file.c 64.7 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 43 44 45 46 47 48 49 50 51 52 53 54 55 56
#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"

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);
	}

57 58 59
	return (READ_CONTROL | FILE_WRITE_ATTRIBUTES | FILE_READ_ATTRIBUTES |
		FILE_WRITE_EA | FILE_APPEND_DATA | FILE_WRITE_DATA |
		FILE_READ_DATA);
60
}
61

62 63 64
static inline fmode_t cifs_posix_convert_flags(unsigned int flags)
{
	fmode_t posix_flags = 0;
65

66 67 68 69 70 71 72 73 74 75 76 77 78 79
	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;
80 81 82 83
	if (flags & O_DSYNC)
		posix_flags |= (fmode_t)O_DSYNC;
	if (flags & __O_SYNC)
		posix_flags |= (fmode_t)__O_SYNC;
84 85 86 87 88 89 90 91
	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
92 93 94 95 96 97 98 99 100 101
}

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

108
/* all arguments to this function must be checked for validity in caller */
109 110 111 112 113
static inline int
cifs_posix_open_inode_helper(struct inode *inode, struct file *file,
			     struct cifsInodeInfo *pCifsInode,
			     struct cifsFileInfo *pCifsFile, __u32 oplock,
			     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;
}

S
Steve French 已提交
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
static struct cifsFileInfo *
cifs_fill_filedata(struct file *file)
{
	struct list_head *tmp;
	struct cifsFileInfo *pCifsFile = NULL;
	struct cifsInodeInfo *pCifsInode = NULL;

	/* search inode for this file and fill in file->private_data */
	pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
	read_lock(&GlobalSMBSeslock);
	list_for_each(tmp, &pCifsInode->openFileList) {
		pCifsFile = list_entry(tmp, struct cifsFileInfo, flist);
		if ((pCifsFile->pfile == NULL) &&
		    (pCifsFile->pid == current->tgid)) {
			/* mode set in cifs_create */

			/* needed for writepage */
			pCifsFile->pfile = file;
			file->private_data = pCifsFile;
			break;
		}
	}
	read_unlock(&GlobalSMBSeslock);

	if (file->private_data != NULL) {
		return pCifsFile;
	} else if ((file->f_flags & O_CREAT) && (file->f_flags & O_EXCL))
193 194
			cERROR(1, "could not find file instance for "
				   "new file %p", file);
S
Steve French 已提交
195 196 197
	return NULL;
}

L
Linus Torvalds 已提交
198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
/* all arguments to this function must be checked for validity in caller */
static inline int cifs_open_inode_helper(struct inode *inode, struct file *file,
	struct cifsInodeInfo *pCifsInode, struct cifsFileInfo *pCifsFile,
	struct cifsTconInfo *pTcon, int *oplock, FILE_ALL_INFO *buf,
	char *full_path, int xid)
{
	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 */
216
	temp = cifs_NTtimeToUnix(buf->LastWriteTime);
217 218
	if (timespec_equal(&file->f_path.dentry->d_inode->i_mtime, &temp) &&
			   (file->f_path.dentry->d_inode->i_size ==
L
Linus Torvalds 已提交
219
			    (loff_t)le64_to_cpu(buf->EndOfFile))) {
220
		cFYI(1, "inode unchanged on server");
L
Linus Torvalds 已提交
221
	} else {
222
		if (file->f_path.dentry->d_inode->i_mapping) {
S
Steve French 已提交
223 224
			/* BB no need to lock inode until after invalidate
			since namei code should already have it locked? */
225 226 227
			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;
L
Linus Torvalds 已提交
228
		}
229 230
		cFYI(1, "invalidating remote inode since open detected it "
			 "changed");
231
		invalidate_remote_inode(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
232 233 234
	}

client_can_cache:
235
	if (pTcon->unix_ext)
236
		rc = cifs_get_inode_info_unix(&file->f_path.dentry->d_inode,
L
Linus Torvalds 已提交
237 238
			full_path, inode->i_sb, xid);
	else
239
		rc = cifs_get_inode_info(&file->f_path.dentry->d_inode,
240
			full_path, buf, inode->i_sb, xid, NULL);
L
Linus Torvalds 已提交
241 242

	if ((*oplock & 0xF) == OPLOCK_EXCLUSIVE) {
243 244
		pCifsInode->clientCanCacheAll = true;
		pCifsInode->clientCanCacheRead = true;
245 246
		cFYI(1, "Exclusive Oplock granted on inode %p",
			 file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
247
	} else if ((*oplock & 0xF) == OPLOCK_READ)
248
		pCifsInode->clientCanCacheRead = true;
L
Linus Torvalds 已提交
249 250 251 252 253 254 255

	return rc;
}

int cifs_open(struct inode *inode, struct file *file)
{
	int rc = -EACCES;
256 257
	int xid;
	__u32 oplock;
L
Linus Torvalds 已提交
258
	struct cifs_sb_info *cifs_sb;
259
	struct cifsTconInfo *tcon;
L
Linus Torvalds 已提交
260 261 262 263 264 265 266 267 268 269 270
	struct cifsFileInfo *pCifsFile;
	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);
271
	tcon = cifs_sb->tcon;
L
Linus Torvalds 已提交
272

273
	pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
S
Steve French 已提交
274 275
	pCifsFile = cifs_fill_filedata(file);
	if (pCifsFile) {
276
		rc = 0;
277
		FreeXid(xid);
278
		return rc;
S
Steve French 已提交
279
	}
L
Linus Torvalds 已提交
280

281
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
282
	if (full_path == NULL) {
283
		rc = -ENOMEM;
L
Linus Torvalds 已提交
284
		FreeXid(xid);
285
		return rc;
L
Linus Torvalds 已提交
286 287
	}

288 289
	cFYI(1, "inode = 0x%p file flags are 0x%x for %s",
		 inode, file->f_flags, full_path);
290 291 292 293 294 295

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

296 297
	if (!tcon->broken_posix_open && tcon->unix_ext &&
	    (tcon->ses->capabilities & CAP_UNIX) &&
298 299 300 301
	    (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 */
302
		rc = cifs_posix_open(full_path, &inode, file->f_path.mnt,
303 304 305
				     cifs_sb->mnt_file_mode /* ignored */,
				     oflags, &oplock, &netfid, xid);
		if (rc == 0) {
306
			cFYI(1, "posix open succeeded");
307 308 309
			/* no need for special case handling of setting mode
			   on read only files needed here */

S
Steve French 已提交
310
			pCifsFile = cifs_fill_filedata(file);
311 312 313
			cifs_posix_open_inode_helper(inode, file, pCifsInode,
						     pCifsFile, oplock, netfid);
			goto out;
314 315
		} else if ((rc == -EINVAL) || (rc == -EOPNOTSUPP)) {
			if (tcon->ses->serverNOS)
316
				cERROR(1, "server %s of type %s returned"
317 318 319 320
					   " unexpected error on SMB posix open"
					   ", disabling posix open support."
					   " Check if server update available.",
					   tcon->ses->serverName,
321
					   tcon->ses->serverNOS);
322
			tcon->broken_posix_open = true;
323 324 325
		} else if ((rc != -EIO) && (rc != -EREMOTE) &&
			 (rc != -EOPNOTSUPP)) /* path not found or net err */
			goto out;
326 327
		/* else fallthrough to retry open the old way on network i/o
		   or DFS errors */
328 329
	}

L
Linus Torvalds 已提交
330 331 332 333
	desiredAccess = cifs_convert_flags(file->f_flags);

/*********************************************************************
 *  open flag mapping table:
S
Steve French 已提交
334
 *
L
Linus Torvalds 已提交
335
 *	POSIX Flag            CIFS Disposition
S
Steve French 已提交
336
 *	----------            ----------------
L
Linus Torvalds 已提交
337 338 339 340 341 342 343
 *	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
344
 *	FILE_SUPERSEDE (ie create whether or not file exists although
L
Linus Torvalds 已提交
345 346 347 348
 *	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
349
 *?  O_SYNC is a reasonable match to CIFS writethrough flag
L
Linus Torvalds 已提交
350 351 352 353 354 355 356 357 358 359 360 361 362 363
 *?  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
364 365
	/* 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
366 367 368 369 370 371
	   able to simply do a filemap_fdatawrite/filemap_fdatawait first */
	buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
	if (!buf) {
		rc = -ENOMEM;
		goto out;
	}
372 373

	if (cifs_sb->tcon->ses->capabilities & CAP_NT_SMBS)
374
		rc = CIFSSMBOpen(xid, tcon, full_path, disposition,
375
			 desiredAccess, CREATE_NOT_DIR, &netfid, &oplock, buf,
376 377
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags
				 & CIFS_MOUNT_MAP_SPECIAL_CHR);
378 379 380
	else
		rc = -EIO; /* no NT SMB support fall into legacy open below */

381 382
	if (rc == -EIO) {
		/* Old server, try legacy style OpenX */
383
		rc = SMBLegacyOpen(xid, tcon, full_path, disposition,
384 385 386 387
			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
388
	if (rc) {
389
		cFYI(1, "cifs_open returned 0x%x", rc);
L
Linus Torvalds 已提交
390 391
		goto out;
	}
392

J
Jeff Layton 已提交
393 394 395
	pCifsFile = cifs_new_fileinfo(inode, netfid, file, file->f_path.mnt,
					file->f_flags);
	file->private_data = pCifsFile;
L
Linus Torvalds 已提交
396 397 398 399 400
	if (file->private_data == NULL) {
		rc = -ENOMEM;
		goto out;
	}

401 402
	rc = cifs_open_inode_helper(inode, file, pCifsInode, pCifsFile, tcon,
				    &oplock, buf, full_path, xid);
L
Linus Torvalds 已提交
403

S
Steve French 已提交
404
	if (oplock & CIFS_CREATE_ACTION) {
L
Linus Torvalds 已提交
405 406
		/* time to set mode which we can not set earlier due to
		   problems creating new read-only files */
407
		if (tcon->unix_ext) {
408 409 410 411 412 413 414 415 416
			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,
			};
417 418 419
			CIFSSMBUnixSetPathInfo(xid, tcon, full_path, &args,
					       cifs_sb->local_nls,
					       cifs_sb->mnt_cifs_flags &
420
						CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
421 422 423 424 425 426 427 428 429 430
		}
	}

out:
	kfree(buf);
	kfree(full_path);
	FreeXid(xid);
	return rc;
}

A
Adrian Bunk 已提交
431
/* Try to reacquire byte range locks that were released when session */
L
Linus Torvalds 已提交
432 433 434 435 436 437 438 439 440 441
/* 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;
}

442
static int cifs_reopen_file(struct file *file, bool can_flush)
L
Linus Torvalds 已提交
443 444
{
	int rc = -EACCES;
445 446
	int xid;
	__u32 oplock;
L
Linus Torvalds 已提交
447
	struct cifs_sb_info *cifs_sb;
448
	struct cifsTconInfo *tcon;
L
Linus Torvalds 已提交
449 450
	struct cifsFileInfo *pCifsFile;
	struct cifsInodeInfo *pCifsInode;
S
Steve French 已提交
451
	struct inode *inode;
L
Linus Torvalds 已提交
452 453 454 455 456
	char *full_path = NULL;
	int desiredAccess;
	int disposition = FILE_OPEN;
	__u16 netfid;

S
Steve French 已提交
457
	if (file->private_data)
L
Linus Torvalds 已提交
458
		pCifsFile = (struct cifsFileInfo *)file->private_data;
S
Steve French 已提交
459
	else
L
Linus Torvalds 已提交
460 461 462
		return -EBADF;

	xid = GetXid();
463
	mutex_lock(&pCifsFile->fh_mutex);
464
	if (!pCifsFile->invalidHandle) {
465
		mutex_unlock(&pCifsFile->fh_mutex);
466
		rc = 0;
L
Linus Torvalds 已提交
467
		FreeXid(xid);
468
		return rc;
L
Linus Torvalds 已提交
469 470
	}

471
	if (file->f_path.dentry == NULL) {
472
		cERROR(1, "no valid name if dentry freed");
473 474 475 476 477 478
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
	}

	inode = file->f_path.dentry->d_inode;
S
Steve French 已提交
479
	if (inode == NULL) {
480
		cERROR(1, "inode not valid");
481 482 483
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
L
Linus Torvalds 已提交
484
	}
485

L
Linus Torvalds 已提交
486
	cifs_sb = CIFS_SB(inode->i_sb);
487
	tcon = cifs_sb->tcon;
488

L
Linus Torvalds 已提交
489 490 491 492
/* 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 */
493
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
494
	if (full_path == NULL) {
495 496
		rc = -ENOMEM;
reopen_error_exit:
497
		mutex_unlock(&pCifsFile->fh_mutex);
L
Linus Torvalds 已提交
498
		FreeXid(xid);
499
		return rc;
L
Linus Torvalds 已提交
500 501
	}

502 503
	cFYI(1, "inode = 0x%p file flags 0x%x for %s",
		 inode, file->f_flags, full_path);
L
Linus Torvalds 已提交
504 505 506 507

	if (oplockEnabled)
		oplock = REQ_OPLOCK;
	else
508
		oplock = 0;
L
Linus Torvalds 已提交
509

510 511 512 513 514
	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 */
515
		rc = cifs_posix_open(full_path, NULL, file->f_path.mnt,
516 517 518
				     cifs_sb->mnt_file_mode /* ignored */,
				     oflags, &oplock, &netfid, xid);
		if (rc == 0) {
519
			cFYI(1, "posix reopen succeeded");
520 521 522 523 524 525 526 527
			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
528
	/* Can not refresh inode by passing in file_info buf to be returned
S
Steve French 已提交
529 530
	   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
531 532 533
	   and server version of file size can be stale. If we knew for sure
	   that inode was not dirty locally we could do this */

534
	rc = CIFSSMBOpen(xid, tcon, full_path, disposition, desiredAccess,
L
Linus Torvalds 已提交
535
			 CREATE_NOT_DIR, &netfid, &oplock, NULL,
S
Steve French 已提交
536
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags &
537
				CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
538
	if (rc) {
539
		mutex_unlock(&pCifsFile->fh_mutex);
540 541
		cFYI(1, "cifs_open returned 0x%x", rc);
		cFYI(1, "oplock: %d", oplock);
L
Linus Torvalds 已提交
542
	} else {
543
reopen_success:
L
Linus Torvalds 已提交
544
		pCifsFile->netfid = netfid;
545
		pCifsFile->invalidHandle = false;
546
		mutex_unlock(&pCifsFile->fh_mutex);
L
Linus Torvalds 已提交
547 548 549
		pCifsInode = CIFS_I(inode);
		if (pCifsInode) {
			if (can_flush) {
550 551 552
				rc = filemap_write_and_wait(inode->i_mapping);
				if (rc != 0)
					CIFS_I(inode)->write_behind_rc = rc;
L
Linus Torvalds 已提交
553 554
			/* temporarily disable caching while we
			   go to server to get inode info */
555 556
				pCifsInode->clientCanCacheAll = false;
				pCifsInode->clientCanCacheRead = false;
557
				if (tcon->unix_ext)
L
Linus Torvalds 已提交
558 559 560 561 562
					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,
563
						xid, NULL);
L
Linus Torvalds 已提交
564 565 566 567 568 569 570
			} /* 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) {
571 572
				pCifsInode->clientCanCacheAll = true;
				pCifsInode->clientCanCacheRead = true;
573 574
				cFYI(1, "Exclusive Oplock granted on inode %p",
					 file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
575
			} else if ((oplock & 0xF) == OPLOCK_READ) {
576 577
				pCifsInode->clientCanCacheRead = true;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
578
			} else {
579 580
				pCifsInode->clientCanCacheRead = false;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
581 582 583 584 585 586 587 588 589 590 591 592
			}
			cifs_relock_file(pCifsFile);
		}
	}
	kfree(full_path);
	FreeXid(xid);
	return rc;
}

int cifs_close(struct inode *inode, struct file *file)
{
	int rc = 0;
593
	int xid, timeout;
L
Linus Torvalds 已提交
594 595 596 597 598 599 600 601 602 603
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	struct cifsFileInfo *pSMBFile =
		(struct cifsFileInfo *)file->private_data;

	xid = GetXid();

	cifs_sb = CIFS_SB(inode->i_sb);
	pTcon = cifs_sb->tcon;
	if (pSMBFile) {
J
[CIFS]  
Jeremy Allison 已提交
604
		struct cifsLockInfo *li, *tmp;
605
		write_lock(&GlobalSMBSeslock);
606
		pSMBFile->closePend = true;
L
Linus Torvalds 已提交
607 608 609
		if (pTcon) {
			/* no sense reconnecting to close a file that is
			   already closed */
S
Steve French 已提交
610
			if (!pTcon->need_reconnect) {
611
				write_unlock(&GlobalSMBSeslock);
612
				timeout = 2;
613
				while ((atomic_read(&pSMBFile->count) != 1)
614
					&& (timeout <= 2048)) {
615 616 617 618 619
					/* 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
620
					but this should give enough time to
621
					clear the socket */
622
					cFYI(DBG2, "close delay, write pending");
623 624
					msleep(timeout);
					timeout *= 4;
625
				}
626 627 628
				if (!pTcon->need_reconnect &&
				    !pSMBFile->invalidHandle)
					rc = CIFSSMBClose(xid, pTcon,
L
Linus Torvalds 已提交
629
						  pSMBFile->netfid);
630 631 632 633
			} else
				write_unlock(&GlobalSMBSeslock);
		} else
			write_unlock(&GlobalSMBSeslock);
J
[CIFS]  
Jeremy Allison 已提交
634 635 636

		/* Delete any outstanding lock records.
		   We'll lose them when the file is closed anyway. */
637
		mutex_lock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
638 639 640 641
		list_for_each_entry_safe(li, tmp, &pSMBFile->llist, llist) {
			list_del(&li->llist);
			kfree(li);
		}
642
		mutex_unlock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
643

644
		write_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
645 646
		list_del(&pSMBFile->flist);
		list_del(&pSMBFile->tlist);
647
		write_unlock(&GlobalSMBSeslock);
648
		cifsFileInfo_put(file->private_data);
L
Linus Torvalds 已提交
649 650 651 652
		file->private_data = NULL;
	} else
		rc = -EBADF;

653
	read_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
654
	if (list_empty(&(CIFS_I(inode)->openFileList))) {
655
		cFYI(1, "closing last open instance for inode %p", inode);
L
Linus Torvalds 已提交
656 657
		/* 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 */
658 659
		CIFS_I(inode)->clientCanCacheRead = false;
		CIFS_I(inode)->clientCanCacheAll  = false;
L
Linus Torvalds 已提交
660
	}
661
	read_unlock(&GlobalSMBSeslock);
S
Steve French 已提交
662
	if ((rc == 0) && CIFS_I(inode)->write_behind_rc)
L
Linus Torvalds 已提交
663 664 665 666 667 668 669 670 671 672 673 674 675
		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;
	struct cifsFileInfo *pCFileStruct =
	    (struct cifsFileInfo *)file->private_data;
	char *ptmp;

676
	cFYI(1, "Closedir inode = 0x%p", inode);
L
Linus Torvalds 已提交
677 678 679 680 681

	xid = GetXid();

	if (pCFileStruct) {
		struct cifsTconInfo *pTcon;
S
Steve French 已提交
682 683
		struct cifs_sb_info *cifs_sb =
			CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
684 685 686

		pTcon = cifs_sb->tcon;

687
		cFYI(1, "Freeing private data in close dir");
688
		write_lock(&GlobalSMBSeslock);
689 690 691
		if (!pCFileStruct->srch_inf.endOfSearch &&
		    !pCFileStruct->invalidHandle) {
			pCFileStruct->invalidHandle = true;
692
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
693
			rc = CIFSFindClose(xid, pTcon, pCFileStruct->netfid);
694 695
			cFYI(1, "Closing uncompleted readdir with rc %d",
				 rc);
L
Linus Torvalds 已提交
696 697
			/* not much we can do if it fails anyway, ignore rc */
			rc = 0;
698 699
		} else
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
700 701
		ptmp = pCFileStruct->srch_inf.ntwrk_buf_start;
		if (ptmp) {
702
			cFYI(1, "closedir free smb buf in srch struct");
L
Linus Torvalds 已提交
703
			pCFileStruct->srch_inf.ntwrk_buf_start = NULL;
S
Steve French 已提交
704
			if (pCFileStruct->srch_inf.smallBuf)
705 706 707
				cifs_small_buf_release(ptmp);
			else
				cifs_buf_release(ptmp);
L
Linus Torvalds 已提交
708 709 710 711 712 713 714 715 716
		}
		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
717 718 719
static int store_file_lock(struct cifsFileInfo *fid, __u64 len,
				__u64 offset, __u8 lockType)
{
S
Steve French 已提交
720 721
	struct cifsLockInfo *li =
		kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
J
[CIFS]  
Jeremy Allison 已提交
722 723 724 725 726
	if (li == NULL)
		return -ENOMEM;
	li->offset = offset;
	li->length = len;
	li->type = lockType;
727
	mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
728
	list_add(&li->llist, &fid->llist);
729
	mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
730 731 732
	return 0;
}

L
Linus Torvalds 已提交
733 734 735 736 737 738
int cifs_lock(struct file *file, int cmd, struct file_lock *pfLock)
{
	int rc, xid;
	__u32 numLock = 0;
	__u32 numUnlock = 0;
	__u64 length;
739
	bool wait_flag = false;
L
Linus Torvalds 已提交
740
	struct cifs_sb_info *cifs_sb;
741
	struct cifsTconInfo *tcon;
742 743
	__u16 netfid;
	__u8 lockType = LOCKING_ANDX_LARGE_FILES;
744
	bool posix_locking = 0;
L
Linus Torvalds 已提交
745 746 747 748 749

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

750
	cFYI(1, "Lock parm: 0x%x flockflags: "
L
Linus Torvalds 已提交
751
		 "0x%x flocktype: 0x%x start: %lld end: %lld",
S
Steve French 已提交
752
		cmd, pfLock->fl_flags, pfLock->fl_type, pfLock->fl_start,
753
		pfLock->fl_end);
L
Linus Torvalds 已提交
754 755

	if (pfLock->fl_flags & FL_POSIX)
756
		cFYI(1, "Posix");
L
Linus Torvalds 已提交
757
	if (pfLock->fl_flags & FL_FLOCK)
758
		cFYI(1, "Flock");
L
Linus Torvalds 已提交
759
	if (pfLock->fl_flags & FL_SLEEP) {
760
		cFYI(1, "Blocking lock");
761
		wait_flag = true;
L
Linus Torvalds 已提交
762 763
	}
	if (pfLock->fl_flags & FL_ACCESS)
764 765
		cFYI(1, "Process suspended by mandatory locking - "
			 "not implemented yet");
L
Linus Torvalds 已提交
766
	if (pfLock->fl_flags & FL_LEASE)
767
		cFYI(1, "Lease on file - not implemented yet");
S
Steve French 已提交
768
	if (pfLock->fl_flags &
L
Linus Torvalds 已提交
769
	    (~(FL_POSIX | FL_FLOCK | FL_SLEEP | FL_ACCESS | FL_LEASE)))
770
		cFYI(1, "Unknown lock flags 0x%x", pfLock->fl_flags);
L
Linus Torvalds 已提交
771 772

	if (pfLock->fl_type == F_WRLCK) {
773
		cFYI(1, "F_WRLCK ");
L
Linus Torvalds 已提交
774 775
		numLock = 1;
	} else if (pfLock->fl_type == F_UNLCK) {
776
		cFYI(1, "F_UNLCK");
L
Linus Torvalds 已提交
777
		numUnlock = 1;
778 779
		/* Check if unlock includes more than
		one lock range */
L
Linus Torvalds 已提交
780
	} else if (pfLock->fl_type == F_RDLCK) {
781
		cFYI(1, "F_RDLCK");
L
Linus Torvalds 已提交
782 783 784
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else if (pfLock->fl_type == F_EXLCK) {
785
		cFYI(1, "F_EXLCK");
L
Linus Torvalds 已提交
786 787
		numLock = 1;
	} else if (pfLock->fl_type == F_SHLCK) {
788
		cFYI(1, "F_SHLCK");
L
Linus Torvalds 已提交
789 790 791
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else
792
		cFYI(1, "Unknown type of lock");
L
Linus Torvalds 已提交
793

794
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
795
	tcon = cifs_sb->tcon;
L
Linus Torvalds 已提交
796 797

	if (file->private_data == NULL) {
798
		rc = -EBADF;
L
Linus Torvalds 已提交
799
		FreeXid(xid);
800
		return rc;
L
Linus Torvalds 已提交
801
	}
802 803
	netfid = ((struct cifsFileInfo *)file->private_data)->netfid;

804 805
	if ((tcon->ses->capabilities & CAP_UNIX) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
S
Steve French 已提交
806
	    ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
807
		posix_locking = 1;
808 809 810
	/* 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
811
	if (IS_GETLK(cmd)) {
S
Steve French 已提交
812
		if (posix_locking) {
813
			int posix_lock_type;
S
Steve French 已提交
814
			if (lockType & LOCKING_ANDX_SHARED_LOCK)
815 816 817
				posix_lock_type = CIFS_RDLCK;
			else
				posix_lock_type = CIFS_WRLCK;
818
			rc = CIFSSMBPosixLock(xid, tcon, netfid, 1 /* get */,
819
					length,	pfLock,
820 821 822 823 824 825
					posix_lock_type, wait_flag);
			FreeXid(xid);
			return rc;
		}

		/* BB we could chain these into one lock request BB */
826
		rc = CIFSSMBLock(xid, tcon, netfid, length, pfLock->fl_start,
827
				 0, 1, lockType, 0 /* wait flag */ );
L
Linus Torvalds 已提交
828
		if (rc == 0) {
829
			rc = CIFSSMBLock(xid, tcon, netfid, length,
L
Linus Torvalds 已提交
830 831 832 833 834
					 pfLock->fl_start, 1 /* numUnlock */ ,
					 0 /* numLock */ , lockType,
					 0 /* wait flag */ );
			pfLock->fl_type = F_UNLCK;
			if (rc != 0)
835 836
				cERROR(1, "Error unlocking previously locked "
					   "range %d during test of lock", rc);
L
Linus Torvalds 已提交
837 838 839 840
			rc = 0;

		} else {
			/* if rc == ERR_SHARING_VIOLATION ? */
841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857
			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)
858
						cERROR(1, "Error unlocking "
859
						"previously locked range %d "
860
						"during test of lock", rc);
861 862 863 864 865 866
					rc = 0;
				} else {
					pfLock->fl_type = F_WRLCK;
					rc = 0;
				}
			}
L
Linus Torvalds 已提交
867 868 869 870 871
		}

		FreeXid(xid);
		return rc;
	}
J
[CIFS]  
Jeremy Allison 已提交
872 873 874 875 876 877 878 879 880

	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) {
881
		int posix_lock_type;
S
Steve French 已提交
882
		if (lockType & LOCKING_ANDX_SHARED_LOCK)
883 884 885
			posix_lock_type = CIFS_RDLCK;
		else
			posix_lock_type = CIFS_WRLCK;
886

S
Steve French 已提交
887
		if (numUnlock == 1)
888
			posix_lock_type = CIFS_UNLCK;
J
[CIFS]  
Jeremy Allison 已提交
889

890
		rc = CIFSSMBPosixLock(xid, tcon, netfid, 0 /* set */,
891
				      length, pfLock,
892
				      posix_lock_type, wait_flag);
J
[CIFS]  
Jeremy Allison 已提交
893
	} else {
S
Steve French 已提交
894 895
		struct cifsFileInfo *fid =
			(struct cifsFileInfo *)file->private_data;
J
[CIFS]  
Jeremy Allison 已提交
896 897

		if (numLock) {
898
			rc = CIFSSMBLock(xid, tcon, netfid, length,
S
Steve French 已提交
899
					pfLock->fl_start,
J
[CIFS]  
Jeremy Allison 已提交
900 901 902 903 904 905 906 907 908 909 910 911 912
					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;

913
			rc = 0;
914
			mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
915 916
			list_for_each_entry_safe(li, tmp, &fid->llist, llist) {
				if (pfLock->fl_start <= li->offset &&
S
Steve French 已提交
917
						(pfLock->fl_start + length) >=
918
						(li->offset + li->length)) {
919
					stored_rc = CIFSSMBLock(xid, tcon,
S
Steve French 已提交
920
							netfid,
J
[CIFS]  
Jeremy Allison 已提交
921
							li->length, li->offset,
922
							1, 0, li->type, false);
J
[CIFS]  
Jeremy Allison 已提交
923 924 925 926 927 928 929
					if (stored_rc)
						rc = stored_rc;

					list_del(&li->llist);
					kfree(li);
				}
			}
930
			mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
931 932 933
		}
	}

934
	if (pfLock->fl_flags & FL_POSIX)
L
Linus Torvalds 已提交
935 936 937 938 939
		posix_lock_file_wait(file, pfLock);
	FreeXid(xid);
	return rc;
}

940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973
/*
 * 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
974 975 976 977 978 979 980 981 982 983
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;
984
	struct cifsInodeInfo *cifsi = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
985

986
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
987 988 989

	pTcon = cifs_sb->tcon;

990 991
	/* 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
992 993 994

	if (file->private_data == NULL)
		return -EBADF;
995
	open_file = (struct cifsFileInfo *) file->private_data;
996

997 998 999 1000
	rc = generic_write_checks(file, poffset, &write_size, 0);
	if (rc)
		return rc;

L
Linus Torvalds 已提交
1001 1002
	xid = GetXid();

1003
	long_op = cifs_write_timeout(cifsi, *poffset);
L
Linus Torvalds 已提交
1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
	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
1015
			}
L
Linus Torvalds 已提交
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
			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 */
1028
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
				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;
			}
1047 1048
		} else {
			cifs_update_eof(cifsi, *poffset, bytes_written);
L
Linus Torvalds 已提交
1049
			*poffset += bytes_written;
1050
		}
1051
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1052 1053 1054
				    15 seconds is plenty */
	}

1055
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1056 1057

	/* since the write may have blocked check these pointers again */
1058 1059
	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
1060 1061
/* Do not update local mtime - server will set its actual value on write
 *		inode->i_ctime = inode->i_mtime =
1062 1063 1064 1065 1066
 * 			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
1067
					*poffset);
1068
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
1069
		}
S
Steve French 已提交
1070
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1071 1072 1073 1074 1075 1076
	}
	FreeXid(xid);
	return total_written;
}

static ssize_t cifs_write(struct file *file, const char *write_data,
N
Nick Piggin 已提交
1077
			  size_t write_size, loff_t *poffset)
L
Linus Torvalds 已提交
1078 1079 1080 1081 1082 1083 1084 1085
{
	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;
1086
	struct cifsInodeInfo *cifsi = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1087

1088
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1089 1090 1091

	pTcon = cifs_sb->tcon;

1092 1093
	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
1094 1095 1096

	if (file->private_data == NULL)
		return -EBADF;
1097
	open_file = (struct cifsFileInfo *)file->private_data;
1098

L
Linus Torvalds 已提交
1099 1100
	xid = GetXid();

1101
	long_op = cifs_write_timeout(cifsi, *poffset);
L
Linus Torvalds 已提交
1102 1103 1104 1105 1106 1107 1108 1109 1110
	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
1111
			   while we blocked so return what we managed to
L
Linus Torvalds 已提交
1112 1113
			   write */
				return total_written;
S
Steve French 已提交
1114
			}
L
Linus Torvalds 已提交
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
			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
1125
				   reopen_file not to flush data to
L
Linus Torvalds 已提交
1126
				   server now */
1127
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
1128 1129 1130
				if (rc != 0)
					break;
			}
S
Steve French 已提交
1131 1132
			if (experimEnabled || (pTcon->ses->server &&
				((pTcon->ses->server->secMode &
S
Steve French 已提交
1133
				(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED))
S
Steve French 已提交
1134
				== 0))) {
1135 1136 1137
				struct kvec iov[2];
				unsigned int len;

1138
				len = min((size_t)cifs_sb->wsize,
1139 1140 1141 1142 1143
					  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;
1144
				rc = CIFSSMBWrite2(xid, pTcon,
1145
						open_file->netfid, len,
1146
						*poffset, &bytes_written,
1147
						iov, 1, long_op);
1148
			} else
1149 1150 1151 1152 1153 1154 1155
				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
1156 1157 1158 1159 1160 1161 1162 1163
		}
		if (rc || (bytes_written == 0)) {
			if (total_written)
				break;
			else {
				FreeXid(xid);
				return rc;
			}
1164 1165
		} else {
			cifs_update_eof(cifsi, *poffset, bytes_written);
L
Linus Torvalds 已提交
1166
			*poffset += bytes_written;
1167
		}
1168
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1169 1170 1171
				    15 seconds is plenty */
	}

1172
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1173 1174

	/* since the write may have blocked check these pointers again */
1175
	if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
1176
/*BB We could make this contingent on superblock ATIME flag too */
1177 1178 1179 1180 1181 1182 1183 1184
/*		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
1185
		}
1186
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1187 1188 1189 1190 1191
	}
	FreeXid(xid);
	return total_written;
}

S
Steve French 已提交
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
#ifdef CONFIG_CIFS_EXPERIMENTAL
struct cifsFileInfo *find_readable_file(struct cifsInodeInfo *cifs_inode)
{
	struct cifsFileInfo *open_file = NULL;

	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;
		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 */
1209
				cifsFileInfo_get(open_file);
S
Steve French 已提交
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222
				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

1223
struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode)
1224 1225
{
	struct cifsFileInfo *open_file;
1226
	bool any_available = false;
1227
	int rc;
1228

1229 1230 1231 1232
	/* 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
1233
	if (cifs_inode == NULL) {
1234
		cERROR(1, "Null inode passed to cifs_writeable_file");
1235 1236 1237 1238
		dump_stack();
		return NULL;
	}

1239
	read_lock(&GlobalSMBSeslock);
1240
refind_writable:
1241
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1242 1243
		if (open_file->closePend ||
		    (!any_available && open_file->pid != current->tgid))
1244
			continue;
1245

1246 1247 1248
		if (open_file->pfile &&
		    ((open_file->pfile->f_flags & O_RDWR) ||
		     (open_file->pfile->f_flags & O_WRONLY))) {
1249
			cifsFileInfo_get(open_file);
1250 1251 1252 1253 1254 1255

			if (!open_file->invalidHandle) {
				/* found a good writable file */
				read_unlock(&GlobalSMBSeslock);
				return open_file;
			}
S
Steve French 已提交
1256

1257
			read_unlock(&GlobalSMBSeslock);
1258
			/* Had to unlock since following call can block */
1259
			rc = cifs_reopen_file(open_file->pfile, false);
S
Steve French 已提交
1260
			if (!rc) {
1261 1262 1263 1264
				if (!open_file->closePend)
					return open_file;
				else { /* start over in case this was deleted */
				       /* since the list could be modified */
1265
					read_lock(&GlobalSMBSeslock);
1266
					cifsFileInfo_put(open_file);
1267
					goto refind_writable;
1268 1269
				}
			}
1270 1271 1272 1273 1274 1275 1276 1277

			/* 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 */
1278
			cFYI(1, "wp failed on reopen file");
1279 1280 1281
			read_lock(&GlobalSMBSeslock);
			/* can not use this handle, no write
			   pending on this one after all */
1282
			cifsFileInfo_put(open_file);
S
Steve French 已提交
1283

1284 1285 1286 1287 1288 1289 1290 1291 1292
			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. */
1293 1294
		}
	}
1295 1296 1297 1298 1299
	/* couldn't find useable FH with same pid, try any available */
	if (!any_available) {
		any_available = true;
		goto refind_writable;
	}
1300 1301 1302 1303
	read_unlock(&GlobalSMBSeslock);
	return NULL;
}

L
Linus Torvalds 已提交
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
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 cifsTconInfo *pTcon;
	struct inode *inode;
1314
	struct cifsFileInfo *open_file;
L
Linus Torvalds 已提交
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339

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

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

	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
1340
		to = (unsigned)(mapping->host->i_size - offset);
L
Linus Torvalds 已提交
1341

1342 1343 1344 1345
	open_file = find_writable_file(CIFS_I(mapping->host));
	if (open_file) {
		bytes_written = cifs_write(open_file->pfile, write_data,
					   to-from, &offset);
1346
		cifsFileInfo_put(open_file);
L
Linus Torvalds 已提交
1347
		/* Does mm or vfs already set times? */
1348
		inode->i_atime = inode->i_mtime = current_fs_time(inode->i_sb);
1349
		if ((bytes_written > 0) && (offset))
1350
			rc = 0;
1351 1352
		else if (bytes_written < 0)
			rc = bytes_written;
1353
	} else {
1354
		cFYI(1, "No writeable filehandles for inode");
L
Linus Torvalds 已提交
1355 1356 1357 1358 1359 1360 1361 1362
		rc = -EIO;
	}

	kunmap(page);
	return rc;
}

static int cifs_writepages(struct address_space *mapping,
1363
			   struct writeback_control *wbc)
L
Linus Torvalds 已提交
1364
{
1365 1366 1367 1368 1369
	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;
1370
	pgoff_t end;
1371
	pgoff_t index;
S
Steve French 已提交
1372 1373
	int range_whole = 0;
	struct kvec *iov;
1374
	int len;
1375 1376 1377 1378
	int n_iov = 0;
	pgoff_t next;
	int nr_pages;
	__u64 offset = 0;
1379
	struct cifsFileInfo *open_file;
1380
	struct cifsInodeInfo *cifsi = CIFS_I(mapping->host);
1381 1382 1383 1384
	struct page *page;
	struct pagevec pvec;
	int rc = 0;
	int scanned = 0;
1385
	int xid, long_op;
L
Linus Torvalds 已提交
1386

1387
	cifs_sb = CIFS_SB(mapping->host->i_sb);
1388

1389 1390 1391 1392 1393 1394 1395
	/*
	 * 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);

S
Steve French 已提交
1396 1397 1398 1399
	if ((cifs_sb->tcon->ses) && (cifs_sb->tcon->ses->server))
		if (cifs_sb->tcon->ses->server->secMode &
				(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED))
			if (!experimEnabled)
1400
				return generic_writepages(mapping, wbc);
1401

1402
	iov = kmalloc(32 * sizeof(struct kvec), GFP_KERNEL);
S
Steve French 已提交
1403
	if (iov == NULL)
1404 1405 1406
		return generic_writepages(mapping, wbc);


1407 1408 1409 1410 1411 1412
	/*
	 * 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;
1413
		kfree(iov);
1414 1415 1416
		return 0;
	}

L
Linus Torvalds 已提交
1417 1418
	xid = GetXid();

1419
	pagevec_init(&pvec, 0);
1420
	if (wbc->range_cyclic) {
1421
		index = mapping->writeback_index; /* Start from prev offset */
1422 1423 1424 1425 1426 1427
		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;
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 1453 1454
		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
1455
			else if (!trylock_page(page))
1456 1457 1458 1459 1460 1461 1462
				break;

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

1463
			if (!wbc->range_cyclic && page->index > end) {
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478
				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) ||
1479
					!clear_page_dirty_for_io(page)) {
1480 1481 1482
				unlock_page(page);
				break;
			}
1483

1484 1485 1486 1487 1488 1489
			/*
			 * This actually clears the dirty bit in the radix tree.
			 * See cifs_writepage() for more commentary.
			 */
			set_page_writeback(page);

1490 1491 1492
			if (page_offset(page) >= mapping->host->i_size) {
				done = 1;
				unlock_page(page);
1493
				end_page_writeback(page);
1494 1495 1496
				break;
			}

1497 1498 1499 1500 1501
			/*
			 * BB can we get rid of this?  pages are held by pvec
			 */
			page_cache_get(page);

1502 1503 1504
			len = min(mapping->host->i_size - page_offset(page),
				  (loff_t)PAGE_CACHE_SIZE);

1505 1506 1507
			/* reserve iov[0] for the smb header */
			n_iov++;
			iov[n_iov].iov_base = kmap(page);
1508 1509
			iov[n_iov].iov_len = len;
			bytes_to_write += len;
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519

			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) {
1520 1521 1522 1523 1524 1525
			/* Search for a writable handle every time we call
			 * CIFSSMBWrite2.  We can't rely on the last handle
			 * we used to still be valid
			 */
			open_file = find_writable_file(CIFS_I(mapping->host));
			if (!open_file) {
1526
				cERROR(1, "No writable handles for inode");
1527
				rc = -EBADF;
S
Steve French 已提交
1528
			} else {
1529
				long_op = cifs_write_timeout(cifsi, offset);
1530 1531 1532 1533
				rc = CIFSSMBWrite2(xid, cifs_sb->tcon,
						   open_file->netfid,
						   bytes_to_write, offset,
						   &bytes_written, iov, n_iov,
1534
						   long_op);
1535
				cifsFileInfo_put(open_file);
1536 1537
				cifs_update_eof(cifsi, offset, bytes_written);

1538
				if (rc || bytes_written < bytes_to_write) {
1539 1540
					cERROR(1, "Write2 ret %d, wrote %d",
						  rc, bytes_written);
1541 1542
					/* BB what if continued retry is
					   requested via mount flags? */
1543 1544 1545 1546
					if (rc == -ENOSPC)
						set_bit(AS_ENOSPC, &mapping->flags);
					else
						set_bit(AS_EIO, &mapping->flags);
1547 1548 1549 1550
				} else {
					cifs_stats_bytes_written(cifs_sb->tcon,
								 bytes_written);
				}
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 1676 1677 1678 1679
	return rc;
}

int cifs_fsync(struct file *file, struct dentry *dentry, int datasync)
{
	int xid;
	int rc = 0;
1680 1681 1682
	struct cifsTconInfo *tcon;
	struct cifsFileInfo *smbfile =
		(struct cifsFileInfo *)file->private_data;
1683
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1684 1685 1686

	xid = GetXid();

1687 1688
	cFYI(1, "Sync file - name: %s datasync: 0x%x",
		dentry->d_name.name, datasync);
1689

1690 1691 1692
	rc = filemap_write_and_wait(inode->i_mapping);
	if (rc == 0) {
		rc = CIFS_I(inode)->write_behind_rc;
L
Linus Torvalds 已提交
1693
		CIFS_I(inode)->write_behind_rc = 0;
1694
		tcon = CIFS_SB(inode->i_sb)->tcon;
1695
		if (!rc && tcon && smbfile &&
S
Steve French 已提交
1696
		   !(CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_NOSSYNC))
1697
			rc = CIFSSMBFlush(xid, tcon, smbfile->netfid);
1698
	}
1699

L
Linus Torvalds 已提交
1700 1701 1702 1703
	FreeXid(xid);
	return rc;
}

1704
/* static void cifs_sync_page(struct page *page)
L
Linus Torvalds 已提交
1705 1706 1707 1708 1709 1710 1711
{
	struct address_space *mapping;
	struct inode *inode;
	unsigned long index = page->index;
	unsigned int rpages = 0;
	int rc = 0;

1712
	cFYI(1, "sync page %p", page);
L
Linus Torvalds 已提交
1713 1714 1715 1716 1717
	mapping = page->mapping;
	if (!mapping)
		return 0;
	inode = mapping->host;
	if (!inode)
1718
		return; */
L
Linus Torvalds 已提交
1719

S
Steve French 已提交
1720
/*	fill in rpages then
L
Linus Torvalds 已提交
1721 1722
	result = cifs_pagein_inode(inode, index, rpages); */ /* BB finish */

1723
/*	cFYI(1, "rpages is %d for sync page of Index %ld", rpages, index);
L
Linus Torvalds 已提交
1724

1725
#if 0
L
Linus Torvalds 已提交
1726 1727 1728
	if (rc < 0)
		return rc;
	return 0;
1729
#endif
L
Linus Torvalds 已提交
1730 1731 1732 1733 1734 1735
} */

/*
 * As file closes, flush all cached write data for this inode checking
 * for write behind errors.
 */
1736
int cifs_flush(struct file *file, fl_owner_t id)
L
Linus Torvalds 已提交
1737
{
S
Steve French 已提交
1738
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750
	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);
1751 1752 1753
	/* 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
1754
		CIFS_I(inode)->write_behind_rc = 0;
1755
	}
1756

1757
	cFYI(1, "Flush inode %p file %p rc %d", inode, file, rc);
L
Linus Torvalds 已提交
1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777

	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();
1778
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1779 1780 1781
	pTcon = cifs_sb->tcon;

	if (file->private_data == NULL) {
1782
		rc = -EBADF;
L
Linus Torvalds 已提交
1783
		FreeXid(xid);
1784
		return rc;
L
Linus Torvalds 已提交
1785 1786 1787
	}
	open_file = (struct cifsFileInfo *)file->private_data;

S
Steve French 已提交
1788
	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
1789
		cFYI(1, "attempting read on write only file instance");
S
Steve French 已提交
1790

L
Linus Torvalds 已提交
1791 1792 1793
	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
1794
		current_read_size = min_t(const int, read_size - total_read,
L
Linus Torvalds 已提交
1795 1796 1797 1798
					  cifs_sb->rsize);
		rc = -EAGAIN;
		smb_read_data = NULL;
		while (rc == -EAGAIN) {
1799
			int buf_type = CIFS_NO_BUFFER;
S
Steve French 已提交
1800
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1801
			    (!open_file->closePend)) {
1802
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1803 1804 1805
				if (rc != 0)
					break;
			}
1806
			rc = CIFSSMBRead(xid, pTcon,
1807 1808 1809 1810
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
L
Linus Torvalds 已提交
1811 1812
			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
			if (smb_read_data) {
1813 1814 1815 1816
				if (copy_to_user(current_offset,
						smb_read_data +
						4 /* RFC1001 length field */ +
						le16_to_cpu(pSMBr->DataOffset),
S
Steve French 已提交
1817
						bytes_read))
1818 1819
					rc = -EFAULT;

S
Steve French 已提交
1820
				if (buf_type == CIFS_SMALL_BUFFER)
1821
					cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1822
				else if (buf_type == CIFS_LARGE_BUFFER)
1823
					cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834
				smb_read_data = NULL;
			}
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1835
			cifs_stats_bytes_read(pTcon, bytes_read);
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
			*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;
1856
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1857 1858

	xid = GetXid();
1859
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1860 1861 1862
	pTcon = cifs_sb->tcon;

	if (file->private_data == NULL) {
1863
		rc = -EBADF;
L
Linus Torvalds 已提交
1864
		FreeXid(xid);
1865
		return rc;
L
Linus Torvalds 已提交
1866 1867 1868 1869
	}
	open_file = (struct cifsFileInfo *)file->private_data;

	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
1870
		cFYI(1, "attempting read on write only file instance");
L
Linus Torvalds 已提交
1871

S
Steve French 已提交
1872
	for (total_read = 0, current_offset = read_data;
L
Linus Torvalds 已提交
1873 1874 1875 1876
	     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);
1877 1878
		/* 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
1879
		if ((pTcon->ses) &&
1880 1881 1882 1883
			!(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
1884 1885
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
1886
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1887
			    (!open_file->closePend)) {
1888
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1889 1890 1891
				if (rc != 0)
					break;
			}
1892
			rc = CIFSSMBRead(xid, pTcon,
1893 1894 1895 1896
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &current_offset,
					 &buf_type);
L
Linus Torvalds 已提交
1897 1898 1899 1900 1901 1902 1903 1904 1905
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1906
			cifs_stats_bytes_read(pTcon, total_read);
L
Linus Torvalds 已提交
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
			*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
1919
	rc = cifs_revalidate_file(file);
L
Linus Torvalds 已提交
1920
	if (rc) {
1921
		cFYI(1, "Validation prior to mmap failed, error=%d", rc);
L
Linus Torvalds 已提交
1922 1923 1924 1925 1926 1927 1928 1929 1930
		FreeXid(xid);
		return rc;
	}
	rc = generic_file_mmap(file, vma);
	FreeXid(xid);
	return rc;
}


S
Steve French 已提交
1931
static void cifs_copy_cache_pages(struct address_space *mapping,
N
Nick Piggin 已提交
1932
	struct list_head *pages, int bytes_read, char *data)
L
Linus Torvalds 已提交
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943
{
	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
1944
		if (add_to_page_cache_lru(page, mapping, page->index,
L
Linus Torvalds 已提交
1945 1946
				      GFP_KERNEL)) {
			page_cache_release(page);
1947
			cFYI(1, "Add page cache failed");
1948 1949
			data += PAGE_CACHE_SIZE;
			bytes_read -= PAGE_CACHE_SIZE;
L
Linus Torvalds 已提交
1950 1951 1952
			continue;
		}

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 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983
			       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;
	}
	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
1984
	unsigned int bytes_read = 0;
S
Steve French 已提交
1985
	unsigned int read_size, i;
L
Linus Torvalds 已提交
1986 1987 1988
	char *smb_read_data = NULL;
	struct smb_com_read_rsp *pSMBr;
	struct cifsFileInfo *open_file;
1989
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1990 1991 1992

	xid = GetXid();
	if (file->private_data == NULL) {
1993
		rc = -EBADF;
L
Linus Torvalds 已提交
1994
		FreeXid(xid);
1995
		return rc;
L
Linus Torvalds 已提交
1996 1997
	}
	open_file = (struct cifsFileInfo *)file->private_data;
1998
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1999
	pTcon = cifs_sb->tcon;
2000

2001
	cFYI(DBG2, "rpages: num pages %d", num_pages);
L
Linus Torvalds 已提交
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
	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
2015
		expected_index =
L
Linus Torvalds 已提交
2016
			list_entry(page_list->prev, struct page, lru)->index;
S
Steve French 已提交
2017
		list_for_each_entry_reverse(tmp_page, page_list, lru) {
L
Linus Torvalds 已提交
2018 2019 2020 2021
			if (tmp_page->index == expected_index) {
				contig_pages++;
				expected_index++;
			} else
S
Steve French 已提交
2022
				break;
L
Linus Torvalds 已提交
2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033
		}
		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);
2034 2035
		cFYI(DBG2, "rpages: read size 0x%x  contiguous pages %d",
				read_size, contig_pages);
L
Linus Torvalds 已提交
2036 2037
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
2038
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
2039
			    (!open_file->closePend)) {
2040
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
2041 2042 2043 2044
				if (rc != 0)
					break;
			}

2045
			rc = CIFSSMBRead(xid, pTcon,
2046 2047 2048 2049
					 open_file->netfid,
					 read_size, offset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
2050
			/* BB more RC checks ? */
S
Steve French 已提交
2051
			if (rc == -EAGAIN) {
L
Linus Torvalds 已提交
2052
				if (smb_read_data) {
S
Steve French 已提交
2053
					if (buf_type == CIFS_SMALL_BUFFER)
2054
						cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2055
					else if (buf_type == CIFS_LARGE_BUFFER)
2056
						cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2057 2058 2059 2060 2061
					smb_read_data = NULL;
				}
			}
		}
		if ((rc < 0) || (smb_read_data == NULL)) {
2062
			cFYI(1, "Read error in readpages: %d", rc);
L
Linus Torvalds 已提交
2063 2064
			break;
		} else if (bytes_read > 0) {
2065
			task_io_account_read(bytes_read);
L
Linus Torvalds 已提交
2066 2067 2068
			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
2069
				le16_to_cpu(pSMBr->DataOffset));
L
Linus Torvalds 已提交
2070 2071

			i +=  bytes_read >> PAGE_CACHE_SHIFT;
2072
			cifs_stats_bytes_read(pTcon, bytes_read);
S
Steve French 已提交
2073
			if ((bytes_read & PAGE_CACHE_MASK) != bytes_read) {
L
Linus Torvalds 已提交
2074 2075
				i++; /* account for partial page */

S
Steve French 已提交
2076
				/* server copy of file can have smaller size
L
Linus Torvalds 已提交
2077
				   than client */
S
Steve French 已提交
2078 2079
				/* 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
2080 2081
				   we will hit it on next read */

O
OGAWA Hirofumi 已提交
2082
				/* break; */
L
Linus Torvalds 已提交
2083 2084
			}
		} else {
2085
			cFYI(1, "No bytes read (%d) at offset %lld . "
2086
				"Cleaning remaining pages from readahead list",
2087
				bytes_read, offset);
S
Steve French 已提交
2088
			/* BB turn off caching and do new lookup on
L
Linus Torvalds 已提交
2089 2090 2091 2092
			   file size at server? */
			break;
		}
		if (smb_read_data) {
S
Steve French 已提交
2093
			if (buf_type == CIFS_SMALL_BUFFER)
2094
				cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2095
			else if (buf_type == CIFS_LARGE_BUFFER)
2096
				cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2097 2098 2099 2100 2101 2102 2103
			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
2104
		if (buf_type == CIFS_SMALL_BUFFER)
2105
			cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
2106
		else if (buf_type == CIFS_LARGE_BUFFER)
2107
			cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
2108
		smb_read_data = NULL;
S
Steve French 已提交
2109
	}
L
Linus Torvalds 已提交
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123

	FreeXid(xid);
	return rc;
}

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

	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
2124

L
Linus Torvalds 已提交
2125
	rc = cifs_read(file, read_data, PAGE_CACHE_SIZE, poffset);
S
Steve French 已提交
2126

L
Linus Torvalds 已提交
2127 2128 2129
	if (rc < 0)
		goto io_error;
	else
2130
		cFYI(1, "Bytes read %d", rc);
S
Steve French 已提交
2131

2132 2133
	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
2134

L
Linus Torvalds 已提交
2135 2136 2137 2138 2139 2140
	if (PAGE_CACHE_SIZE > rc)
		memset(read_data + rc, 0, PAGE_CACHE_SIZE - rc);

	flush_dcache_page(page);
	SetPageUptodate(page);
	rc = 0;
S
Steve French 已提交
2141

L
Linus Torvalds 已提交
2142
io_error:
S
Steve French 已提交
2143
	kunmap(page);
L
Linus Torvalds 已提交
2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156
	page_cache_release(page);
	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) {
2157
		rc = -EBADF;
L
Linus Torvalds 已提交
2158
		FreeXid(xid);
2159
		return rc;
L
Linus Torvalds 已提交
2160 2161
	}

2162 2163
	cFYI(1, "readpage %p at offset %d 0x%x\n",
		 page, (int)offset, (int)offset);
L
Linus Torvalds 已提交
2164 2165 2166 2167 2168 2169 2170 2171 2172

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

	unlock_page(page);

	FreeXid(xid);
	return rc;
}

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
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
2192 2193 2194
/* 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
2195
   refreshing the inode only on increases in the file size
L
Linus Torvalds 已提交
2196 2197
   but this is tricky to do without racing with writebehind
   page caching in the current Linux kernel design */
2198
bool is_size_safe_to_change(struct cifsInodeInfo *cifsInode, __u64 end_of_file)
L
Linus Torvalds 已提交
2199
{
2200
	if (!cifsInode)
2201
		return true;
2202

2203 2204
	if (is_inode_writable(cifsInode)) {
		/* This inode is open for write at least once */
2205 2206 2207
		struct cifs_sb_info *cifs_sb;

		cifs_sb = CIFS_SB(cifsInode->vfs_inode.i_sb);
S
Steve French 已提交
2208
		if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DIRECT_IO) {
S
Steve French 已提交
2209
			/* since no page cache to corrupt on directio
2210
			we can change size safely */
2211
			return true;
2212 2213
		}

S
Steve French 已提交
2214
		if (i_size_read(&cifsInode->vfs_inode) < end_of_file)
2215
			return true;
2216

2217
		return false;
2218
	} else
2219
		return true;
L
Linus Torvalds 已提交
2220 2221
}

N
Nick Piggin 已提交
2222 2223 2224
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
2225
{
N
Nick Piggin 已提交
2226 2227
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	loff_t offset = pos & (PAGE_CACHE_SIZE - 1);
2228 2229 2230 2231
	loff_t page_start = pos & PAGE_MASK;
	loff_t i_size;
	struct page *page;
	int rc = 0;
N
Nick Piggin 已提交
2232

2233
	cFYI(1, "write_begin from %lld len %d", (long long)pos, len);
N
Nick Piggin 已提交
2234

2235
	page = grab_cache_page_write_begin(mapping, index, flags);
2236 2237 2238 2239
	if (!page) {
		rc = -ENOMEM;
		goto out;
	}
2240

2241 2242
	if (PageUptodate(page))
		goto out;
2243

2244 2245 2246 2247 2248 2249 2250
	/*
	 * 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;
2251

2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274
	/*
	 * 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
2275

2276 2277 2278 2279 2280 2281 2282
	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);
2283 2284 2285 2286
	} 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
2287
		   this will be written out by write_end so is fine */
L
Linus Torvalds 已提交
2288
	}
2289 2290 2291
out:
	*pagep = page;
	return rc;
L
Linus Torvalds 已提交
2292 2293
}

2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
static void
cifs_oplock_break(struct slow_work *work)
{
	struct cifsFileInfo *cfile = container_of(work, struct cifsFileInfo,
						  oplock_break);
	struct inode *inode = cfile->pInode;
	struct cifsInodeInfo *cinode = CIFS_I(inode);
	struct cifs_sb_info *cifs_sb = CIFS_SB(cfile->mnt->mnt_sb);
	int rc, waitrc = 0;

	if (inode && S_ISREG(inode->i_mode)) {
#ifdef CONFIG_CIFS_EXPERIMENTAL
		if (cinode->clientCanCacheAll == 0)
2307
			break_lease(inode, O_RDONLY);
2308
		else if (cinode->clientCanCacheRead == 0)
2309
			break_lease(inode, O_WRONLY);
2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
#endif
		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;
2320
		cFYI(1, "Oplock flush inode %p rc %d", inode, rc);
2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
	}

	/*
	 * 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) {
		rc = CIFSSMBLock(0, cifs_sb->tcon, cfile->netfid, 0, 0, 0, 0,
				 LOCKING_ANDX_OPLOCK_RELEASE, false);
2332
		cFYI(1, "Oplock release rc = %d", rc);
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
	}
}

static int
cifs_oplock_break_get(struct slow_work *work)
{
	struct cifsFileInfo *cfile = container_of(work, struct cifsFileInfo,
						  oplock_break);
	mntget(cfile->mnt);
	cifsFileInfo_get(cfile);
	return 0;
}

static void
cifs_oplock_break_put(struct slow_work *work)
{
	struct cifsFileInfo *cfile = container_of(work, struct cifsFileInfo,
						  oplock_break);
	mntput(cfile->mnt);
	cifsFileInfo_put(cfile);
}

const struct slow_work_ops cifs_oplock_break_ops = {
	.get_ref	= cifs_oplock_break_get,
	.put_ref	= cifs_oplock_break_put,
	.execute	= cifs_oplock_break,
};

2361
const struct address_space_operations cifs_addr_ops = {
L
Linus Torvalds 已提交
2362 2363 2364
	.readpage = cifs_readpage,
	.readpages = cifs_readpages,
	.writepage = cifs_writepage,
2365
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2366 2367
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
L
Linus Torvalds 已提交
2368 2369 2370 2371
	.set_page_dirty = __set_page_dirty_nobuffers,
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};
D
Dave Kleikamp 已提交
2372 2373 2374 2375 2376 2377

/*
 * 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.
 */
2378
const struct address_space_operations cifs_addr_ops_smallbuf = {
D
Dave Kleikamp 已提交
2379 2380 2381
	.readpage = cifs_readpage,
	.writepage = cifs_writepage,
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2382 2383
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
D
Dave Kleikamp 已提交
2384 2385 2386 2387
	.set_page_dirty = __set_page_dirty_nobuffers,
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};