file.c 64.4 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
static u32 cifs_posix_convert_flags(unsigned int flags)
64
{
65
	u32 posix_flags = 0;
66

67
	if ((flags & O_ACCMODE) == O_RDONLY)
68
		posix_flags = SMB_O_RDONLY;
69
	else if ((flags & O_ACCMODE) == O_WRONLY)
70 71 72 73 74 75 76 77 78 79 80
		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 */
81
	if (flags & O_DSYNC)
82
		posix_flags |= SMB_O_SYNC;
83
	if (flags & O_DIRECTORY)
84
		posix_flags |= SMB_O_DIRECTORY;
85
	if (flags & O_NOFOLLOW)
86
		posix_flags |= SMB_O_NOFOLLOW;
87
	if (flags & O_DIRECT)
88
		posix_flags |= SMB_O_DIRECT;
89 90

	return posix_flags;
L
Linus Torvalds 已提交
91 92 93 94 95 96 97 98 99 100
}

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;
101 102
	else if ((flags & O_TRUNC) == O_TRUNC)
		return FILE_OVERWRITE;
L
Linus Torvalds 已提交
103 104 105 106
	else
		return FILE_OPEN;
}

107
static inline int cifs_open_inode_helper(struct inode *inode,
108
	struct cifsTconInfo *pTcon, __u32 oplock, FILE_ALL_INFO *buf,
L
Linus Torvalds 已提交
109 110
	char *full_path, int xid)
{
111
	struct cifsInodeInfo *pCifsInode = CIFS_I(inode);
L
Linus Torvalds 已提交
112 113 114 115 116 117 118 119 120 121 122 123
	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 */
124
	temp = cifs_NTtimeToUnix(buf->LastWriteTime);
125 126
	if (timespec_equal(&inode->i_mtime, &temp) &&
			   (inode->i_size ==
L
Linus Torvalds 已提交
127
			    (loff_t)le64_to_cpu(buf->EndOfFile))) {
128
		cFYI(1, "inode unchanged on server");
L
Linus Torvalds 已提交
129
	} else {
130
		if (inode->i_mapping) {
S
Steve French 已提交
131 132
			/* BB no need to lock inode until after invalidate
			since namei code should already have it locked? */
133
			rc = filemap_write_and_wait(inode->i_mapping);
134
			if (rc != 0)
135
				pCifsInode->write_behind_rc = rc;
L
Linus Torvalds 已提交
136
		}
137 138
		cFYI(1, "invalidating remote inode since open detected it "
			 "changed");
139
		invalidate_remote_inode(inode);
L
Linus Torvalds 已提交
140 141 142
	}

client_can_cache:
143
	if (pTcon->unix_ext)
144 145
		rc = cifs_get_inode_info_unix(&inode, full_path, inode->i_sb,
					      xid);
L
Linus Torvalds 已提交
146
	else
147 148
		rc = cifs_get_inode_info(&inode, full_path, buf, inode->i_sb,
					 xid, NULL);
L
Linus Torvalds 已提交
149

150
	if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) {
151 152
		pCifsInode->clientCanCacheAll = true;
		pCifsInode->clientCanCacheRead = true;
153
		cFYI(1, "Exclusive Oplock granted on inode %p", inode);
154
	} else if ((oplock & 0xF) == OPLOCK_READ)
155
		pCifsInode->clientCanCacheRead = true;
L
Linus Torvalds 已提交
156 157 158 159

	return rc;
}

160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
int cifs_posix_open(char *full_path, struct inode **pinode,
			struct super_block *sb, int mode, unsigned int f_flags,
			__u32 *poplock, __u16 *pnetfid, int xid)
{
	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;
	struct cifsTconInfo *tcon;

	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;
}

L
Linus Torvalds 已提交
222 223 224
int cifs_open(struct inode *inode, struct file *file)
{
	int rc = -EACCES;
225 226
	int xid;
	__u32 oplock;
L
Linus Torvalds 已提交
227
	struct cifs_sb_info *cifs_sb;
228
	struct cifsTconInfo *tcon;
229
	struct tcon_link *tlink;
230
	struct cifsFileInfo *pCifsFile = NULL;
L
Linus Torvalds 已提交
231 232 233 234 235 236 237 238 239 240
	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);
241 242 243 244 245 246
	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
247

248
	pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
249

250
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
251
	if (full_path == NULL) {
252
		rc = -ENOMEM;
253
		goto out;
L
Linus Torvalds 已提交
254 255
	}

256 257
	cFYI(1, "inode = 0x%p file flags are 0x%x for %s",
		 inode, file->f_flags, full_path);
258 259 260 261 262 263

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

264 265
	if (!tcon->broken_posix_open && tcon->unix_ext &&
	    (tcon->ses->capabilities & CAP_UNIX) &&
266 267 268
	    (CIFS_UNIX_POSIX_PATH_OPS_CAP &
			le64_to_cpu(tcon->fsUnixInfo.Capability))) {
		/* can not refresh inode info since size could be stale */
269
		rc = cifs_posix_open(full_path, &inode, inode->i_sb,
270
				cifs_sb->mnt_file_mode /* ignored */,
271
				file->f_flags, &oplock, &netfid, xid);
272
		if (rc == 0) {
273
			cFYI(1, "posix open succeeded");
274

275 276
			pCifsFile = cifs_new_fileinfo(netfid, file, tlink,
						      oplock);
277 278 279 280
			if (pCifsFile == NULL) {
				CIFSSMBClose(xid, tcon, netfid);
				rc = -ENOMEM;
			}
281 282 283

			cifs_fscache_set_inode_cookie(inode, file);

284
			goto out;
285 286
		} else if ((rc == -EINVAL) || (rc == -EOPNOTSUPP)) {
			if (tcon->ses->serverNOS)
287
				cERROR(1, "server %s of type %s returned"
288 289 290 291
					   " unexpected error on SMB posix open"
					   ", disabling posix open support."
					   " Check if server update available.",
					   tcon->ses->serverName,
292
					   tcon->ses->serverNOS);
293
			tcon->broken_posix_open = true;
294 295 296
		} else if ((rc != -EIO) && (rc != -EREMOTE) &&
			 (rc != -EOPNOTSUPP)) /* path not found or net err */
			goto out;
297 298
		/* else fallthrough to retry open the old way on network i/o
		   or DFS errors */
299 300
	}

L
Linus Torvalds 已提交
301 302 303 304
	desiredAccess = cifs_convert_flags(file->f_flags);

/*********************************************************************
 *  open flag mapping table:
S
Steve French 已提交
305
 *
L
Linus Torvalds 已提交
306
 *	POSIX Flag            CIFS Disposition
S
Steve French 已提交
307
 *	----------            ----------------
L
Linus Torvalds 已提交
308 309 310 311 312 313 314
 *	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
315
 *	FILE_SUPERSEDE (ie create whether or not file exists although
L
Linus Torvalds 已提交
316 317 318 319
 *	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
320
 *?  O_SYNC is a reasonable match to CIFS writethrough flag
L
Linus Torvalds 已提交
321 322 323 324 325 326 327 328 329 330 331 332 333 334
 *?  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
335 336
	/* 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
337 338 339 340 341 342
	   able to simply do a filemap_fdatawrite/filemap_fdatawait first */
	buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
	if (!buf) {
		rc = -ENOMEM;
		goto out;
	}
343

344
	if (tcon->ses->capabilities & CAP_NT_SMBS)
345
		rc = CIFSSMBOpen(xid, tcon, full_path, disposition,
346
			 desiredAccess, CREATE_NOT_DIR, &netfid, &oplock, buf,
347 348
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags
				 & CIFS_MOUNT_MAP_SPECIAL_CHR);
349 350 351
	else
		rc = -EIO; /* no NT SMB support fall into legacy open below */

352 353
	if (rc == -EIO) {
		/* Old server, try legacy style OpenX */
354
		rc = SMBLegacyOpen(xid, tcon, full_path, disposition,
355 356 357 358
			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
359
	if (rc) {
360
		cFYI(1, "cifs_open returned 0x%x", rc);
L
Linus Torvalds 已提交
361 362
		goto out;
	}
363

364
	rc = cifs_open_inode_helper(inode, tcon, oplock, buf, full_path, xid);
365 366 367
	if (rc != 0)
		goto out;

368
	pCifsFile = cifs_new_fileinfo(netfid, file, tlink, oplock);
369
	if (pCifsFile == NULL) {
L
Linus Torvalds 已提交
370 371 372 373
		rc = -ENOMEM;
		goto out;
	}

374 375
	cifs_fscache_set_inode_cookie(inode, file);

S
Steve French 已提交
376
	if (oplock & CIFS_CREATE_ACTION) {
L
Linus Torvalds 已提交
377 378
		/* time to set mode which we can not set earlier due to
		   problems creating new read-only files */
379
		if (tcon->unix_ext) {
380 381 382 383 384 385 386 387 388
			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,
			};
389 390 391
			CIFSSMBUnixSetPathInfo(xid, tcon, full_path, &args,
					       cifs_sb->local_nls,
					       cifs_sb->mnt_cifs_flags &
392
						CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
393 394 395 396 397 398 399
		}
	}

out:
	kfree(buf);
	kfree(full_path);
	FreeXid(xid);
400
	cifs_put_tlink(tlink);
L
Linus Torvalds 已提交
401 402 403
	return rc;
}

A
Adrian Bunk 已提交
404
/* Try to reacquire byte range locks that were released when session */
L
Linus Torvalds 已提交
405 406 407 408 409 410 411 412 413 414
/* 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;
}

415
static int cifs_reopen_file(struct file *file, bool can_flush)
L
Linus Torvalds 已提交
416 417
{
	int rc = -EACCES;
418 419
	int xid;
	__u32 oplock;
L
Linus Torvalds 已提交
420
	struct cifs_sb_info *cifs_sb;
421
	struct cifsTconInfo *tcon;
L
Linus Torvalds 已提交
422 423
	struct cifsFileInfo *pCifsFile;
	struct cifsInodeInfo *pCifsInode;
S
Steve French 已提交
424
	struct inode *inode;
L
Linus Torvalds 已提交
425 426 427 428 429
	char *full_path = NULL;
	int desiredAccess;
	int disposition = FILE_OPEN;
	__u16 netfid;

S
Steve French 已提交
430
	if (file->private_data)
431
		pCifsFile = file->private_data;
S
Steve French 已提交
432
	else
L
Linus Torvalds 已提交
433 434 435
		return -EBADF;

	xid = GetXid();
436
	mutex_lock(&pCifsFile->fh_mutex);
437
	if (!pCifsFile->invalidHandle) {
438
		mutex_unlock(&pCifsFile->fh_mutex);
439
		rc = 0;
L
Linus Torvalds 已提交
440
		FreeXid(xid);
441
		return rc;
L
Linus Torvalds 已提交
442 443
	}

444
	if (file->f_path.dentry == NULL) {
445
		cERROR(1, "no valid name if dentry freed");
446 447 448 449 450 451
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
	}

	inode = file->f_path.dentry->d_inode;
S
Steve French 已提交
452
	if (inode == NULL) {
453
		cERROR(1, "inode not valid");
454 455 456
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
L
Linus Torvalds 已提交
457
	}
458

L
Linus Torvalds 已提交
459
	cifs_sb = CIFS_SB(inode->i_sb);
460
	tcon = tlink_tcon(pCifsFile->tlink);
461

L
Linus Torvalds 已提交
462 463 464 465
/* 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 */
466
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
467
	if (full_path == NULL) {
468 469
		rc = -ENOMEM;
reopen_error_exit:
470
		mutex_unlock(&pCifsFile->fh_mutex);
L
Linus Torvalds 已提交
471
		FreeXid(xid);
472
		return rc;
L
Linus Torvalds 已提交
473 474
	}

475 476
	cFYI(1, "inode = 0x%p file flags 0x%x for %s",
		 inode, file->f_flags, full_path);
L
Linus Torvalds 已提交
477 478 479 480

	if (oplockEnabled)
		oplock = REQ_OPLOCK;
	else
481
		oplock = 0;
L
Linus Torvalds 已提交
482

483 484 485
	if (tcon->unix_ext && (tcon->ses->capabilities & CAP_UNIX) &&
	    (CIFS_UNIX_POSIX_PATH_OPS_CAP &
			le64_to_cpu(tcon->fsUnixInfo.Capability))) {
486 487 488 489 490 491 492

		/*
		 * O_CREAT, O_EXCL and O_TRUNC already had their effect on the
		 * original open. Must mask them off for a reopen.
		 */
		unsigned int oflags = file->f_flags & ~(O_CREAT|O_EXCL|O_TRUNC);

493
		rc = cifs_posix_open(full_path, NULL, inode->i_sb,
494 495
				cifs_sb->mnt_file_mode /* ignored */,
				oflags, &oplock, &netfid, xid);
496
		if (rc == 0) {
497
			cFYI(1, "posix reopen succeeded");
498 499 500 501 502 503 504 505
			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
506
	/* Can not refresh inode by passing in file_info buf to be returned
S
Steve French 已提交
507 508
	   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
509 510 511
	   and server version of file size can be stale. If we knew for sure
	   that inode was not dirty locally we could do this */

512
	rc = CIFSSMBOpen(xid, tcon, full_path, disposition, desiredAccess,
L
Linus Torvalds 已提交
513
			 CREATE_NOT_DIR, &netfid, &oplock, NULL,
S
Steve French 已提交
514
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags &
515
				CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
516
	if (rc) {
517
		mutex_unlock(&pCifsFile->fh_mutex);
518 519
		cFYI(1, "cifs_open returned 0x%x", rc);
		cFYI(1, "oplock: %d", oplock);
L
Linus Torvalds 已提交
520
	} else {
521
reopen_success:
L
Linus Torvalds 已提交
522
		pCifsFile->netfid = netfid;
523
		pCifsFile->invalidHandle = false;
524
		mutex_unlock(&pCifsFile->fh_mutex);
L
Linus Torvalds 已提交
525 526 527
		pCifsInode = CIFS_I(inode);
		if (pCifsInode) {
			if (can_flush) {
528 529 530
				rc = filemap_write_and_wait(inode->i_mapping);
				if (rc != 0)
					CIFS_I(inode)->write_behind_rc = rc;
L
Linus Torvalds 已提交
531 532
			/* temporarily disable caching while we
			   go to server to get inode info */
533 534
				pCifsInode->clientCanCacheAll = false;
				pCifsInode->clientCanCacheRead = false;
535
				if (tcon->unix_ext)
L
Linus Torvalds 已提交
536 537 538 539 540
					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,
541
						xid, NULL);
L
Linus Torvalds 已提交
542 543 544 545 546 547 548
			} /* 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) {
549 550
				pCifsInode->clientCanCacheAll = true;
				pCifsInode->clientCanCacheRead = true;
551 552
				cFYI(1, "Exclusive Oplock granted on inode %p",
					 file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
553
			} else if ((oplock & 0xF) == OPLOCK_READ) {
554 555
				pCifsInode->clientCanCacheRead = true;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
556
			} else {
557 558
				pCifsInode->clientCanCacheRead = false;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
559 560 561 562 563 564 565 566 567 568 569 570
			}
			cifs_relock_file(pCifsFile);
		}
	}
	kfree(full_path);
	FreeXid(xid);
	return rc;
}

int cifs_close(struct inode *inode, struct file *file)
{
	int rc = 0;
571
	int xid, timeout;
L
Linus Torvalds 已提交
572 573
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
574
	struct cifsFileInfo *pSMBFile = file->private_data;
L
Linus Torvalds 已提交
575 576 577 578

	xid = GetXid();

	cifs_sb = CIFS_SB(inode->i_sb);
579
	pTcon = tlink_tcon(pSMBFile->tlink);
L
Linus Torvalds 已提交
580
	if (pSMBFile) {
J
[CIFS]  
Jeremy Allison 已提交
581
		struct cifsLockInfo *li, *tmp;
582
		write_lock(&GlobalSMBSeslock);
583
		pSMBFile->closePend = true;
L
Linus Torvalds 已提交
584 585 586
		if (pTcon) {
			/* no sense reconnecting to close a file that is
			   already closed */
S
Steve French 已提交
587
			if (!pTcon->need_reconnect) {
588
				write_unlock(&GlobalSMBSeslock);
589
				timeout = 2;
590
				while ((atomic_read(&pSMBFile->count) != 1)
591
					&& (timeout <= 2048)) {
592 593 594 595 596
					/* 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
597
					but this should give enough time to
598
					clear the socket */
599
					cFYI(DBG2, "close delay, write pending");
600 601
					msleep(timeout);
					timeout *= 4;
602
				}
603 604 605
				if (!pTcon->need_reconnect &&
				    !pSMBFile->invalidHandle)
					rc = CIFSSMBClose(xid, pTcon,
L
Linus Torvalds 已提交
606
						  pSMBFile->netfid);
607 608 609 610
			} else
				write_unlock(&GlobalSMBSeslock);
		} else
			write_unlock(&GlobalSMBSeslock);
J
[CIFS]  
Jeremy Allison 已提交
611 612 613

		/* Delete any outstanding lock records.
		   We'll lose them when the file is closed anyway. */
614
		mutex_lock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
615 616 617 618
		list_for_each_entry_safe(li, tmp, &pSMBFile->llist, llist) {
			list_del(&li->llist);
			kfree(li);
		}
619
		mutex_unlock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
620

621
		write_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
622 623
		list_del(&pSMBFile->flist);
		list_del(&pSMBFile->tlist);
624
		write_unlock(&GlobalSMBSeslock);
625
		cifsFileInfo_put(file->private_data);
L
Linus Torvalds 已提交
626 627 628 629
		file->private_data = NULL;
	} else
		rc = -EBADF;

630
	read_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
631
	if (list_empty(&(CIFS_I(inode)->openFileList))) {
632
		cFYI(1, "closing last open instance for inode %p", inode);
L
Linus Torvalds 已提交
633 634
		/* 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 */
635 636
		CIFS_I(inode)->clientCanCacheRead = false;
		CIFS_I(inode)->clientCanCacheAll  = false;
L
Linus Torvalds 已提交
637
	}
638
	read_unlock(&GlobalSMBSeslock);
S
Steve French 已提交
639
	if ((rc == 0) && CIFS_I(inode)->write_behind_rc)
L
Linus Torvalds 已提交
640 641 642 643 644 645 646 647 648
		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;
649
	struct cifsFileInfo *pCFileStruct = file->private_data;
L
Linus Torvalds 已提交
650 651
	char *ptmp;

652
	cFYI(1, "Closedir inode = 0x%p", inode);
L
Linus Torvalds 已提交
653 654 655 656

	xid = GetXid();

	if (pCFileStruct) {
657
		struct cifsTconInfo *pTcon = tlink_tcon(pCFileStruct->tlink);
L
Linus Torvalds 已提交
658

659
		cFYI(1, "Freeing private data in close dir");
660
		write_lock(&GlobalSMBSeslock);
661 662 663
		if (!pCFileStruct->srch_inf.endOfSearch &&
		    !pCFileStruct->invalidHandle) {
			pCFileStruct->invalidHandle = true;
664
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
665
			rc = CIFSFindClose(xid, pTcon, pCFileStruct->netfid);
666 667
			cFYI(1, "Closing uncompleted readdir with rc %d",
				 rc);
L
Linus Torvalds 已提交
668 669
			/* not much we can do if it fails anyway, ignore rc */
			rc = 0;
670 671
		} else
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
672 673
		ptmp = pCFileStruct->srch_inf.ntwrk_buf_start;
		if (ptmp) {
674
			cFYI(1, "closedir free smb buf in srch struct");
L
Linus Torvalds 已提交
675
			pCFileStruct->srch_inf.ntwrk_buf_start = NULL;
S
Steve French 已提交
676
			if (pCFileStruct->srch_inf.smallBuf)
677 678 679
				cifs_small_buf_release(ptmp);
			else
				cifs_buf_release(ptmp);
L
Linus Torvalds 已提交
680
		}
681
		cifs_put_tlink(pCFileStruct->tlink);
L
Linus Torvalds 已提交
682 683 684 685 686 687 688 689
		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
690 691 692
static int store_file_lock(struct cifsFileInfo *fid, __u64 len,
				__u64 offset, __u8 lockType)
{
S
Steve French 已提交
693 694
	struct cifsLockInfo *li =
		kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
J
[CIFS]  
Jeremy Allison 已提交
695 696 697 698 699
	if (li == NULL)
		return -ENOMEM;
	li->offset = offset;
	li->length = len;
	li->type = lockType;
700
	mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
701
	list_add(&li->llist, &fid->llist);
702
	mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
703 704 705
	return 0;
}

L
Linus Torvalds 已提交
706 707 708 709 710 711
int cifs_lock(struct file *file, int cmd, struct file_lock *pfLock)
{
	int rc, xid;
	__u32 numLock = 0;
	__u32 numUnlock = 0;
	__u64 length;
712
	bool wait_flag = false;
L
Linus Torvalds 已提交
713
	struct cifs_sb_info *cifs_sb;
714
	struct cifsTconInfo *tcon;
715 716
	__u16 netfid;
	__u8 lockType = LOCKING_ANDX_LARGE_FILES;
717
	bool posix_locking = 0;
L
Linus Torvalds 已提交
718 719 720 721 722

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

723
	cFYI(1, "Lock parm: 0x%x flockflags: "
L
Linus Torvalds 已提交
724
		 "0x%x flocktype: 0x%x start: %lld end: %lld",
S
Steve French 已提交
725
		cmd, pfLock->fl_flags, pfLock->fl_type, pfLock->fl_start,
726
		pfLock->fl_end);
L
Linus Torvalds 已提交
727 728

	if (pfLock->fl_flags & FL_POSIX)
729
		cFYI(1, "Posix");
L
Linus Torvalds 已提交
730
	if (pfLock->fl_flags & FL_FLOCK)
731
		cFYI(1, "Flock");
L
Linus Torvalds 已提交
732
	if (pfLock->fl_flags & FL_SLEEP) {
733
		cFYI(1, "Blocking lock");
734
		wait_flag = true;
L
Linus Torvalds 已提交
735 736
	}
	if (pfLock->fl_flags & FL_ACCESS)
737 738
		cFYI(1, "Process suspended by mandatory locking - "
			 "not implemented yet");
L
Linus Torvalds 已提交
739
	if (pfLock->fl_flags & FL_LEASE)
740
		cFYI(1, "Lease on file - not implemented yet");
S
Steve French 已提交
741
	if (pfLock->fl_flags &
L
Linus Torvalds 已提交
742
	    (~(FL_POSIX | FL_FLOCK | FL_SLEEP | FL_ACCESS | FL_LEASE)))
743
		cFYI(1, "Unknown lock flags 0x%x", pfLock->fl_flags);
L
Linus Torvalds 已提交
744 745

	if (pfLock->fl_type == F_WRLCK) {
746
		cFYI(1, "F_WRLCK ");
L
Linus Torvalds 已提交
747 748
		numLock = 1;
	} else if (pfLock->fl_type == F_UNLCK) {
749
		cFYI(1, "F_UNLCK");
L
Linus Torvalds 已提交
750
		numUnlock = 1;
751 752
		/* Check if unlock includes more than
		one lock range */
L
Linus Torvalds 已提交
753
	} else if (pfLock->fl_type == F_RDLCK) {
754
		cFYI(1, "F_RDLCK");
L
Linus Torvalds 已提交
755 756 757
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else if (pfLock->fl_type == F_EXLCK) {
758
		cFYI(1, "F_EXLCK");
L
Linus Torvalds 已提交
759 760
		numLock = 1;
	} else if (pfLock->fl_type == F_SHLCK) {
761
		cFYI(1, "F_SHLCK");
L
Linus Torvalds 已提交
762 763 764
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else
765
		cFYI(1, "Unknown type of lock");
L
Linus Torvalds 已提交
766

767
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
768
	tcon = tlink_tcon(((struct cifsFileInfo *)file->private_data)->tlink);
L
Linus Torvalds 已提交
769 770

	if (file->private_data == NULL) {
771
		rc = -EBADF;
L
Linus Torvalds 已提交
772
		FreeXid(xid);
773
		return rc;
L
Linus Torvalds 已提交
774
	}
775 776
	netfid = ((struct cifsFileInfo *)file->private_data)->netfid;

777 778
	if ((tcon->ses->capabilities & CAP_UNIX) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
S
Steve French 已提交
779
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
780
		posix_locking = 1;
781 782 783
	/* 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
784
	if (IS_GETLK(cmd)) {
S
Steve French 已提交
785
		if (posix_locking) {
786
			int posix_lock_type;
S
Steve French 已提交
787
			if (lockType & LOCKING_ANDX_SHARED_LOCK)
788 789 790
				posix_lock_type = CIFS_RDLCK;
			else
				posix_lock_type = CIFS_WRLCK;
791
			rc = CIFSSMBPosixLock(xid, tcon, netfid, 1 /* get */,
792
					length,	pfLock,
793 794 795 796 797 798
					posix_lock_type, wait_flag);
			FreeXid(xid);
			return rc;
		}

		/* BB we could chain these into one lock request BB */
799
		rc = CIFSSMBLock(xid, tcon, netfid, length, pfLock->fl_start,
800
				 0, 1, lockType, 0 /* wait flag */ );
L
Linus Torvalds 已提交
801
		if (rc == 0) {
802
			rc = CIFSSMBLock(xid, tcon, netfid, length,
L
Linus Torvalds 已提交
803 804 805 806 807
					 pfLock->fl_start, 1 /* numUnlock */ ,
					 0 /* numLock */ , lockType,
					 0 /* wait flag */ );
			pfLock->fl_type = F_UNLCK;
			if (rc != 0)
808 809
				cERROR(1, "Error unlocking previously locked "
					   "range %d during test of lock", rc);
L
Linus Torvalds 已提交
810 811 812 813
			rc = 0;

		} else {
			/* if rc == ERR_SHARING_VIOLATION ? */
814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830
			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)
831
						cERROR(1, "Error unlocking "
832
						"previously locked range %d "
833
						"during test of lock", rc);
834 835 836 837 838 839
					rc = 0;
				} else {
					pfLock->fl_type = F_WRLCK;
					rc = 0;
				}
			}
L
Linus Torvalds 已提交
840 841 842 843 844
		}

		FreeXid(xid);
		return rc;
	}
J
[CIFS]  
Jeremy Allison 已提交
845 846 847 848 849 850 851 852 853

	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) {
854
		int posix_lock_type;
S
Steve French 已提交
855
		if (lockType & LOCKING_ANDX_SHARED_LOCK)
856 857 858
			posix_lock_type = CIFS_RDLCK;
		else
			posix_lock_type = CIFS_WRLCK;
859

S
Steve French 已提交
860
		if (numUnlock == 1)
861
			posix_lock_type = CIFS_UNLCK;
J
[CIFS]  
Jeremy Allison 已提交
862

863
		rc = CIFSSMBPosixLock(xid, tcon, netfid, 0 /* set */,
864
				      length, pfLock,
865
				      posix_lock_type, wait_flag);
J
[CIFS]  
Jeremy Allison 已提交
866
	} else {
867
		struct cifsFileInfo *fid = file->private_data;
J
[CIFS]  
Jeremy Allison 已提交
868 869

		if (numLock) {
870
			rc = CIFSSMBLock(xid, tcon, netfid, length,
S
Steve French 已提交
871
					pfLock->fl_start,
J
[CIFS]  
Jeremy Allison 已提交
872 873 874 875 876 877 878 879 880 881 882 883 884
					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;

885
			rc = 0;
886
			mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
887 888
			list_for_each_entry_safe(li, tmp, &fid->llist, llist) {
				if (pfLock->fl_start <= li->offset &&
S
Steve French 已提交
889
						(pfLock->fl_start + length) >=
890
						(li->offset + li->length)) {
891
					stored_rc = CIFSSMBLock(xid, tcon,
S
Steve French 已提交
892
							netfid,
J
[CIFS]  
Jeremy Allison 已提交
893
							li->length, li->offset,
894
							1, 0, li->type, false);
J
[CIFS]  
Jeremy Allison 已提交
895 896
					if (stored_rc)
						rc = stored_rc;
897 898 899 900
					else {
						list_del(&li->llist);
						kfree(li);
					}
J
[CIFS]  
Jeremy Allison 已提交
901 902
				}
			}
903
			mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
904 905 906
		}
	}

907
	if (pfLock->fl_flags & FL_POSIX)
L
Linus Torvalds 已提交
908 909 910 911 912
		posix_lock_file_wait(file, pfLock);
	FreeXid(xid);
	return rc;
}

913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946
/*
 * 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
947 948 949 950 951 952 953 954 955 956
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;
957
	struct cifsInodeInfo *cifsi = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
958

959
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
960

961 962
	/* 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
963 964 965

	if (file->private_data == NULL)
		return -EBADF;
966

967
	open_file = file->private_data;
968
	pTcon = tlink_tcon(open_file->tlink);
969

970 971 972 973
	rc = generic_write_checks(file, poffset, &write_size, 0);
	if (rc)
		return rc;

L
Linus Torvalds 已提交
974 975
	xid = GetXid();

976
	long_op = cifs_write_timeout(cifsi, *poffset);
L
Linus Torvalds 已提交
977 978 979 980 981 982 983 984 985 986 987
	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
988
			}
L
Linus Torvalds 已提交
989 990 991 992 993 994 995 996 997 998 999 1000
			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 */
1001
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
				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;
			}
1020 1021
		} else {
			cifs_update_eof(cifsi, *poffset, bytes_written);
L
Linus Torvalds 已提交
1022
			*poffset += bytes_written;
1023
		}
1024
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1025 1026 1027
				    15 seconds is plenty */
	}

1028
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1029 1030

	/* since the write may have blocked check these pointers again */
1031 1032
	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
1033 1034
/* Do not update local mtime - server will set its actual value on write
 *		inode->i_ctime = inode->i_mtime =
1035 1036 1037 1038 1039
 * 			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
1040
					*poffset);
1041
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1042
		}
S
Steve French 已提交
1043
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1044 1045 1046 1047 1048 1049
	}
	FreeXid(xid);
	return total_written;
}

static ssize_t cifs_write(struct file *file, const char *write_data,
N
Nick Piggin 已提交
1050
			  size_t write_size, loff_t *poffset)
L
Linus Torvalds 已提交
1051 1052 1053 1054 1055 1056 1057 1058
{
	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;
1059
	struct cifsInodeInfo *cifsi = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1060

1061
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1062

1063 1064
	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
1065 1066 1067

	if (file->private_data == NULL)
		return -EBADF;
1068
	open_file = file->private_data;
1069
	pTcon = tlink_tcon(open_file->tlink);
1070

L
Linus Torvalds 已提交
1071 1072
	xid = GetXid();

1073
	long_op = cifs_write_timeout(cifsi, *poffset);
L
Linus Torvalds 已提交
1074 1075 1076 1077 1078 1079 1080 1081 1082
	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
1083
			   while we blocked so return what we managed to
L
Linus Torvalds 已提交
1084 1085
			   write */
				return total_written;
S
Steve French 已提交
1086
			}
L
Linus Torvalds 已提交
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
			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
1097
				   reopen_file not to flush data to
L
Linus Torvalds 已提交
1098
				   server now */
1099
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
1100 1101 1102
				if (rc != 0)
					break;
			}
S
Steve French 已提交
1103 1104
			if (experimEnabled || (pTcon->ses->server &&
				((pTcon->ses->server->secMode &
S
Steve French 已提交
1105
				(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED))
S
Steve French 已提交
1106
				== 0))) {
1107 1108 1109
				struct kvec iov[2];
				unsigned int len;

1110
				len = min((size_t)cifs_sb->wsize,
1111 1112 1113 1114 1115
					  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;
1116
				rc = CIFSSMBWrite2(xid, pTcon,
1117
						open_file->netfid, len,
1118
						*poffset, &bytes_written,
1119
						iov, 1, long_op);
1120
			} else
1121 1122 1123 1124 1125 1126 1127
				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
1128 1129 1130 1131 1132 1133 1134 1135
		}
		if (rc || (bytes_written == 0)) {
			if (total_written)
				break;
			else {
				FreeXid(xid);
				return rc;
			}
1136 1137
		} else {
			cifs_update_eof(cifsi, *poffset, bytes_written);
L
Linus Torvalds 已提交
1138
			*poffset += bytes_written;
1139
		}
1140
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1141 1142 1143
				    15 seconds is plenty */
	}

1144
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1145 1146

	/* since the write may have blocked check these pointers again */
1147
	if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
1148
/*BB We could make this contingent on superblock ATIME flag too */
1149 1150 1151 1152 1153 1154 1155 1156
/*		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
1157
		}
1158
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1159 1160 1161 1162 1163
	}
	FreeXid(xid);
	return total_written;
}

S
Steve French 已提交
1164
#ifdef CONFIG_CIFS_EXPERIMENTAL
1165 1166
struct cifsFileInfo *find_readable_file(struct cifsInodeInfo *cifs_inode,
					bool fsuid_only)
S
Steve French 已提交
1167 1168
{
	struct cifsFileInfo *open_file = NULL;
1169 1170 1171 1172 1173
	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
1174 1175 1176 1177 1178 1179 1180 1181

	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;
1182 1183
		if (fsuid_only && open_file->uid != current_fsuid())
			continue;
S
Steve French 已提交
1184 1185 1186 1187 1188
		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 */
1189
				cifsFileInfo_get(open_file);
S
Steve French 已提交
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
				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

1203 1204
struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode,
					bool fsuid_only)
1205 1206
{
	struct cifsFileInfo *open_file;
1207
	struct cifs_sb_info *cifs_sb = CIFS_SB(cifs_inode->vfs_inode.i_sb);
1208
	bool any_available = false;
1209
	int rc;
1210

1211 1212 1213 1214
	/* 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
1215
	if (cifs_inode == NULL) {
1216
		cERROR(1, "Null inode passed to cifs_writeable_file");
1217 1218 1219 1220
		dump_stack();
		return NULL;
	}

1221 1222 1223 1224
	/* only filter by fsuid on multiuser mounts */
	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MULTIUSER))
		fsuid_only = false;

1225
	read_lock(&GlobalSMBSeslock);
1226
refind_writable:
1227
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1228 1229 1230 1231 1232
		if (open_file->closePend)
			continue;
		if (!any_available && open_file->pid != current->tgid)
			continue;
		if (fsuid_only && open_file->uid != current_fsuid())
1233 1234 1235 1236
			continue;
		if (open_file->pfile &&
		    ((open_file->pfile->f_flags & O_RDWR) ||
		     (open_file->pfile->f_flags & O_WRONLY))) {
1237
			cifsFileInfo_get(open_file);
1238 1239 1240 1241 1242 1243

			if (!open_file->invalidHandle) {
				/* found a good writable file */
				read_unlock(&GlobalSMBSeslock);
				return open_file;
			}
S
Steve French 已提交
1244

1245
			read_unlock(&GlobalSMBSeslock);
1246
			/* Had to unlock since following call can block */
1247
			rc = cifs_reopen_file(open_file->pfile, false);
S
Steve French 已提交
1248
			if (!rc) {
1249 1250 1251 1252
				if (!open_file->closePend)
					return open_file;
				else { /* start over in case this was deleted */
				       /* since the list could be modified */
1253
					read_lock(&GlobalSMBSeslock);
1254
					cifsFileInfo_put(open_file);
1255
					goto refind_writable;
1256 1257
				}
			}
1258 1259 1260 1261 1262 1263 1264 1265

			/* 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 */
1266
			cFYI(1, "wp failed on reopen file");
1267 1268 1269
			read_lock(&GlobalSMBSeslock);
			/* can not use this handle, no write
			   pending on this one after all */
1270
			cifsFileInfo_put(open_file);
S
Steve French 已提交
1271

1272 1273 1274 1275 1276 1277 1278 1279 1280
			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. */
1281 1282
		}
	}
1283 1284 1285 1286 1287
	/* couldn't find useable FH with same pid, try any available */
	if (!any_available) {
		any_available = true;
		goto refind_writable;
	}
1288 1289 1290 1291
	read_unlock(&GlobalSMBSeslock);
	return NULL;
}

L
Linus Torvalds 已提交
1292 1293 1294 1295 1296 1297 1298 1299 1300
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;
1301
	struct cifsFileInfo *open_file;
L
Linus Torvalds 已提交
1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325

	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
1326
		to = (unsigned)(mapping->host->i_size - offset);
L
Linus Torvalds 已提交
1327

1328
	open_file = find_writable_file(CIFS_I(mapping->host), false);
1329 1330 1331
	if (open_file) {
		bytes_written = cifs_write(open_file->pfile, write_data,
					   to-from, &offset);
1332
		cifsFileInfo_put(open_file);
L
Linus Torvalds 已提交
1333
		/* Does mm or vfs already set times? */
1334
		inode->i_atime = inode->i_mtime = current_fs_time(inode->i_sb);
1335
		if ((bytes_written > 0) && (offset))
1336
			rc = 0;
1337 1338
		else if (bytes_written < 0)
			rc = bytes_written;
1339
	} else {
1340
		cFYI(1, "No writeable filehandles for inode");
L
Linus Torvalds 已提交
1341 1342 1343 1344 1345 1346 1347 1348
		rc = -EIO;
	}

	kunmap(page);
	return rc;
}

static int cifs_writepages(struct address_space *mapping,
1349
			   struct writeback_control *wbc)
L
Linus Torvalds 已提交
1350
{
1351 1352 1353 1354 1355
	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;
1356
	pgoff_t end;
1357
	pgoff_t index;
S
Steve French 已提交
1358 1359
	int range_whole = 0;
	struct kvec *iov;
1360
	int len;
1361 1362 1363 1364
	int n_iov = 0;
	pgoff_t next;
	int nr_pages;
	__u64 offset = 0;
1365
	struct cifsFileInfo *open_file;
1366
	struct cifsTconInfo *tcon;
1367
	struct cifsInodeInfo *cifsi = CIFS_I(mapping->host);
1368 1369 1370 1371
	struct page *page;
	struct pagevec pvec;
	int rc = 0;
	int scanned = 0;
1372
	int xid, long_op;
L
Linus Torvalds 已提交
1373

1374 1375 1376 1377 1378 1379 1380 1381 1382
	/*
	 * 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;
	}

1383
	cifs_sb = CIFS_SB(mapping->host->i_sb);
1384

1385 1386 1387 1388 1389 1390 1391
	/*
	 * 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);

1392
	iov = kmalloc(32 * sizeof(struct kvec), GFP_KERNEL);
S
Steve French 已提交
1393
	if (iov == NULL)
1394 1395
		return generic_writepages(mapping, wbc);

1396
	/*
1397 1398 1399
	 * 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?
1400
	 */
1401
	open_file = find_writable_file(CIFS_I(mapping->host), false);
1402
	if (!open_file) {
1403
		kfree(iov);
1404 1405 1406
		return generic_writepages(mapping, wbc);
	}

1407
	tcon = tlink_tcon(open_file->tlink);
1408 1409 1410 1411
	if (!experimEnabled && tcon->ses->server->secMode &
			(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED)) {
		cifsFileInfo_put(open_file);
		return generic_writepages(mapping, wbc);
1412
	}
1413
	cifsFileInfo_put(open_file);
1414

L
Linus Torvalds 已提交
1415 1416
	xid = GetXid();

1417
	pagevec_init(&pvec, 0);
1418
	if (wbc->range_cyclic) {
1419
		index = mapping->writeback_index; /* Start from prev offset */
1420 1421 1422 1423 1424 1425
		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;
1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
		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
1453
			else if (!trylock_page(page))
1454 1455 1456 1457 1458 1459 1460
				break;

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

1461
			if (!wbc->range_cyclic && page->index > end) {
1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476
				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) ||
1477
					!clear_page_dirty_for_io(page)) {
1478 1479 1480
				unlock_page(page);
				break;
			}
1481

1482 1483 1484 1485 1486 1487
			/*
			 * This actually clears the dirty bit in the radix tree.
			 * See cifs_writepage() for more commentary.
			 */
			set_page_writeback(page);

1488 1489 1490
			if (page_offset(page) >= mapping->host->i_size) {
				done = 1;
				unlock_page(page);
1491
				end_page_writeback(page);
1492 1493 1494
				break;
			}

1495 1496 1497 1498 1499
			/*
			 * BB can we get rid of this?  pages are held by pvec
			 */
			page_cache_get(page);

1500 1501 1502
			len = min(mapping->host->i_size - page_offset(page),
				  (loff_t)PAGE_CACHE_SIZE);

1503 1504 1505
			/* reserve iov[0] for the smb header */
			n_iov++;
			iov[n_iov].iov_base = kmap(page);
1506 1507
			iov[n_iov].iov_len = len;
			bytes_to_write += len;
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517

			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) {
1518 1519
			open_file = find_writable_file(CIFS_I(mapping->host),
							false);
1520
			if (!open_file) {
1521
				cERROR(1, "No writable handles for inode");
1522
				rc = -EBADF;
S
Steve French 已提交
1523
			} else {
1524
				long_op = cifs_write_timeout(cifsi, offset);
1525
				rc = CIFSSMBWrite2(xid, tcon, open_file->netfid,
1526 1527
						   bytes_to_write, offset,
						   &bytes_written, iov, n_iov,
1528
						   long_op);
1529
				cifsFileInfo_put(open_file);
1530
				cifs_update_eof(cifsi, offset, bytes_written);
1531
			}
1532

1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
			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);
1544
			}
1545

1546 1547
			for (i = 0; i < n_iov; i++) {
				page = pvec.pages[first + i];
1548 1549 1550 1551
				/* 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
1552 1553
				when page marked as error */
				if (rc)
1554
					SetPageError(page);
1555 1556
				kunmap(page);
				unlock_page(page);
1557
				end_page_writeback(page);
1558 1559 1560 1561 1562
				page_cache_release(page);
			}
			if ((wbc->nr_to_write -= n_iov) <= 0)
				done = 1;
			index = next;
1563 1564 1565 1566
		} else
			/* Need to re-find the pages we skipped */
			index = pvec.pages[0]->index + 1;

1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
		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;
	}
1578
	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1579 1580
		mapping->writeback_index = index;

L
Linus Torvalds 已提交
1581
	FreeXid(xid);
1582
	kfree(iov);
L
Linus Torvalds 已提交
1583 1584 1585
	return rc;
}

S
Steve French 已提交
1586
static int cifs_writepage(struct page *page, struct writeback_control *wbc)
L
Linus Torvalds 已提交
1587 1588 1589 1590 1591 1592 1593
{
	int rc = -EFAULT;
	int xid;

	xid = GetXid();
/* BB add check for wbc flags */
	page_cache_get(page);
S
Steve French 已提交
1594
	if (!PageUptodate(page))
1595
		cFYI(1, "ppw - page not up to date");
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606

	/*
	 * 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
1607
	set_page_writeback(page);
L
Linus Torvalds 已提交
1608 1609 1610
	rc = cifs_partialpagewrite(page, 0, PAGE_CACHE_SIZE);
	SetPageUptodate(page); /* BB add check for error and Clearuptodate? */
	unlock_page(page);
1611 1612
	end_page_writeback(page);
	page_cache_release(page);
L
Linus Torvalds 已提交
1613 1614 1615 1616
	FreeXid(xid);
	return rc;
}

N
Nick Piggin 已提交
1617 1618 1619
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
1620
{
N
Nick Piggin 已提交
1621 1622
	int rc;
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
1623

1624 1625
	cFYI(1, "write_end for page %p from pos %lld with %d bytes",
		 page, pos, copied);
N
Nick Piggin 已提交
1626

1627 1628 1629 1630 1631
	if (PageChecked(page)) {
		if (copied == len)
			SetPageUptodate(page);
		ClearPageChecked(page);
	} else if (!PageUptodate(page) && copied == PAGE_CACHE_SIZE)
N
Nick Piggin 已提交
1632
		SetPageUptodate(page);
S
Steve French 已提交
1633

L
Linus Torvalds 已提交
1634
	if (!PageUptodate(page)) {
N
Nick Piggin 已提交
1635 1636 1637 1638 1639
		char *page_data;
		unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
		int xid;

		xid = GetXid();
L
Linus Torvalds 已提交
1640 1641 1642 1643 1644 1645
		/* 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
1646 1647
		rc = cifs_write(file, page_data + offset, copied, &pos);
		/* if (rc < 0) should we set writebehind rc? */
L
Linus Torvalds 已提交
1648
		kunmap(page);
N
Nick Piggin 已提交
1649 1650

		FreeXid(xid);
S
Steve French 已提交
1651
	} else {
N
Nick Piggin 已提交
1652 1653
		rc = copied;
		pos += copied;
L
Linus Torvalds 已提交
1654 1655 1656
		set_page_dirty(page);
	}

N
Nick Piggin 已提交
1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
	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
1667 1668 1669
	return rc;
}

1670
int cifs_fsync(struct file *file, int datasync)
L
Linus Torvalds 已提交
1671 1672 1673
{
	int xid;
	int rc = 0;
1674
	struct cifsTconInfo *tcon;
1675
	struct cifsFileInfo *smbfile = file->private_data;
1676
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1677 1678 1679

	xid = GetXid();

1680
	cFYI(1, "Sync file - name: %s datasync: 0x%x",
1681
		file->f_path.dentry->d_name.name, datasync);
1682

1683 1684 1685
	rc = filemap_write_and_wait(inode->i_mapping);
	if (rc == 0) {
		rc = CIFS_I(inode)->write_behind_rc;
L
Linus Torvalds 已提交
1686
		CIFS_I(inode)->write_behind_rc = 0;
1687
		tcon = tlink_tcon(smbfile->tlink);
1688
		if (!rc && tcon && smbfile &&
S
Steve French 已提交
1689
		   !(CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC))
1690
			rc = CIFSSMBFlush(xid, tcon, smbfile->netfid);
1691
	}
1692

L
Linus Torvalds 已提交
1693 1694 1695 1696
	FreeXid(xid);
	return rc;
}

1697
/* static void cifs_sync_page(struct page *page)
L
Linus Torvalds 已提交
1698 1699 1700 1701 1702 1703 1704
{
	struct address_space *mapping;
	struct inode *inode;
	unsigned long index = page->index;
	unsigned int rpages = 0;
	int rc = 0;

1705
	cFYI(1, "sync page %p", page);
L
Linus Torvalds 已提交
1706 1707 1708 1709 1710
	mapping = page->mapping;
	if (!mapping)
		return 0;
	inode = mapping->host;
	if (!inode)
1711
		return; */
L
Linus Torvalds 已提交
1712

S
Steve French 已提交
1713
/*	fill in rpages then
L
Linus Torvalds 已提交
1714 1715
	result = cifs_pagein_inode(inode, index, rpages); */ /* BB finish */

1716
/*	cFYI(1, "rpages is %d for sync page of Index %ld", rpages, index);
L
Linus Torvalds 已提交
1717

1718
#if 0
L
Linus Torvalds 已提交
1719 1720 1721
	if (rc < 0)
		return rc;
	return 0;
1722
#endif
L
Linus Torvalds 已提交
1723 1724 1725 1726 1727 1728
} */

/*
 * As file closes, flush all cached write data for this inode checking
 * for write behind errors.
 */
1729
int cifs_flush(struct file *file, fl_owner_t id)
L
Linus Torvalds 已提交
1730
{
S
Steve French 已提交
1731
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
	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);
1744 1745 1746
	/* 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
1747
		CIFS_I(inode)->write_behind_rc = 0;
1748
	}
1749

1750
	cFYI(1, "Flush inode %p file %p rc %d", inode, file, rc);
L
Linus Torvalds 已提交
1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770

	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();
1771
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1772 1773

	if (file->private_data == NULL) {
1774
		rc = -EBADF;
L
Linus Torvalds 已提交
1775
		FreeXid(xid);
1776
		return rc;
L
Linus Torvalds 已提交
1777
	}
1778
	open_file = file->private_data;
1779
	pTcon = tlink_tcon(open_file->tlink);
L
Linus Torvalds 已提交
1780

S
Steve French 已提交
1781
	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
1782
		cFYI(1, "attempting read on write only file instance");
S
Steve French 已提交
1783

L
Linus Torvalds 已提交
1784 1785 1786
	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
1787
		current_read_size = min_t(const int, read_size - total_read,
L
Linus Torvalds 已提交
1788 1789 1790 1791
					  cifs_sb->rsize);
		rc = -EAGAIN;
		smb_read_data = NULL;
		while (rc == -EAGAIN) {
1792
			int buf_type = CIFS_NO_BUFFER;
S
Steve French 已提交
1793
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1794
			    (!open_file->closePend)) {
1795
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1796 1797 1798
				if (rc != 0)
					break;
			}
1799
			rc = CIFSSMBRead(xid, pTcon,
1800 1801 1802 1803
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
L
Linus Torvalds 已提交
1804 1805
			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
			if (smb_read_data) {
1806 1807 1808 1809
				if (copy_to_user(current_offset,
						smb_read_data +
						4 /* RFC1001 length field */ +
						le16_to_cpu(pSMBr->DataOffset),
S
Steve French 已提交
1810
						bytes_read))
1811 1812
					rc = -EFAULT;

S
Steve French 已提交
1813
				if (buf_type == CIFS_SMALL_BUFFER)
1814
					cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1815
				else if (buf_type == CIFS_LARGE_BUFFER)
1816
					cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827
				smb_read_data = NULL;
			}
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1828
			cifs_stats_bytes_read(pTcon, bytes_read);
L
Linus Torvalds 已提交
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
			*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;
1849
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1850 1851

	xid = GetXid();
1852
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1853 1854

	if (file->private_data == NULL) {
1855
		rc = -EBADF;
L
Linus Torvalds 已提交
1856
		FreeXid(xid);
1857
		return rc;
L
Linus Torvalds 已提交
1858
	}
1859
	open_file = file->private_data;
1860
	pTcon = tlink_tcon(open_file->tlink);
L
Linus Torvalds 已提交
1861 1862

	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
1863
		cFYI(1, "attempting read on write only file instance");
L
Linus Torvalds 已提交
1864

S
Steve French 已提交
1865
	for (total_read = 0, current_offset = read_data;
L
Linus Torvalds 已提交
1866 1867 1868 1869
	     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);
1870 1871
		/* 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
1872
		if ((pTcon->ses) &&
1873 1874 1875 1876
			!(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
1877 1878
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
1879
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1880
			    (!open_file->closePend)) {
1881
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1882 1883 1884
				if (rc != 0)
					break;
			}
1885
			rc = CIFSSMBRead(xid, pTcon,
1886 1887 1888 1889
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &current_offset,
					 &buf_type);
L
Linus Torvalds 已提交
1890 1891 1892 1893 1894 1895 1896 1897 1898
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1899
			cifs_stats_bytes_read(pTcon, total_read);
L
Linus Torvalds 已提交
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
			*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
1912
	rc = cifs_revalidate_file(file);
L
Linus Torvalds 已提交
1913
	if (rc) {
1914
		cFYI(1, "Validation prior to mmap failed, error=%d", rc);
L
Linus Torvalds 已提交
1915 1916 1917 1918 1919 1920 1921 1922 1923
		FreeXid(xid);
		return rc;
	}
	rc = generic_file_mmap(file, vma);
	FreeXid(xid);
	return rc;
}


S
Steve French 已提交
1924
static void cifs_copy_cache_pages(struct address_space *mapping,
N
Nick Piggin 已提交
1925
	struct list_head *pages, int bytes_read, char *data)
L
Linus Torvalds 已提交
1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
{
	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
1937
		if (add_to_page_cache_lru(page, mapping, page->index,
L
Linus Torvalds 已提交
1938 1939
				      GFP_KERNEL)) {
			page_cache_release(page);
1940
			cFYI(1, "Add page cache failed");
1941 1942
			data += PAGE_CACHE_SIZE;
			bytes_read -= PAGE_CACHE_SIZE;
L
Linus Torvalds 已提交
1943 1944
			continue;
		}
J
Jeff Layton 已提交
1945
		page_cache_release(page);
L
Linus Torvalds 已提交
1946

S
Steve French 已提交
1947
		target = kmap_atomic(page, KM_USER0);
L
Linus Torvalds 已提交
1948 1949 1950 1951

		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
1952
			memset(target + bytes_read, 0,
L
Linus Torvalds 已提交
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964
			       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;
1965 1966 1967

		/* add page to FS-Cache */
		cifs_readpage_to_fscache(mapping->host, page);
L
Linus Torvalds 已提交
1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
	}
	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
1981
	unsigned int bytes_read = 0;
S
Steve French 已提交
1982
	unsigned int read_size, i;
L
Linus Torvalds 已提交
1983 1984 1985
	char *smb_read_data = NULL;
	struct smb_com_read_rsp *pSMBr;
	struct cifsFileInfo *open_file;
1986
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1987 1988 1989

	xid = GetXid();
	if (file->private_data == NULL) {
1990
		rc = -EBADF;
L
Linus Torvalds 已提交
1991
		FreeXid(xid);
1992
		return rc;
L
Linus Torvalds 已提交
1993
	}
1994
	open_file = file->private_data;
1995
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
1996
	pTcon = tlink_tcon(open_file->tlink);
1997

1998 1999 2000 2001 2002 2003 2004 2005 2006
	/*
	 * 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;

2007
	cFYI(DBG2, "rpages: num pages %d", num_pages);
L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
	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
2021
		expected_index =
L
Linus Torvalds 已提交
2022
			list_entry(page_list->prev, struct page, lru)->index;
S
Steve French 已提交
2023
		list_for_each_entry_reverse(tmp_page, page_list, lru) {
L
Linus Torvalds 已提交
2024 2025 2026 2027
			if (tmp_page->index == expected_index) {
				contig_pages++;
				expected_index++;
			} else
S
Steve French 已提交
2028
				break;
L
Linus Torvalds 已提交
2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039
		}
		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);
2040 2041
		cFYI(DBG2, "rpages: read size 0x%x  contiguous pages %d",
				read_size, contig_pages);
L
Linus Torvalds 已提交
2042 2043
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
2044
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
2045
			    (!open_file->closePend)) {
2046
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
2047 2048 2049 2050
				if (rc != 0)
					break;
			}

2051
			rc = CIFSSMBRead(xid, pTcon,
2052 2053 2054 2055
					 open_file->netfid,
					 read_size, offset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
2056
			/* BB more RC checks ? */
S
Steve French 已提交
2057
			if (rc == -EAGAIN) {
L
Linus Torvalds 已提交
2058
				if (smb_read_data) {
S
Steve French 已提交
2059
					if (buf_type == CIFS_SMALL_BUFFER)
2060
						cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2061
					else if (buf_type == CIFS_LARGE_BUFFER)
2062
						cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2063 2064 2065 2066 2067
					smb_read_data = NULL;
				}
			}
		}
		if ((rc < 0) || (smb_read_data == NULL)) {
2068
			cFYI(1, "Read error in readpages: %d", rc);
L
Linus Torvalds 已提交
2069 2070
			break;
		} else if (bytes_read > 0) {
2071
			task_io_account_read(bytes_read);
L
Linus Torvalds 已提交
2072 2073 2074
			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
2075
				le16_to_cpu(pSMBr->DataOffset));
L
Linus Torvalds 已提交
2076 2077

			i +=  bytes_read >> PAGE_CACHE_SHIFT;
2078
			cifs_stats_bytes_read(pTcon, bytes_read);
S
Steve French 已提交
2079
			if ((bytes_read & PAGE_CACHE_MASK) != bytes_read) {
L
Linus Torvalds 已提交
2080 2081
				i++; /* account for partial page */

S
Steve French 已提交
2082
				/* server copy of file can have smaller size
L
Linus Torvalds 已提交
2083
				   than client */
S
Steve French 已提交
2084 2085
				/* 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
2086 2087
				   we will hit it on next read */

O
OGAWA Hirofumi 已提交
2088
				/* break; */
L
Linus Torvalds 已提交
2089 2090
			}
		} else {
2091
			cFYI(1, "No bytes read (%d) at offset %lld . "
2092
				"Cleaning remaining pages from readahead list",
2093
				bytes_read, offset);
S
Steve French 已提交
2094
			/* BB turn off caching and do new lookup on
L
Linus Torvalds 已提交
2095 2096 2097 2098
			   file size at server? */
			break;
		}
		if (smb_read_data) {
S
Steve French 已提交
2099
			if (buf_type == CIFS_SMALL_BUFFER)
2100
				cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2101
			else if (buf_type == CIFS_LARGE_BUFFER)
2102
				cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2103 2104 2105 2106 2107 2108 2109
			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
2110
		if (buf_type == CIFS_SMALL_BUFFER)
2111
			cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2112
		else if (buf_type == CIFS_LARGE_BUFFER)
2113
			cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2114
		smb_read_data = NULL;
S
Steve French 已提交
2115
	}
L
Linus Torvalds 已提交
2116

2117
read_complete:
L
Linus Torvalds 已提交
2118 2119 2120 2121 2122 2123 2124 2125 2126 2127
	FreeXid(xid);
	return rc;
}

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

2128 2129 2130 2131 2132
	/* 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
2133 2134 2135
	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
2136

L
Linus Torvalds 已提交
2137
	rc = cifs_read(file, read_data, PAGE_CACHE_SIZE, poffset);
S
Steve French 已提交
2138

L
Linus Torvalds 已提交
2139 2140 2141
	if (rc < 0)
		goto io_error;
	else
2142
		cFYI(1, "Bytes read %d", rc);
S
Steve French 已提交
2143

2144 2145
	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
2146

L
Linus Torvalds 已提交
2147 2148 2149 2150 2151
	if (PAGE_CACHE_SIZE > rc)
		memset(read_data + rc, 0, PAGE_CACHE_SIZE - rc);

	flush_dcache_page(page);
	SetPageUptodate(page);
2152 2153 2154 2155

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

L
Linus Torvalds 已提交
2156
	rc = 0;
S
Steve French 已提交
2157

L
Linus Torvalds 已提交
2158
io_error:
S
Steve French 已提交
2159
	kunmap(page);
L
Linus Torvalds 已提交
2160
	page_cache_release(page);
2161 2162

read_complete:
L
Linus Torvalds 已提交
2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174
	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) {
2175
		rc = -EBADF;
L
Linus Torvalds 已提交
2176
		FreeXid(xid);
2177
		return rc;
L
Linus Torvalds 已提交
2178 2179
	}

2180 2181
	cFYI(1, "readpage %p at offset %d 0x%x\n",
		 page, (int)offset, (int)offset);
L
Linus Torvalds 已提交
2182 2183 2184 2185 2186 2187 2188 2189 2190

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

	unlock_page(page);

	FreeXid(xid);
	return rc;
}

2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
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
2210 2211 2212
/* 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
2213
   refreshing the inode only on increases in the file size
L
Linus Torvalds 已提交
2214 2215
   but this is tricky to do without racing with writebehind
   page caching in the current Linux kernel design */
2216
bool is_size_safe_to_change(struct cifsInodeInfo *cifsInode, __u64 end_of_file)
L
Linus Torvalds 已提交
2217
{
2218
	if (!cifsInode)
2219
		return true;
2220

2221 2222
	if (is_inode_writable(cifsInode)) {
		/* This inode is open for write at least once */
2223 2224 2225
		struct cifs_sb_info *cifs_sb;

		cifs_sb = CIFS_SB(cifsInode->vfs_inode.i_sb);
S
Steve French 已提交
2226
		if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DIRECT_IO) {
S
Steve French 已提交
2227
			/* since no page cache to corrupt on directio
2228
			we can change size safely */
2229
			return true;
2230 2231
		}

S
Steve French 已提交
2232
		if (i_size_read(&cifsInode->vfs_inode) < end_of_file)
2233
			return true;
2234

2235
		return false;
2236
	} else
2237
		return true;
L
Linus Torvalds 已提交
2238 2239
}

N
Nick Piggin 已提交
2240 2241 2242
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
2243
{
N
Nick Piggin 已提交
2244 2245
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	loff_t offset = pos & (PAGE_CACHE_SIZE - 1);
2246 2247 2248 2249
	loff_t page_start = pos & PAGE_MASK;
	loff_t i_size;
	struct page *page;
	int rc = 0;
N
Nick Piggin 已提交
2250

2251
	cFYI(1, "write_begin from %lld len %d", (long long)pos, len);
N
Nick Piggin 已提交
2252

2253
	page = grab_cache_page_write_begin(mapping, index, flags);
2254 2255 2256 2257
	if (!page) {
		rc = -ENOMEM;
		goto out;
	}
2258

2259 2260
	if (PageUptodate(page))
		goto out;
2261

2262 2263 2264 2265 2266 2267 2268
	/*
	 * 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;
2269

2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
	/*
	 * 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
2293

2294 2295 2296 2297 2298 2299 2300
	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);
2301 2302 2303 2304
	} 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
2305
		   this will be written out by write_end so is fine */
L
Linus Torvalds 已提交
2306
	}
2307 2308 2309
out:
	*pagep = page;
	return rc;
L
Linus Torvalds 已提交
2310 2311
}

2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
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);
}

2328
void cifs_oplock_break(struct work_struct *work)
2329 2330 2331
{
	struct cifsFileInfo *cfile = container_of(work, struct cifsFileInfo,
						  oplock_break);
2332
	struct inode *inode = cfile->dentry->d_inode;
2333 2334 2335 2336
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	int rc, waitrc = 0;

	if (inode && S_ISREG(inode->i_mode)) {
S
Steve French 已提交
2337
		if (cinode->clientCanCacheRead)
2338
			break_lease(inode, O_RDONLY);
S
Steve French 已提交
2339
		else
2340
			break_lease(inode, O_WRONLY);
2341 2342 2343 2344 2345 2346 2347 2348 2349
		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;
2350
		cFYI(1, "Oplock flush inode %p rc %d", inode, rc);
2351 2352 2353 2354 2355 2356 2357 2358 2359
	}

	/*
	 * 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) {
2360 2361
		rc = CIFSSMBLock(0, tlink_tcon(cfile->tlink), cfile->netfid, 0,
				 0, 0, 0, LOCKING_ANDX_OPLOCK_RELEASE, false);
2362
		cFYI(1, "Oplock release rc = %d", rc);
2363
	}
2364 2365 2366 2367 2368 2369 2370 2371 2372 2373

	/*
	 * 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);
2374 2375
}

2376
void cifs_oplock_break_get(struct cifsFileInfo *cfile)
2377
{
2378
	cifs_sb_active(cfile->dentry->d_sb);
2379 2380 2381
	cifsFileInfo_get(cfile);
}

2382
void cifs_oplock_break_put(struct cifsFileInfo *cfile)
2383 2384
{
	cifsFileInfo_put(cfile);
2385
	cifs_sb_deactive(cfile->dentry->d_sb);
2386 2387
}

2388
const struct address_space_operations cifs_addr_ops = {
L
Linus Torvalds 已提交
2389 2390 2391
	.readpage = cifs_readpage,
	.readpages = cifs_readpages,
	.writepage = cifs_writepage,
2392
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2393 2394
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
L
Linus Torvalds 已提交
2395
	.set_page_dirty = __set_page_dirty_nobuffers,
2396 2397
	.releasepage = cifs_release_page,
	.invalidatepage = cifs_invalidate_page,
L
Linus Torvalds 已提交
2398 2399 2400
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};
D
Dave Kleikamp 已提交
2401 2402 2403 2404 2405 2406

/*
 * 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.
 */
2407
const struct address_space_operations cifs_addr_ops_smallbuf = {
D
Dave Kleikamp 已提交
2408 2409 2410
	.readpage = cifs_readpage,
	.writepage = cifs_writepage,
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2411 2412
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
D
Dave Kleikamp 已提交
2413
	.set_page_dirty = __set_page_dirty_nobuffers,
2414 2415
	.releasepage = cifs_release_page,
	.invalidatepage = cifs_invalidate_page,
D
Dave Kleikamp 已提交
2416 2417 2418
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};