = CIFS_SB(inode->i_sb);
471
	tcon = tlink_tcon(pCifsFile->tlink);
472

L
Linus Torvalds 已提交
473 474 475 476
/* 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 */
477
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
478
	if (full_path == NULL) {
479 480
		rc = -ENOMEM;
reopen_error_exit:
481
		mutex_unlock(&pCifsFile->fh_mutex);
L
Linus Torvalds 已提交
482
		FreeXid(xid);
483
		return rc;
L
Linus Torvalds 已提交
484 485
	}

486 487
	cFYI(1, "inode = 0x%p file flags 0x%x for %s",
		 inode, file->f_flags, full_path);
L
Linus Torvalds 已提交
488 489 490 491

	if (oplockEnabled)
		oplock = REQ_OPLOCK;
	else
492
		oplock = 0;
L
Linus Torvalds 已提交
493

494 495 496 497 498
	if (tcon->unix_ext && (tcon->ses->capabilities & CAP_UNIX) &&
	    (CIFS_UNIX_POSIX_PATH_OPS_CAP &
			le64_to_cpu(tcon->fsUnixInfo.Capability))) {
		int oflags = (int) cifs_posix_convert_flags(file->f_flags);
		/* can not refresh inode info since size could be stale */
499
		rc = cifs_posix_open(full_path, NULL, inode->i_sb,
500 501
				cifs_sb->mnt_file_mode /* ignored */,
				oflags, &oplock, &netfid, xid);
502
		if (rc == 0) {
503
			cFYI(1, "posix reopen succeeded");
504 505 506 507 508 509 510 511
			goto reopen_success;
		}
		/* fallthrough to retry open the old way on errors, especially
		   in the reconnect path it is important to retry hard */
	}

	desiredAccess = cifs_convert_flags(file->f_flags);

L
Linus Torvalds 已提交
512
	/* Can not refresh inode by passing in file_info buf to be returned
S
Steve French 已提交
513 514
	   by SMBOpen and then calling get_inode_info with returned buf
	   since file might have write behind data that needs to be flushed
L
Linus Torvalds 已提交
515 516 517
	   and server version of file size can be stale. If we knew for sure
	   that inode was not dirty locally we could do this */

518
	rc = CIFSSMBOpen(xid, tcon, full_path, disposition, desiredAccess,
L
Linus Torvalds 已提交
519
			 CREATE_NOT_DIR, &netfid, &oplock, NULL,
S
Steve French 已提交
520
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags &
521
				CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
522
	if (rc) {
523
		mutex_unlock(&pCifsFile->fh_mutex);
524 525
		cFYI(1, "cifs_open returned 0x%x", rc);
		cFYI(1, "oplock: %d", oplock);
L
Linus Torvalds 已提交
526
	} else {
527
reopen_success:
L
Linus Torvalds 已提交
528
		pCifsFile->netfid = netfid;
529
		pCifsFile->invalidHandle = false;
530
		mutex_unlock(&pCifsFile->fh_mutex);
L
Linus Torvalds 已提交
531 532 533
		pCifsInode = CIFS_I(inode);
		if (pCifsInode) {
			if (can_flush) {
534 535 536
				rc = filemap_write_and_wait(inode->i_mapping);
				if (rc != 0)
					CIFS_I(inode)->write_behind_rc = rc;
L
Linus Torvalds 已提交
537 538
			/* temporarily disable caching while we
			   go to server to get inode info */
539 540
				pCifsInode->clientCanCacheAll = false;
				pCifsInode->clientCanCacheRead = false;
541
				if (tcon->unix_ext)
L
Linus Torvalds 已提交
542 543 544 545 546
					rc = cifs_get_inode_info_unix(&inode,
						full_path, inode->i_sb, xid);
				else
					rc = cifs_get_inode_info(&inode,
						full_path, NULL, inode->i_sb,
547
						xid, NULL);
L
Linus Torvalds 已提交
548 549 550 551 552 553 554
			} /* 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 inod
			     info */
			if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) {
555 556
				pCifsInode->clientCanCacheAll = true;
				pCifsInode->clientCanCacheRead = true;
557 558
				cFYI(1, "Exclusive Oplock granted on inode %p",
					 file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
559
			} else if ((oplock & 0xF) == OPLOCK_READ) {
560 561
				pCifsInode->clientCanCacheRead = true;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
562
			} else {
563 564
				pCifsInode->clientCanCacheRead = false;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
565 566 567 568 569 570 571 572 573 574 575 576
			}
			cifs_relock_file(pCifsFile);
		}
	}
	kfree(full_path);
	FreeXid(xid);
	return rc;
}

int cifs_close(struct inode *inode, struct file *file)
{
	int rc = 0;
577
	int xid, timeout;
L
Linus Torvalds 已提交
578 579
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
580
	struct cifsFileInfo *pSMBFile = file->private_data;
L
Linus Torvalds 已提交
581 582 583 584

	xid = GetXid();

	cifs_sb = CIFS_SB(inode->i_sb);
585
	pTcon = tlink_tcon(pSMBFile->tlink);
L
Linus Torvalds 已提交
586
	if (pSMBFile) {
J
[CIFS]  
Jeremy Allison 已提交
587
		struct cifsLockInfo *li, *tmp;
588
		write_lock(&GlobalSMBSeslock);
589
		pSMBFile->closePend = true;
L
Linus Torvalds 已提交
590 591 592
		if (pTcon) {
			/* no sense reconnecting to close a file that is
			   already closed */
S
Steve French 已提交
593
			if (!pTcon->need_reconnect) {
594
				write_unlock(&GlobalSMBSeslock);
595
				timeout = 2;
596
				while ((atomic_read(&pSMBFile->count) != 1)
597
					&& (timeout <= 2048)) {
598 599 600 601 602
					/* Give write a better chance to get to
					server ahead of the close.  We do not
					want to add a wait_q here as it would
					increase the memory utilization as
					the struct would be in each open file,
S
Steve French 已提交
603
					but this should give enough time to
604
					clear the socket */
605
					cFYI(DBG2, "close delay, write pending");
606 607
					msleep(timeout);
					timeout *= 4;
608
				}
609 610 611
				if (!pTcon->need_reconnect &&
				    !pSMBFile->invalidHandle)
					rc = CIFSSMBClose(xid, pTcon,
L
Linus Torvalds 已提交
612
						  pSMBFile->netfid);
613 614 615 616
			} else
				write_unlock(&GlobalSMBSeslock);
		} else
			write_unlock(&GlobalSMBSeslock);
J
[CIFS]  
Jeremy Allison 已提交
617 618 619

		/* Delete any outstanding lock records.
		   We'll lose them when the file is closed anyway. */
620
		mutex_lock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
621 622 623 624
		list_for_each_entry_safe(li, tmp, &pSMBFile->llist, llist) {
			list_del(&li->llist);
			kfree(li);
		}
625
		mutex_unlock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
626

627
		write_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
628 629
		list_del(&pSMBFile->flist);
		list_del(&pSMBFile->tlist);
630
		write_unlock(&GlobalSMBSeslock);
631
		cifsFileInfo_put(file->private_data);
L
Linus Torvalds 已提交
632 633 634 635
		file->private_data = NULL;
	} else
		rc = -EBADF;

636
	read_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
637
	if (list_empty(&(CIFS_I(inode)->openFileList))) {
638
		cFYI(1, "closing last open instance for inode %p", inode);
L
Linus Torvalds 已提交
639 640
		/* if the file is not open we do not know if we can cache info
		   on this inode, much less write behind and read ahead */
641 642
		CIFS_I(inode)->clientCanCacheRead = false;
		CIFS_I(inode)->clientCanCacheAll  = false;
L
Linus Torvalds 已提交
643
	}
644
	read_unlock(&GlobalSMBSeslock);
S
Steve French 已提交
645
	if ((rc == 0) && CIFS_I(inode)->write_behind_rc)
L
Linus Torvalds 已提交
646 647 648 649 650 651 652 653 654
		rc = CIFS_I(inode)->write_behind_rc;
	FreeXid(xid);
	return rc;
}

int cifs_closedir(struct inode *inode, struct file *file)
{
	int rc = 0;
	int xid;
655
	struct cifsFileInfo *pCFileStruct = file->private_data;
L
Linus Torvalds 已提交
656 657
	char *ptmp;

658
	cFYI(1, "Closedir inode = 0x%p", inode);
L
Linus Torvalds 已提交
659 660 661 662

	xid = GetXid();

	if (pCFileStruct) {
663
		struct cifsTconInfo *pTcon = tlink_tcon(pCFileStruct->tlink);
L
Linus Torvalds 已提交
664

665
		cFYI(1, "Freeing private data in close dir");
666
		write_lock(&GlobalSMBSeslock);
667 668 669
		if (!pCFileStruct->srch_inf.endOfSearch &&
		    !pCFileStruct->invalidHandle) {
			pCFileStruct->invalidHandle = true;
670
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
671
			rc = CIFSFindClose(xid, pTcon, pCFileStruct->netfid);
672 673
			cFYI(1, "Closing uncompleted readdir with rc %d",
				 rc);
L
Linus Torvalds 已提交
674 675
			/* not much we can do if it fails anyway, ignore rc */
			rc = 0;
676 677
		} else
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
678 679
		ptmp = pCFileStruct->srch_inf.ntwrk_buf_start;
		if (ptmp) {
680
			cFYI(1, "closedir free smb buf in srch struct");
L
Linus Torvalds 已提交
681
			pCFileStruct->srch_inf.ntwrk_buf_start = NULL;
S
Steve French 已提交
682
			if (pCFileStruct->srch_inf.smallBuf)
683 684 685
				cifs_small_buf_release(ptmp);
			else
				cifs_buf_release(ptmp);
L
Linus Torvalds 已提交
686
		}
687
		cifs_put_tlink(pCFileStruct->tlink);
L
Linus Torvalds 已提交
688 689 690 691 692 693 694 695
		kfree(file->private_data);
		file->private_data = NULL;
	}
	/* BB can we lock the filestruct while this is going on? */
	FreeXid(xid);
	return rc;
}

J
[CIFS]  
Jeremy Allison 已提交
696 697 698
static int store_file_lock(struct cifsFileInfo *fid, __u64 len,
				__u64 offset, __u8 lockType)
{
S
Steve French 已提交
699 700
	struct cifsLockInfo *li =
		kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
J
[CIFS]  
Jeremy Allison 已提交
701 702 703 704 705
	if (li == NULL)
		return -ENOMEM;
	li->offset = offset;
	li->length = len;
	li->type = lockType;
706
	mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
707
	list_add(&li->llist, &fid->llist);
708
	mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
709 710 711
	return 0;
}

L
Linus Torvalds 已提交
712 713 714 715 716 717
int cifs_lock(struct file *file, int cmd, struct file_lock *pfLock)
{
	int rc, xid;
	__u32 numLock = 0;
	__u32 numUnlock = 0;
	__u64 length;
718
	bool wait_flag = false;
L
Linus Torvalds 已提交
719
	struct cifs_sb_info *cifs_sb;
720
	struct cifsTconInfo *tcon;
721 722
	__u16 netfid;
	__u8 lockType = LOCKING_ANDX_LARGE_FILES;
723
	bool posix_locking = 0;
L
Linus Torvalds 已提交
724 725 726 727 728

	length = 1 + pfLock->fl_end - pfLock->fl_start;
	rc = -EACCES;
	xid = GetXid();

729
	cFYI(1, "Lock parm: 0x%x flockflags: "
L
Linus Torvalds 已提交
730
		 "0x%x flocktype: 0x%x start: %lld end: %lld",
S
Steve French 已提交
731
		cmd, pfLock->fl_flags, pfLock->fl_type, pfLock->fl_start,
732
		pfLock->fl_end);
L
Linus Torvalds 已提交
733 734

	if (pfLock->fl_flags & FL_POSIX)
735
		cFYI(1, "Posix");
L
Linus Torvalds 已提交
736
	if (pfLock->fl_flags & FL_FLOCK)
737
		cFYI(1, "Flock");
L
Linus Torvalds 已提交
738
	if (pfLock->fl_flags & FL_SLEEP) {
739
		cFYI(1, "Blocking lock");
740
		wait_flag = true;
L
Linus Torvalds 已提交
741 742
	}
	if (pfLock->fl_flags & FL_ACCESS)
743 744
		cFYI(1, "Process suspended by mandatory locking - "
			 "not implemented yet");
L
Linus Torvalds 已提交
745
	if (pfLock->fl_flags & FL_LEASE)
746
		cFYI(1, "Lease on file - not implemented yet");
S
Steve French 已提交
747
	if (pfLock->fl_flags &
L
Linus Torvalds 已提交
748
	    (~(FL_POSIX | FL_FLOCK | FL_SLEEP | FL_ACCESS | FL_LEASE)))
749
		cFYI(1, "Unknown lock flags 0x%x", pfLock->fl_flags);
L
Linus Torvalds 已提交
750 751

	if (pfLock->fl_type == F_WRLCK) {
752
		cFYI(1, "F_WRLCK ");
L
Linus Torvalds 已提交
753 754
		numLock = 1;
	} else if (pfLock->fl_type == F_UNLCK) {
755
		cFYI(1, "F_UNLCK");
L
Linus Torvalds 已提交
756
		numUnlock = 1;
757 758
		/* Check if unlock includes more than
		one lock range */
L
Linus Torvalds 已提交
759
	} else if (pfLock->fl_type == F_RDLCK) {
760
		cFYI(1, "F_RDLCK");
L
Linus Torvalds 已提交
761 762 763
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else if (pfLock->fl_type == F_EXLCK) {
764
		cFYI(1, "F_EXLCK");
L
Linus Torvalds 已提交
765 766
		numLock = 1;
	} else if (pfLock->fl_type == F_SHLCK) {
767
		cFYI(1, "F_SHLCK");
L
Linus Torvalds 已提交
768 769 770
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else
771
		cFYI(1, "Unknown type of lock");
L
Linus Torvalds 已提交
772

773
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
774
	tcon = tlink_tcon(((struct cifsFileInfo *)file->private_data)->tlink);
L
Linus Torvalds 已提交
775 776

	if (file->private_data == NULL) {
777
		rc = -EBADF;
L
Linus Torvalds 已提交
778
		FreeXid(xid);
779
		return rc;
L
Linus Torvalds 已提交
780
	}
781 782
	netfid = ((struct cifsFileInfo *)file->private_data)->netfid;

783 784
	if ((tcon->ses->capabilities & CAP_UNIX) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
S
Steve French 已提交
785
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
786
		posix_locking = 1;
787 788 789
	/* BB add code here to normalize offset and length to
	account for negative length which we can not accept over the
	wire */
L
Linus Torvalds 已提交
790
	if (IS_GETLK(cmd)) {
S
Steve French 已提交
791
		if (posix_locking) {
792
			int posix_lock_type;
S
Steve French 已提交
793
			if (lockType & LOCKING_ANDX_SHARED_LOCK)
794 795 796
				posix_lock_type = CIFS_RDLCK;
			else
				posix_lock_type = CIFS_WRLCK;
797
			rc = CIFSSMBPosixLock(xid, tcon, netfid, 1 /* get */,
798
					length,	pfLock,
799 800 801 802 803 804
					posix_lock_type, wait_flag);
			FreeXid(xid);
			return rc;
		}

		/* BB we could chain these into one lock request BB */
805
		rc = CIFSSMBLock(xid, tcon, netfid, length, pfLock->fl_start,
806
				 0, 1, lockType, 0 /* wait flag */ );
L
Linus Torvalds 已提交
807
		if (rc == 0) {
808
			rc = CIFSSMBLock(xid, tcon, netfid, length,
L
Linus Torvalds 已提交
809 810 811 812 813
					 pfLock->fl_start, 1 /* numUnlock */ ,
					 0 /* numLock */ , lockType,
					 0 /* wait flag */ );
			pfLock->fl_type = F_UNLCK;
			if (rc != 0)
814 815
				cERROR(1, "Error unlocking previously locked "
					   "range %d during test of lock", rc);
L
Linus Torvalds 已提交
816 817 818 819
			rc = 0;

		} else {
			/* if rc == ERR_SHARING_VIOLATION ? */
820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836
			rc = 0;

			if (lockType & LOCKING_ANDX_SHARED_LOCK) {
				pfLock->fl_type = F_WRLCK;
			} else {
				rc = CIFSSMBLock(xid, tcon, netfid, length,
					pfLock->fl_start, 0, 1,
					lockType | LOCKING_ANDX_SHARED_LOCK,
					0 /* wait flag */);
				if (rc == 0) {
					rc = CIFSSMBLock(xid, tcon, netfid,
						length, pfLock->fl_start, 1, 0,
						lockType |
						LOCKING_ANDX_SHARED_LOCK,
						0 /* wait flag */);
					pfLock->fl_type = F_RDLCK;
					if (rc != 0)
837
						cERROR(1, "Error unlocking "
838
						"previously locked range %d "
839
						"during test of lock", rc);
840 841 842 843 844 845
					rc = 0;
				} else {
					pfLock->fl_type = F_WRLCK;
					rc = 0;
				}
			}
L
Linus Torvalds 已提交
846 847 848 849 850
		}

		FreeXid(xid);
		return rc;
	}
J
[CIFS]  
Jeremy Allison 已提交
851 852 853 854 855 856 857 858 859

	if (!numLock && !numUnlock) {
		/* if no lock or unlock then nothing
		to do since we do not know what it is */
		FreeXid(xid);
		return -EOPNOTSUPP;
	}

	if (posix_locking) {
860
		int posix_lock_type;
S
Steve French 已提交
861
		if (lockType & LOCKING_ANDX_SHARED_LOCK)
862 863 864
			posix_lock_type = CIFS_RDLCK;
		else
			posix_lock_type = CIFS_WRLCK;
865

S
Steve French 已提交
866
		if (numUnlock == 1)
867
			posix_lock_type = CIFS_UNLCK;
J
[CIFS]  
Jeremy Allison 已提交
868

869
		rc = CIFSSMBPosixLock(xid, tcon, netfid, 0 /* set */,
870
				      length, pfLock,
871
				      posix_lock_type, wait_flag);
J
[CIFS]  
Jeremy Allison 已提交
872
	} else {
873
		struct cifsFileInfo *fid = file->private_data;
J
[CIFS]  
Jeremy Allison 已提交
874 875

		if (numLock) {
876
			rc = CIFSSMBLock(xid, tcon, netfid, length,
S
Steve French 已提交
877
					pfLock->fl_start,
J
[CIFS]  
Jeremy Allison 已提交
878 879 880 881 882 883 884 885 886 887 888 889 890
					0, numLock, lockType, wait_flag);

			if (rc == 0) {
				/* For Windows locks we must store them. */
				rc = store_file_lock(fid, length,
						pfLock->fl_start, lockType);
			}
		} else if (numUnlock) {
			/* For each stored lock that this unlock overlaps
			   completely, unlock it. */
			int stored_rc = 0;
			struct cifsLockInfo *li, *tmp;

891
			rc = 0;
892
			mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
893 894
			list_for_each_entry_safe(li, tmp, &fid->llist, llist) {
				if (pfLock->fl_start <= li->offset &&
S
Steve French 已提交
895
						(pfLock->fl_start + length) >=
896
						(li->offset + li->length)) {
897
					stored_rc = CIFSSMBLock(xid, tcon,
S
Steve French 已提交
898
							netfid,
J
[CIFS]  
Jeremy Allison 已提交
899
							li->length, li->offset,
900
							1, 0, li->type, false);
J
[CIFS]  
Jeremy Allison 已提交
901 902
					if (stored_rc)
						rc = stored_rc;
903 904 905 906
					else {
						list_del(&li->llist);
						kfree(li);
					}
J
[CIFS]  
Jeremy Allison 已提交
907 908
				}
			}
909
			mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
910 911 912
		}
	}

913
	if (pfLock->fl_flags & FL_POSIX)
L
Linus Torvalds 已提交
914 915 916 917 918
		posix_lock_file_wait(file, pfLock);
	FreeXid(xid);
	return rc;
}

919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
/*
 * Set the timeout on write requests past EOF. For some servers (Windows)
 * these calls can be very long.
 *
 * If we're writing >10M past the EOF we give a 180s timeout. Anything less
 * than that gets a 45s timeout. Writes not past EOF get 15s timeouts.
 * The 10M cutoff is totally arbitrary. A better scheme for this would be
 * welcome if someone wants to suggest one.
 *
 * We may be able to do a better job with this if there were some way to
 * declare that a file should be sparse.
 */
static int
cifs_write_timeout(struct cifsInodeInfo *cifsi, loff_t offset)
{
	if (offset <= cifsi->server_eof)
		return CIFS_STD_OP;
	else if (offset > (cifsi->server_eof + (10 * 1024 * 1024)))
		return CIFS_VLONG_OP;
	else
		return CIFS_LONG_OP;
}

/* update the file size (if needed) after a write */
static void
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;
}

L
Linus Torvalds 已提交
953 954 955 956 957 958 959 960 961 962
ssize_t cifs_user_write(struct file *file, const char __user *write_data,
	size_t write_size, loff_t *poffset)
{
	int rc = 0;
	unsigned int bytes_written = 0;
	unsigned int total_written;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	int xid, long_op;
	struct cifsFileInfo *open_file;
963
	struct cifsInodeInfo *cifsi = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
964

965
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
966

967 968
	/* cFYI(1, " write %d bytes to offset %lld of %s", write_size,
	   *poffset, file->f_path.dentry->d_name.name); */
L
Linus Torvalds 已提交
969 970 971

	if (file->private_data == NULL)
		return -EBADF;
972

973
	open_file = file->private_data;
974
	pTcon = tlink_tcon(open_file->tlink);
975

976 977 978 979
	rc = generic_write_checks(file, poffset, &write_size, 0);
	if (rc)
		return rc;

L
Linus Torvalds 已提交
980 981
	xid = GetXid();

982
	long_op = cifs_write_timeout(cifsi, *poffset);
L
Linus Torvalds 已提交
983 984 985 986 987 988 989 990 991 992 993
	for (total_written = 0; write_size > total_written;
	     total_written += bytes_written) {
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
			if (file->private_data == NULL) {
				/* file has been closed on us */
				FreeXid(xid);
			/* if we have gotten here we have written some data
			   and blocked, and the file has been freed on us while
			   we blocked so return what we managed to write */
				return total_written;
S
Steve French 已提交
994
			}
L
Linus Torvalds 已提交
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006
			if (open_file->closePend) {
				FreeXid(xid);
				if (total_written)
					return total_written;
				else
					return -EBADF;
			}
			if (open_file->invalidHandle) {
				/* we could deadlock if we called
				   filemap_fdatawait from here so tell
				   reopen_file not to flush data to server
				   now */
1007
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
				if (rc != 0)
					break;
			}

			rc = CIFSSMBWrite(xid, pTcon,
				open_file->netfid,
				min_t(const int, cifs_sb->wsize,
				      write_size - total_written),
				*poffset, &bytes_written,
				NULL, write_data + total_written, long_op);
		}
		if (rc || (bytes_written == 0)) {
			if (total_written)
				break;
			else {
				FreeXid(xid);
				return rc;
			}
1026 1027
		} else {
			cifs_update_eof(cifsi, *poffset, bytes_written);
L
Linus Torvalds 已提交
1028
			*poffset += bytes_written;
1029
		}
1030
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1031 1032 1033
				    15 seconds is plenty */
	}

1034
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1035 1036

	/* since the write may have blocked check these pointers again */
1037 1038
	if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
		struct inode *inode = file->f_path.dentry->d_inode;
S
Steve French 已提交
1039 1040
/* Do not update local mtime - server will set its actual value on write
 *		inode->i_ctime = inode->i_mtime =
1041 1042 1043 1044 1045
 * 			current_fs_time(inode->i_sb);*/
		if (total_written > 0) {
			spin_lock(&inode->i_lock);
			if (*poffset > file->f_path.dentry->d_inode->i_size)
				i_size_write(file->f_path.dentry->d_inode,
L
Linus Torvalds 已提交
1046
					*poffset);
1047
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1048
		}
S
Steve French 已提交
1049
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1050 1051 1052 1053 1054 1055
	}
	FreeXid(xid);
	return total_written;
}

static ssize_t cifs_write(struct file *file, const char *write_data,
N
Nick Piggin 已提交
1056
			  size_t write_size, loff_t *poffset)
L
Linus Torvalds 已提交
1057 1058 1059 1060 1061 1062 1063 1064
{
	int rc = 0;
	unsigned int bytes_written = 0;
	unsigned int total_written;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	int xid, long_op;
	struct cifsFileInfo *open_file;
1065
	struct cifsInodeInfo *cifsi = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1066

1067
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1068

1069 1070
	cFYI(1, "write %zd bytes to offset %lld of %s", write_size,
	   *poffset, file->f_path.dentry->d_name.name);
L
Linus Torvalds 已提交
1071 1072 1073

	if (file->private_data == NULL)
		return -EBADF;
1074
	open_file = file->private_data;
1075
	pTcon = tlink_tcon(open_file->tlink);
1076

L
Linus Torvalds 已提交
1077 1078
	xid = GetXid();

1079
	long_op = cifs_write_timeout(cifsi, *poffset);
L
Linus Torvalds 已提交
1080 1081 1082 1083 1084 1085 1086 1087 1088
	for (total_written = 0; write_size > total_written;
	     total_written += bytes_written) {
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
			if (file->private_data == NULL) {
				/* file has been closed on us */
				FreeXid(xid);
			/* if we have gotten here we have written some data
			   and blocked, and the file has been freed on us
S
Steve French 已提交
1089
			   while we blocked so return what we managed to
L
Linus Torvalds 已提交
1090 1091
			   write */
				return total_written;
S
Steve French 已提交
1092
			}
L
Linus Torvalds 已提交
1093 1094 1095 1096 1097 1098 1099 1100 1101 1102
			if (open_file->closePend) {
				FreeXid(xid);
				if (total_written)
					return total_written;
				else
					return -EBADF;
			}
			if (open_file->invalidHandle) {
				/* we could deadlock if we called
				   filemap_fdatawait from here so tell
S
Steve French 已提交
1103
				   reopen_file not to flush data to
L
Linus Torvalds 已提交
1104
				   server now */
1105
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
1106 1107 1108
				if (rc != 0)
					break;
			}
S
Steve French 已提交
1109 1110
			if (experimEnabled || (pTcon->ses->server &&
				((pTcon->ses->server->secMode &
S
Steve French 已提交
1111
				(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED))
S
Steve French 已提交
1112
				== 0))) {
1113 1114 1115
				struct kvec iov[2];
				unsigned int len;

1116
				len = min((size_t)cifs_sb->wsize,
1117 1118 1119 1120 1121
					  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;
1122
				rc = CIFSSMBWrite2(xid, pTcon,
1123
						open_file->netfid, len,
1124
						*poffset, &bytes_written,
1125
						iov, 1, long_op);
1126
			} else
1127 1128 1129 1130 1131 1132 1133
				rc = CIFSSMBWrite(xid, pTcon,
					 open_file->netfid,
					 min_t(const int, cifs_sb->wsize,
					       write_size - total_written),
					 *poffset, &bytes_written,
					 write_data + total_written,
					 NULL, long_op);
L
Linus Torvalds 已提交
1134 1135 1136 1137 1138 1139 1140 1141
		}
		if (rc || (bytes_written == 0)) {
			if (total_written)
				break;
			else {
				FreeXid(xid);
				return rc;
			}
1142 1143
		} else {
			cifs_update_eof(cifsi, *poffset, bytes_written);
L
Linus Torvalds 已提交
1144
			*poffset += bytes_written;
1145
		}
1146
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1147 1148 1149
				    15 seconds is plenty */
	}

1150
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1151 1152

	/* since the write may have blocked check these pointers again */
1153
	if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
1154
/*BB We could make this contingent on superblock ATIME flag too */
1155 1156 1157 1158 1159 1160 1161 1162
/*		file->f_path.dentry->d_inode->i_ctime =
		file->f_path.dentry->d_inode->i_mtime = CURRENT_TIME;*/
		if (total_written > 0) {
			spin_lock(&file->f_path.dentry->d_inode->i_lock);
			if (*poffset > file->f_path.dentry->d_inode->i_size)
				i_size_write(file->f_path.dentry->d_inode,
					     *poffset);
			spin_unlock(&file->f_path.dentry->d_inode->i_lock);
L
Linus Torvalds 已提交
1163
		}
1164
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1165 1166 1167 1168 1169
	}
	FreeXid(xid);
	return total_written;
}

S
Steve French 已提交
1170
#ifdef CONFIG_CIFS_EXPERIMENTAL
1171 1172
struct cifsFileInfo *find_readable_file(struct cifsInodeInfo *cifs_inode,
					bool fsuid_only)
S
Steve French 已提交
1173 1174
{
	struct cifsFileInfo *open_file = NULL;
1175 1176 1177 1178 1179
	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
1180 1181 1182 1183 1184 1185 1186 1187

	read_lock(&GlobalSMBSeslock);
	/* 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) {
		if (open_file->closePend)
			continue;
1188 1189
		if (fsuid_only && open_file->uid != current_fsuid())
			continue;
S
Steve French 已提交
1190 1191 1192 1193 1194
		if (open_file->pfile && ((open_file->pfile->f_flags & O_RDWR) ||
		    (open_file->pfile->f_flags & O_RDONLY))) {
			if (!open_file->invalidHandle) {
				/* found a good file */
				/* lock it so it will not be closed on us */
1195
				cifsFileInfo_get(open_file);
S
Steve French 已提交
1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
				read_unlock(&GlobalSMBSeslock);
				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 */
	}
	read_unlock(&GlobalSMBSeslock);
	return NULL;
}
#endif

1209 1210
struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode,
					bool fsuid_only)
1211 1212
{
	struct cifsFileInfo *open_file;
1213
	struct cifs_sb_info *cifs_sb = CIFS_SB(cifs_inode->vfs_inode.i_sb);
1214
	bool any_available = false;
1215
	int rc;
1216

1217 1218 1219 1220
	/* 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
1221
	if (cifs_inode == NULL) {
1222
		cERROR(1, "Null inode passed to cifs_writeable_file");
1223 1224 1225 1226
		dump_stack();
		return NULL;
	}

1227 1228 1229 1230
	/* only filter by fsuid on multiuser mounts */
	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MULTIUSER))
		fsuid_only = false;

1231
	read_lock(&GlobalSMBSeslock);
1232
refind_writable:
1233
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1234 1235 1236 1237 1238
		if (open_file->closePend)
			continue;
		if (!any_available && open_file->pid != current->tgid)
			continue;
		if (fsuid_only && open_file->uid != current_fsuid())
1239 1240 1241 1242
			continue;
		if (open_file->pfile &&
		    ((open_file->pfile->f_flags & O_RDWR) ||
		     (open_file->pfile->f_flags & O_WRONLY))) {
1243
			cifsFileInfo_get(open_file);
1244 1245 1246 1247 1248 1249

			if (!open_file->invalidHandle) {
				/* found a good writable file */
				read_unlock(&GlobalSMBSeslock);
				return open_file;
			}
S
Steve French 已提交
1250

1251
			read_unlock(&GlobalSMBSeslock);
1252
			/* Had to unlock since following call can block */
1253
			rc = cifs_reopen_file(open_file->pfile, false);
S
Steve French 已提交
1254
			if (!rc) {
1255 1256 1257 1258
				if (!open_file->closePend)
					return open_file;
				else { /* start over in case this was deleted */
				       /* since the list could be modified */
1259
					read_lock(&GlobalSMBSeslock);
1260
					cifsFileInfo_put(open_file);
1261
					goto refind_writable;
1262 1263
				}
			}
1264 1265 1266 1267 1268 1269 1270 1271

			/* if it fails, try another handle if possible -
			(we can not do this if closePending since
			loop could be modified - in which case we
			have to start at the beginning of the list
			again. Note that it would be bad
			to hold up writepages here (rather than
			in caller) with continuous retries */
1272
			cFYI(1, "wp failed on reopen file");
1273 1274 1275
			read_lock(&GlobalSMBSeslock);
			/* can not use this handle, no write
			   pending on this one after all */
1276
			cifsFileInfo_put(open_file);
S
Steve French 已提交
1277

1278 1279 1280 1281 1282 1283 1284 1285 1286
			if (open_file->closePend) /* list could have changed */
				goto refind_writable;
			/* else we simply continue to the next entry. Thus
			   we do not loop on reopen errors.  If we
			   can not reopen the file, for example if we
			   reconnected to a server with another client
			   racing to delete or lock the file we would not
			   make progress if we restarted before the beginning
			   of the loop here. */
1287 1288
		}
	}
1289 1290 1291 1292 1293
	/* couldn't find useable FH with same pid, try any available */
	if (!any_available) {
		any_available = true;
		goto refind_writable;
	}
1294 1295 1296 1297
	read_unlock(&GlobalSMBSeslock);
	return NULL;
}

L
Linus Torvalds 已提交
1298 1299 1300 1301 1302 1303 1304 1305 1306
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 cifs_sb_info *cifs_sb;
	struct inode *inode;
1307
	struct cifsFileInfo *open_file;
L
Linus Torvalds 已提交
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331

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

	inode = page->mapping->host;
	cifs_sb = CIFS_SB(inode->i_sb);

	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
1332
		to = (unsigned)(mapping->host->i_size - offset);
L
Linus Torvalds 已提交
1333

1334
	open_file = find_writable_file(CIFS_I(mapping->host), false);
1335 1336 1337
	if (open_file) {
		bytes_written = cifs_write(open_file->pfile, write_data,
					   to-from, &offset);
1338
		cifsFileInfo_put(open_file);
L
Linus Torvalds 已提交
1339
		/* Does mm or vfs already set times? */
1340
		inode->i_atime = inode->i_mtime = current_fs_time(inode->i_sb);
1341
		if ((bytes_written > 0) && (offset))
1342
			rc = 0;
1343 1344
		else if (bytes_written < 0)
			rc = bytes_written;
1345
	} else {
1346
		cFYI(1, "No writeable filehandles for inode");
L
Linus Torvalds 已提交
1347 1348 1349 1350 1351 1352 1353 1354
		rc = -EIO;
	}

	kunmap(page);
	return rc;
}

static int cifs_writepages(struct address_space *mapping,
1355
			   struct writeback_control *wbc)
L
Linus Torvalds 已提交
1356
{
1357 1358 1359 1360 1361
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	unsigned int bytes_to_write;
	unsigned int bytes_written;
	struct cifs_sb_info *cifs_sb;
	int done = 0;
1362
	pgoff_t end;
1363
	pgoff_t index;
S
Steve French 已提交
1364 1365
	int range_whole = 0;
	struct kvec *iov;
1366
	int len;
1367 1368 1369 1370
	int n_iov = 0;
	pgoff_t next;
	int nr_pages;
	__u64 offset = 0;
1371
	struct cifsFileInfo *open_file;
1372
	struct cifsTconInfo *tcon;
1373
	struct cifsInodeInfo *cifsi = CIFS_I(mapping->host);
1374 1375 1376 1377
	struct page *page;
	struct pagevec pvec;
	int rc = 0;
	int scanned = 0;
1378
	int xid, long_op;
L
Linus Torvalds 已提交
1379

1380 1381 1382 1383 1384 1385 1386 1387 1388
	/*
	 * BB: Is this meaningful for a non-block-device file system?
	 * If it is, we should test it again after we do I/O
	 */
	if (wbc->nonblocking && bdi_write_congested(bdi)) {
		wbc->encountered_congestion = 1;
		return 0;
	}

1389
	cifs_sb = CIFS_SB(mapping->host->i_sb);
1390

1391 1392 1393 1394 1395 1396 1397
	/*
	 * If wsize is smaller that the page cache size, default to writing
	 * one page at a time via cifs_writepage
	 */
	if (cifs_sb->wsize < PAGE_CACHE_SIZE)
		return generic_writepages(mapping, wbc);

1398
	iov = kmalloc(32 * sizeof(struct kvec), GFP_KERNEL);
S
Steve French 已提交
1399
	if (iov == NULL)
1400 1401
		return generic_writepages(mapping, wbc);

1402
	/*
1403 1404 1405
	 * if there's no open file, then this is likely to fail too,
	 * but it'll at least handle the return. Maybe it should be
	 * a BUG() instead?
1406
	 */
1407
	open_file = find_writable_file(CIFS_I(mapping->host), false);
1408
	if (!open_file) {
1409
		kfree(iov);
1410 1411 1412
		return generic_writepages(mapping, wbc);
	}

1413
	tcon = tlink_tcon(open_file->tlink);
1414 1415 1416 1417
	if (!experimEnabled && tcon->ses->server->secMode &
			(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED)) {
		cifsFileInfo_put(open_file);
		return generic_writepages(mapping, wbc);
1418
	}
1419
	cifsFileInfo_put(open_file);
1420

L
Linus Torvalds 已提交
1421 1422
	xid = GetXid();

1423
	pagevec_init(&pvec, 0);
1424
	if (wbc->range_cyclic) {
1425
		index = mapping->writeback_index; /* Start from prev offset */
1426 1427 1428 1429 1430 1431
		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)
			range_whole = 1;
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
		scanned = 1;
	}
retry:
	while (!done && (index <= end) &&
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
			PAGECACHE_TAG_DIRTY,
			min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1))) {
		int first;
		unsigned int i;

		first = -1;
		next = 0;
		n_iov = 0;
		bytes_to_write = 0;

		for (i = 0; i < nr_pages; i++) {
			page = pvec.pages[i];
			/*
			 * 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
			 */

			if (first < 0)
				lock_page(page);
N
Nick Piggin 已提交
1459
			else if (!trylock_page(page))
1460 1461 1462 1463 1464 1465 1466
				break;

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

1467
			if (!wbc->range_cyclic && page->index > end) {
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
				done = 1;
				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) ||
1483
					!clear_page_dirty_for_io(page)) {
1484 1485 1486
				unlock_page(page);
				break;
			}
1487

1488 1489 1490 1491 1492 1493
			/*
			 * This actually clears the dirty bit in the radix tree.
			 * See cifs_writepage() for more commentary.
			 */
			set_page_writeback(page);

1494 1495 1496
			if (page_offset(page) >= mapping->host->i_size) {
				done = 1;
				unlock_page(page);
1497
				end_page_writeback(page);
1498 1499 1500
				break;
			}

1501 1502 1503 1504 1505
			/*
			 * BB can we get rid of this?  pages are held by pvec
			 */
			page_cache_get(page);

1506 1507 1508
			len = min(mapping->host->i_size - page_offset(page),
				  (loff_t)PAGE_CACHE_SIZE);

1509 1510 1511
			/* reserve iov[0] for the smb header */
			n_iov++;
			iov[n_iov].iov_base = kmap(page);
1512 1513
			iov[n_iov].iov_len = len;
			bytes_to_write += len;
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523

			if (first < 0) {
				first = i;
				offset = page_offset(page);
			}
			next = page->index + 1;
			if (bytes_to_write + PAGE_CACHE_SIZE > cifs_sb->wsize)
				break;
		}
		if (n_iov) {
1524 1525
			open_file = find_writable_file(CIFS_I(mapping->host),
							false);
1526
			if (!open_file) {
1527
				cERROR(1, "No writable handles for inode");
1528
				rc = -EBADF;
S
Steve French 已提交
1529
			} else {
1530
				long_op = cifs_write_timeout(cifsi, offset);
1531
				rc = CIFSSMBWrite2(xid, tcon, open_file->netfid,
1532 1533
						   bytes_to_write, offset,
						   &bytes_written, iov, n_iov,
1534
						   long_op);
1535
				cifsFileInfo_put(open_file);
1536
				cifs_update_eof(cifsi, offset, bytes_written);
1537
			}
1538

1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
			if (rc || bytes_written < bytes_to_write) {
				cERROR(1, "Write2 ret %d, wrote %d",
					  rc, bytes_written);
				/* BB what if continued retry is
				   requested via mount flags? */
				if (rc == -ENOSPC)
					set_bit(AS_ENOSPC, &mapping->flags);
				else
					set_bit(AS_EIO, &mapping->flags);
			} else {
				cifs_stats_bytes_written(tcon, bytes_written);
1550
			}
1551

1552 1553
			for (i = 0; i < n_iov; i++) {
				page = pvec.pages[first + i];
1554 1555 1556 1557
				/* Should we also set page error on
				success rc but too little data written? */
				/* BB investigate retry logic on temporary
				server crash cases and how recovery works
S
Steve French 已提交
1558 1559
				when page marked as error */
				if (rc)
1560
					SetPageError(page);
1561 1562
				kunmap(page);
				unlock_page(page);
1563
				end_page_writeback(page);
1564 1565 1566 1567 1568
				page_cache_release(page);
			}
			if ((wbc->nr_to_write -= n_iov) <= 0)
				done = 1;
			index = next;
1569 1570 1571 1572
		} else
			/* Need to re-find the pages we skipped */
			index = pvec.pages[0]->index + 1;

1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
		pagevec_release(&pvec);
	}
	if (!scanned && !done) {
		/*
		 * We hit the last page and there is more work to be done: wrap
		 * back to the start of the file
		 */
		scanned = 1;
		index = 0;
		goto retry;
	}
1584
	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1585 1586
		mapping->writeback_index = index;

L
Linus Torvalds 已提交
1587
	FreeXid(xid);
1588
	kfree(iov);
L
Linus Torvalds 已提交
1589 1590 1591
	return rc;
}

S
Steve French 已提交
1592
static int cifs_writepage(struct page *page, struct writeback_control *wbc)
L
Linus Torvalds 已提交
1593 1594 1595 1596 1597 1598 1599
{
	int rc = -EFAULT;
	int xid;

	xid = GetXid();
/* BB add check for wbc flags */
	page_cache_get(page);
S
Steve French 已提交
1600
	if (!PageUptodate(page))
1601
		cFYI(1, "ppw - page not up to date");
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612

	/*
	 * 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
1613
	set_page_writeback(page);
L
Linus Torvalds 已提交
1614 1615 1616
	rc = cifs_partialpagewrite(page, 0, PAGE_CACHE_SIZE);
	SetPageUptodate(page); /* BB add check for error and Clearuptodate? */
	unlock_page(page);
1617 1618
	end_page_writeback(page);
	page_cache_release(page);
L
Linus Torvalds 已提交
1619 1620 1621 1622
	FreeXid(xid);
	return rc;
}

N
Nick Piggin 已提交
1623 1624 1625
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
1626
{
N
Nick Piggin 已提交
1627 1628
	int rc;
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
1629

1630 1631
	cFYI(1, "write_end for page %p from pos %lld with %d bytes",
		 page, pos, copied);
N
Nick Piggin 已提交
1632

1633 1634 1635 1636 1637
	if (PageChecked(page)) {
		if (copied == len)
			SetPageUptodate(page);
		ClearPageChecked(page);
	} else if (!PageUptodate(page) && copied == PAGE_CACHE_SIZE)
N
Nick Piggin 已提交
1638
		SetPageUptodate(page);
S
Steve French 已提交
1639

L
Linus Torvalds 已提交
1640
	if (!PageUptodate(page)) {
N
Nick Piggin 已提交
1641 1642 1643 1644 1645
		char *page_data;
		unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
		int xid;

		xid = GetXid();
L
Linus Torvalds 已提交
1646 1647 1648 1649 1650 1651
		/* 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);
N
Nick Piggin 已提交
1652 1653
		rc = cifs_write(file, page_data + offset, copied, &pos);
		/* if (rc < 0) should we set writebehind rc? */
L
Linus Torvalds 已提交
1654
		kunmap(page);
N
Nick Piggin 已提交
1655 1656

		FreeXid(xid);
S
Steve French 已提交
1657
	} else {
N
Nick Piggin 已提交
1658 1659
		rc = copied;
		pos += copied;
L
Linus Torvalds 已提交
1660 1661 1662
		set_page_dirty(page);
	}

N
Nick Piggin 已提交
1663 1664 1665 1666 1667 1668 1669 1670 1671 1672
	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
1673 1674 1675
	return rc;
}

1676
int cifs_fsync(struct file *file, int datasync)
L
Linus Torvalds 已提交
1677 1678 1679
{
	int xid;
	int rc = 0;
1680
	struct cifsTconInfo *tcon;
1681
	struct cifsFileInfo *smbfile = file->private_data;
1682
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1683 1684 1685

	xid = GetXid();

1686
	cFYI(1, "Sync file - name: %s datasync: 0x%x",
1687
		file->f_path.dentry->d_name.name, datasync);
1688

1689 1690 1691
	rc = filemap_write_and_wait(inode->i_mapping);
	if (rc == 0) {
		rc = CIFS_I(inode)->write_behind_rc;
L
Linus Torvalds 已提交
1692
		CIFS_I(inode)->write_behind_rc = 0;
1693
		tcon = tlink_tcon(smbfile->tlink);
1694
		if (!rc && tcon && smbfile &&
S
Steve French 已提交
1695
		   !(CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC))
1696
			rc = CIFSSMBFlush(xid, tcon, smbfile->netfid);
1697
	}
1698

L
Linus Torvalds 已提交
1699 1700 1701 1702
	FreeXid(xid);
	return rc;
}

1703
/* static void cifs_sync_page(struct page *page)
L
Linus Torvalds 已提交
1704 1705 1706 1707 1708 1709 1710
{
	struct address_space *mapping;
	struct inode *inode;
	unsigned long index = page->index;
	unsigned int rpages = 0;
	int rc = 0;

1711
	cFYI(1, "sync page %p", page);
L
Linus Torvalds 已提交
1712 1713 1714 1715 1716
	mapping = page->mapping;
	if (!mapping)
		return 0;
	inode = mapping->host;
	if (!inode)
1717
		return; */
L
Linus Torvalds 已提交
1718

S
Steve French 已提交
1719
/*	fill in rpages then
L
Linus Torvalds 已提交
1720 1721
	result = cifs_pagein_inode(inode, index, rpages); */ /* BB finish */

1722
/*	cFYI(1, "rpages is %d for sync page of Index %ld", rpages, index);
L
Linus Torvalds 已提交
1723

1724
#if 0
L
Linus Torvalds 已提交
1725 1726 1727
	if (rc < 0)
		return rc;
	return 0;
1728
#endif
L
Linus Torvalds 已提交
1729 1730 1731 1732 1733 1734
} */

/*
 * As file closes, flush all cached write data for this inode checking
 * for write behind errors.
 */
1735
int cifs_flush(struct file *file, fl_owner_t id)
L
Linus Torvalds 已提交
1736
{
S
Steve French 已提交
1737
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749
	int rc = 0;

	/* Rather than do the steps manually:
	   lock the inode for writing
	   loop through pages looking for write behind data (dirty pages)
	   coalesce into contiguous 16K (or smaller) chunks to write to server
	   send to server (prefer in parallel)
	   deal with writebehind errors
	   unlock inode for writing
	   filemapfdatawrite appears easier for the time being */

	rc = filemap_fdatawrite(inode->i_mapping);
1750 1751 1752
	/* reset wb rc if we were able to write out dirty pages */
	if (!rc) {
		rc = CIFS_I(inode)->write_behind_rc;
L
Linus Torvalds 已提交
1753
		CIFS_I(inode)->write_behind_rc = 0;
1754
	}
1755

1756
	cFYI(1, "Flush inode %p file %p rc %d", inode, file, rc);
L
Linus Torvalds 已提交
1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776

	return rc;
}

ssize_t cifs_user_read(struct file *file, char __user *read_data,
	size_t read_size, loff_t *poffset)
{
	int rc = -EACCES;
	unsigned int bytes_read = 0;
	unsigned int total_read = 0;
	unsigned int current_read_size;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	int xid;
	struct cifsFileInfo *open_file;
	char *smb_read_data;
	char __user *current_offset;
	struct smb_com_read_rsp *pSMBr;

	xid = GetXid();
1777
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1778 1779

	if (file->private_data == NULL) {
1780
		rc = -EBADF;
L
Linus Torvalds 已提交
1781
		FreeXid(xid);
1782
		return rc;
L
Linus Torvalds 已提交
1783
	}
1784
	open_file = file->private_data;
1785
	pTcon = tlink_tcon(open_file->tlink);
L
Linus Torvalds 已提交
1786

S
Steve French 已提交
1787
	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
1788
		cFYI(1, "attempting read on write only file instance");
S
Steve French 已提交
1789

L
Linus Torvalds 已提交
1790 1791 1792
	for (total_read = 0, current_offset = read_data;
	     read_size > total_read;
	     total_read += bytes_read, current_offset += bytes_read) {
S
Steve French 已提交
1793
		current_read_size = min_t(const int, read_size - total_read,
L
Linus Torvalds 已提交
1794 1795 1796 1797
					  cifs_sb->rsize);
		rc = -EAGAIN;
		smb_read_data = NULL;
		while (rc == -EAGAIN) {
1798
			int buf_type = CIFS_NO_BUFFER;
S
Steve French 已提交
1799
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1800
			    (!open_file->closePend)) {
1801
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1802 1803 1804
				if (rc != 0)
					break;
			}
1805
			rc = CIFSSMBRead(xid, pTcon,
1806 1807 1808 1809
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
L
Linus Torvalds 已提交
1810 1811
			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
			if (smb_read_data) {
1812 1813 1814 1815
				if (copy_to_user(current_offset,
						smb_read_data +
						4 /* RFC1001 length field */ +
						le16_to_cpu(pSMBr->DataOffset),
S
Steve French 已提交
1816
						bytes_read))
1817 1818
					rc = -EFAULT;

S
Steve French 已提交
1819
				if (buf_type == CIFS_SMALL_BUFFER)
1820
					cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1821
				else if (buf_type == CIFS_LARGE_BUFFER)
1822
					cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833
				smb_read_data = NULL;
			}
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1834
			cifs_stats_bytes_read(pTcon, bytes_read);
L
Linus Torvalds 已提交
1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854
			*poffset += bytes_read;
		}
	}
	FreeXid(xid);
	return total_read;
}


static ssize_t cifs_read(struct file *file, char *read_data, size_t read_size,
	loff_t *poffset)
{
	int rc = -EACCES;
	unsigned int bytes_read = 0;
	unsigned int total_read;
	unsigned int current_read_size;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	int xid;
	char *current_offset;
	struct cifsFileInfo *open_file;
1855
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1856 1857

	xid = GetXid();
1858
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1859 1860

	if (file->private_data == NULL) {
1861
		rc = -EBADF;
L
Linus Torvalds 已提交
1862
		FreeXid(xid);
1863
		return rc;
L
Linus Torvalds 已提交
1864
	}
1865
	open_file = file->private_data;
1866
	pTcon = tlink_tcon(open_file->tlink);
L
Linus Torvalds 已提交
1867 1868

	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
1869
		cFYI(1, "attempting read on write only file instance");
L
Linus Torvalds 已提交
1870

S
Steve French 已提交
1871
	for (total_read = 0, current_offset = read_data;
L
Linus Torvalds 已提交
1872 1873 1874 1875
	     read_size > total_read;
	     total_read += bytes_read, current_offset += bytes_read) {
		current_read_size = min_t(const int, read_size - total_read,
					  cifs_sb->rsize);
1876 1877
		/* For windows me and 9x we do not want to request more
		than it negotiated since it will refuse the read then */
S
Steve French 已提交
1878
		if ((pTcon->ses) &&
1879 1880 1881 1882
			!(pTcon->ses->capabilities & CAP_LARGE_FILES)) {
			current_read_size = min_t(const int, current_read_size,
					pTcon->ses->server->maxBuf - 128);
		}
L
Linus Torvalds 已提交
1883 1884
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
1885
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1886
			    (!open_file->closePend)) {
1887
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1888 1889 1890
				if (rc != 0)
					break;
			}
1891
			rc = CIFSSMBRead(xid, pTcon,
1892 1893 1894 1895
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &current_offset,
					 &buf_type);
L
Linus Torvalds 已提交
1896 1897 1898 1899 1900 1901 1902 1903 1904
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1905
			cifs_stats_bytes_read(pTcon, total_read);
L
Linus Torvalds 已提交
1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917
			*poffset += bytes_read;
		}
	}
	FreeXid(xid);
	return total_read;
}

int cifs_file_mmap(struct file *file, struct vm_area_struct *vma)
{
	int rc, xid;

	xid = GetXid();
J
Jeff Layton 已提交
1918
	rc = cifs_revalidate_file(file);
L
Linus Torvalds 已提交
1919
	if (rc) {
1920
		cFYI(1, "Validation prior to mmap failed, error=%d", rc);
L
Linus Torvalds 已提交
1921 1922 1923 1924 1925 1926 1927 1928 1929
		FreeXid(xid);
		return rc;
	}
	rc = generic_file_mmap(file, vma);
	FreeXid(xid);
	return rc;
}


S
Steve French 已提交
1930
static void cifs_copy_cache_pages(struct address_space *mapping,
N
Nick Piggin 已提交
1931
	struct list_head *pages, int bytes_read, char *data)
L
Linus Torvalds 已提交
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
{
	struct page *page;
	char *target;

	while (bytes_read > 0) {
		if (list_empty(pages))
			break;

		page = list_entry(pages->prev, struct page, lru);
		list_del(&page->lru);

N
Nick Piggin 已提交
1943
		if (add_to_page_cache_lru(page, mapping, page->index,
L
Linus Torvalds 已提交
1944 1945
				      GFP_KERNEL)) {
			page_cache_release(page);
1946
			cFYI(1, "Add page cache failed");
1947 1948
			data += PAGE_CACHE_SIZE;
			bytes_read -= PAGE_CACHE_SIZE;
L
Linus Torvalds 已提交
1949 1950
			continue;
		}
J
Jeff Layton 已提交
1951
		page_cache_release(page);
L
Linus Torvalds 已提交
1952

S
Steve French 已提交
1953
		target = kmap_atomic(page, KM_USER0);
L
Linus Torvalds 已提交
1954 1955 1956 1957

		if (PAGE_CACHE_SIZE > bytes_read) {
			memcpy(target, data, bytes_read);
			/* zero the tail end of this partial page */
S
Steve French 已提交
1958
			memset(target + bytes_read, 0,
L
Linus Torvalds 已提交
1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970
			       PAGE_CACHE_SIZE - bytes_read);
			bytes_read = 0;
		} else {
			memcpy(target, data, PAGE_CACHE_SIZE);
			bytes_read -= PAGE_CACHE_SIZE;
		}
		kunmap_atomic(target, KM_USER0);

		flush_dcache_page(page);
		SetPageUptodate(page);
		unlock_page(page);
		data += PAGE_CACHE_SIZE;
1971 1972 1973

		/* add page to FS-Cache */
		cifs_readpage_to_fscache(mapping->host, page);
L
Linus Torvalds 已提交
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
	}
	return;
}

static int cifs_readpages(struct file *file, struct address_space *mapping,
	struct list_head *page_list, unsigned num_pages)
{
	int rc = -EACCES;
	int xid;
	loff_t offset;
	struct page *page;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
S
Steve French 已提交
1987
	unsigned int bytes_read = 0;
S
Steve French 已提交
1988
	unsigned int read_size, i;
L
Linus Torvalds 已提交
1989 1990 1991
	char *smb_read_data = NULL;
	struct smb_com_read_rsp *pSMBr;
	struct cifsFileInfo *open_file;
1992
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1993 1994 1995

	xid = GetXid();
	if (file->private_data == NULL) {
1996
		rc = -EBADF;
L
Linus Torvalds 已提交
1997
		FreeXid(xid);
1998
		return rc;
L
Linus Torvalds 已提交
1999
	}
2000
	open_file = file->private_data;
2001
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
2002
	pTcon = tlink_tcon(open_file->tlink);
2003

2004 2005 2006 2007 2008 2009 2010 2011 2012
	/*
	 * 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)
		goto read_complete;

2013
	cFYI(DBG2, "rpages: num pages %d", num_pages);
L
Linus Torvalds 已提交
2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026
	for (i = 0; i < num_pages; ) {
		unsigned contig_pages;
		struct page *tmp_page;
		unsigned long expected_index;

		if (list_empty(page_list))
			break;

		page = list_entry(page_list->prev, struct page, lru);
		offset = (loff_t)page->index << PAGE_CACHE_SHIFT;

		/* count adjacent pages that we will read into */
		contig_pages = 0;
S
Steve French 已提交
2027
		expected_index =
L
Linus Torvalds 已提交
2028
			list_entry(page_list->prev, struct page, lru)->index;
S
Steve French 已提交
2029
		list_for_each_entry_reverse(tmp_page, page_list, lru) {
L
Linus Torvalds 已提交
2030 2031 2032 2033
			if (tmp_page->index == expected_index) {
				contig_pages++;
				expected_index++;
			} else
S
Steve French 已提交
2034
				break;
L
Linus Torvalds 已提交
2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
		}
		if (contig_pages + i >  num_pages)
			contig_pages = num_pages - i;

		/* for reads over a certain size could initiate async
		   read ahead */

		read_size = contig_pages * PAGE_CACHE_SIZE;
		/* Read size needs to be in multiples of one page */
		read_size = min_t(const unsigned int, read_size,
				  cifs_sb->rsize & PAGE_CACHE_MASK);
2046 2047
		cFYI(DBG2, "rpages: read size 0x%x  contiguous pages %d",
				read_size, contig_pages);
L
Linus Torvalds 已提交
2048 2049
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
2050
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
2051
			    (!open_file->closePend)) {
2052
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
2053 2054 2055 2056
				if (rc != 0)
					break;
			}

2057
			rc = CIFSSMBRead(xid, pTcon,
2058 2059 2060 2061
					 open_file->netfid,
					 read_size, offset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
2062
			/* BB more RC checks ? */
S
Steve French 已提交
2063
			if (rc == -EAGAIN) {
L
Linus Torvalds 已提交
2064
				if (smb_read_data) {
S
Steve French 已提交
2065
					if (buf_type == CIFS_SMALL_BUFFER)
2066
						cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2067
					else if (buf_type == CIFS_LARGE_BUFFER)
2068
						cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2069 2070 2071 2072 2073
					smb_read_data = NULL;
				}
			}
		}
		if ((rc < 0) || (smb_read_data == NULL)) {
2074
			cFYI(1, "Read error in readpages: %d", rc);
L
Linus Torvalds 已提交
2075 2076
			break;
		} else if (bytes_read > 0) {
2077
			task_io_account_read(bytes_read);
L
Linus Torvalds 已提交
2078 2079 2080
			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
			cifs_copy_cache_pages(mapping, page_list, bytes_read,
				smb_read_data + 4 /* RFC1001 hdr */ +
N
Nick Piggin 已提交
2081
				le16_to_cpu(pSMBr->DataOffset));
L
Linus Torvalds 已提交
2082 2083

			i +=  bytes_read >> PAGE_CACHE_SHIFT;
2084
			cifs_stats_bytes_read(pTcon, bytes_read);
S
Steve French 已提交
2085
			if ((bytes_read & PAGE_CACHE_MASK) != bytes_read) {
L
Linus Torvalds 已提交
2086 2087
				i++; /* account for partial page */

S
Steve French 已提交
2088
				/* server copy of file can have smaller size
L
Linus Torvalds 已提交
2089
				   than client */
S
Steve French 已提交
2090 2091
				/* BB do we need to verify this common case ?
				   this case is ok - if we are at server EOF
L
Linus Torvalds 已提交
2092 2093
				   we will hit it on next read */

O
OGAWA Hirofumi 已提交
2094
				/* break; */
L
Linus Torvalds 已提交
2095 2096
			}
		} else {
2097
			cFYI(1, "No bytes read (%d) at offset %lld . "
2098
				"Cleaning remaining pages from readahead list",
2099
				bytes_read, offset);
S
Steve French 已提交
2100
			/* BB turn off caching and do new lookup on
L
Linus Torvalds 已提交
2101 2102 2103 2104
			   file size at server? */
			break;
		}
		if (smb_read_data) {
S
Steve French 已提交
2105
			if (buf_type == CIFS_SMALL_BUFFER)
2106
				cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2107
			else if (buf_type == CIFS_LARGE_BUFFER)
2108
				cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2109 2110 2111 2112 2113 2114 2115
			smb_read_data = NULL;
		}
		bytes_read = 0;
	}

/* need to free smb_read_data buf before exit */
	if (smb_read_data) {
S
Steve French 已提交
2116
		if (buf_type == CIFS_SMALL_BUFFER)
2117
			cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2118
		else if (buf_type == CIFS_LARGE_BUFFER)
2119
			cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2120
		smb_read_data = NULL;
S
Steve French 已提交
2121
	}
L
Linus Torvalds 已提交
2122

2123
read_complete:
L
Linus Torvalds 已提交
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133
	FreeXid(xid);
	return rc;
}

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

2134 2135 2136 2137 2138
	/* 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
2139 2140 2141
	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
2142

L
Linus Torvalds 已提交
2143
	rc = cifs_read(file, read_data, PAGE_CACHE_SIZE, poffset);
S
Steve French 已提交
2144

L
Linus Torvalds 已提交
2145 2146 2147
	if (rc < 0)
		goto io_error;
	else
2148
		cFYI(1, "Bytes read %d", rc);
S
Steve French 已提交
2149

2150 2151
	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
2152

L
Linus Torvalds 已提交
2153 2154 2155 2156 2157
	if (PAGE_CACHE_SIZE > rc)
		memset(read_data + rc, 0, PAGE_CACHE_SIZE - rc);

	flush_dcache_page(page);
	SetPageUptodate(page);
2158 2159 2160 2161

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

L
Linus Torvalds 已提交
2162
	rc = 0;
S
Steve French 已提交
2163

L
Linus Torvalds 已提交
2164
io_error:
S
Steve French 已提交
2165
	kunmap(page);
L
Linus Torvalds 已提交
2166
	page_cache_release(page);
2167 2168

read_complete:
L
Linus Torvalds 已提交
2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180
	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;
	int xid;

	xid = GetXid();

	if (file->private_data == NULL) {
2181
		rc = -EBADF;
L
Linus Torvalds 已提交
2182
		FreeXid(xid);
2183
		return rc;
L
Linus Torvalds 已提交
2184 2185
	}

2186 2187
	cFYI(1, "readpage %p at offset %d 0x%x\n",
		 page, (int)offset, (int)offset);
L
Linus Torvalds 已提交
2188 2189 2190 2191 2192 2193 2194 2195 2196

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

	unlock_page(page);

	FreeXid(xid);
	return rc;
}

2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215
static int is_inode_writable(struct cifsInodeInfo *cifs_inode)
{
	struct cifsFileInfo *open_file;

	read_lock(&GlobalSMBSeslock);
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
		if (open_file->closePend)
			continue;
		if (open_file->pfile &&
		    ((open_file->pfile->f_flags & O_RDWR) ||
		     (open_file->pfile->f_flags & O_WRONLY))) {
			read_unlock(&GlobalSMBSeslock);
			return 1;
		}
	}
	read_unlock(&GlobalSMBSeslock);
	return 0;
}

L
Linus Torvalds 已提交
2216 2217 2218
/* 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
2219
   refreshing the inode only on increases in the file size
L
Linus Torvalds 已提交
2220 2221
   but this is tricky to do without racing with writebehind
   page caching in the current Linux kernel design */
2222
bool is_size_safe_to_change(struct cifsInodeInfo *cifsInode, __u64 end_of_file)
L
Linus Torvalds 已提交
2223
{
2224
	if (!cifsInode)
2225
		return true;
2226

2227 2228
	if (is_inode_writable(cifsInode)) {
		/* This inode is open for write at least once */
2229 2230 2231
		struct cifs_sb_info *cifs_sb;

		cifs_sb = CIFS_SB(cifsInode->vfs_inode.i_sb);
S
Steve French 已提交
2232
		if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DIRECT_IO) {
S
Steve French 已提交
2233
			/* since no page cache to corrupt on directio
2234
			we can change size safely */
2235
			return true;
2236 2237
		}

S
Steve French 已提交
2238
		if (i_size_read(&cifsInode->vfs_inode) < end_of_file)
2239
			return true;
2240

2241
		return false;
2242
	} else
2243
		return true;
L
Linus Torvalds 已提交
2244 2245
}

N
Nick Piggin 已提交
2246 2247 2248
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
2249
{
N
Nick Piggin 已提交
2250 2251
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	loff_t offset = pos & (PAGE_CACHE_SIZE - 1);
2252 2253 2254 2255
	loff_t page_start = pos & PAGE_MASK;
	loff_t i_size;
	struct page *page;
	int rc = 0;
N
Nick Piggin 已提交
2256

2257
	cFYI(1, "write_begin from %lld len %d", (long long)pos, len);
N
Nick Piggin 已提交
2258

2259
	page = grab_cache_page_write_begin(mapping, index, flags);
2260 2261 2262 2263
	if (!page) {
		rc = -ENOMEM;
		goto out;
	}
2264

2265 2266
	if (PageUptodate(page))
		goto out;
2267

2268 2269 2270 2271 2272 2273 2274
	/*
	 * 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;
2275

2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298
	/*
	 * 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
2299

2300 2301 2302 2303 2304 2305 2306
	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);
2307 2308 2309 2310
	} 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
2311
		   this will be written out by write_end so is fine */
L
Linus Torvalds 已提交
2312
	}
2313 2314 2315
out:
	*pagep = page;
	return rc;
L
Linus Torvalds 已提交
2316 2317
}

2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333
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);
}

2334
void cifs_oplock_break(struct work_struct *work)
2335 2336 2337
{
	struct cifsFileInfo *cfile = container_of(work, struct cifsFileInfo,
						  oplock_break);
2338
	struct inode *inode = cfile->dentry->d_inode;
2339 2340 2341 2342
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	int rc, waitrc = 0;

	if (inode && S_ISREG(inode->i_mode)) {
S
Steve French 已提交
2343
		if (cinode->clientCanCacheRead)
2344
			break_lease(inode, O_RDONLY);
S
Steve French 已提交
2345
		else
2346
			break_lease(inode, O_WRONLY);
2347 2348 2349 2350 2351 2352 2353 2354 2355
		rc = filemap_fdatawrite(inode->i_mapping);
		if (cinode->clientCanCacheRead == 0) {
			waitrc = filemap_fdatawait(inode->i_mapping);
			invalidate_remote_inode(inode);
		}
		if (!rc)
			rc = waitrc;
		if (rc)
			cinode->write_behind_rc = rc;
2356
		cFYI(1, "Oplock flush inode %p rc %d", inode, rc);
2357 2358 2359 2360 2361 2362 2363 2364 2365
	}

	/*
	 * 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
	 */
	if (!cfile->closePend && !cfile->oplock_break_cancelled) {
2366 2367
		rc = CIFSSMBLock(0, tlink_tcon(cfile->tlink), cfile->netfid, 0,
				 0, 0, 0, LOCKING_ANDX_OPLOCK_RELEASE, false);
2368
		cFYI(1, "Oplock release rc = %d", rc);
2369
	}
2370 2371 2372 2373 2374 2375 2376 2377 2378 2379

	/*
	 * We might have kicked in before is_valid_oplock_break()
	 * finished grabbing reference for us.  Make sure it's done by
	 * waiting for GlobalSMSSeslock.
	 */
	write_lock(&GlobalSMBSeslock);
	write_unlock(&GlobalSMBSeslock);

	cifs_oplock_break_put(cfile);
2380 2381
}

2382
void cifs_oplock_break_get(struct cifsFileInfo *cfile)
2383 2384 2385 2386 2387
{
	mntget(cfile->mnt);
	cifsFileInfo_get(cfile);
}

2388
void cifs_oplock_break_put(struct cifsFileInfo *cfile)
2389 2390 2391 2392 2393
{
	mntput(cfile->mnt);
	cifsFileInfo_put(cfile);
}

2394
const struct address_space_operations cifs_addr_ops = {
L
Linus Torvalds 已提交
2395 2396 2397
	.readpage = cifs_readpage,
	.readpages = cifs_readpages,
	.writepage = cifs_writepage,
2398
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2399 2400
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
L
Linus Torvalds 已提交
2401
	.set_page_dirty = __set_page_dirty_nobuffers,
2402 2403
	.releasepage = cifs_release_page,
	.invalidatepage = cifs_invalidate_page,
L
Linus Torvalds 已提交
2404 2405 2406
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};
D
Dave Kleikamp 已提交
2407 2408 2409 2410 2411 2412

/*
 * 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.
 */
2413
const struct address_space_operations cifs_addr_ops_smallbuf = {
D
Dave Kleikamp 已提交
2414 2415 2416
	.readpage = cifs_readpage,
	.writepage = cifs_writepage,
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2417 2418
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
D
Dave Kleikamp 已提交
2419
	.set_page_dirty = __set_page_dirty_nobuffers,
2420 2421
	.releasepage = cifs_release_page,
	.invalidatepage = cifs_invalidate_page,
D
Dave Kleikamp 已提交
2422 2423 2424
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};