file.c 58.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4
/*
 *   fs/cifs/file.c
 *
 *   vfs operations that deal with files
S
Steve French 已提交
5 6
 *
 *   Copyright (C) International Business Machines  Corp., 2002,2007
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>
L
Linus Torvalds 已提交
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
#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 struct cifsFileInfo *cifs_init_private(
	struct cifsFileInfo *private_data, struct inode *inode,
	struct file *file, __u16 netfid)
{
	memset(private_data, 0, sizeof(struct cifsFileInfo));
	private_data->netfid = netfid;
S
Steve French 已提交
48
	private_data->pid = current->tgid;
L
Linus Torvalds 已提交
49
	init_MUTEX(&private_data->fh_sem);
50
	mutex_init(&private_data->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
51
	INIT_LIST_HEAD(&private_data->llist);
L
Linus Torvalds 已提交
52 53
	private_data->pfile = file; /* needed for writepage */
	private_data->pInode = inode;
54 55
	private_data->invalidHandle = false;
	private_data->closePend = false;
56 57 58 59
	/* we have to track num writers to the inode, since writepages
	does not tell us which handle the write is for so there can
	be a close (overlapping with write) of the filehandle that
	cifs_writepages chose to use */
S
Steve French 已提交
60
	atomic_set(&private_data->wrtPending, 0);
L
Linus Torvalds 已提交
61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77

	return private_data;
}

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);
	}

78 79 80 81 82
	return (READ_CONTROL | FILE_WRITE_ATTRIBUTES | FILE_READ_ATTRIBUTES |
		FILE_WRITE_EA | FILE_APPEND_DATA | FILE_WRITE_DATA |
		FILE_READ_DATA);


L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90 91 92
}

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;
93 94
	else if ((flags & O_TRUNC) == O_TRUNC)
		return FILE_OVERWRITE;
L
Linus Torvalds 已提交
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109
	else
		return FILE_OPEN;
}

/* 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;

	/* want handles we can use to read with first
	   in the list so we do not have to walk the
N
Nick Piggin 已提交
110
	   list to search for one in write_begin */
L
Linus Torvalds 已提交
111
	if ((file->f_flags & O_ACCMODE) == O_WRONLY) {
S
Steve French 已提交
112
		list_add_tail(&pCifsFile->flist,
L
Linus Torvalds 已提交
113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
			      &pCifsInode->openFileList);
	} else {
		list_add(&pCifsFile->flist,
			 &pCifsInode->openFileList);
	}
	write_unlock(&GlobalSMBSeslock);
	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 */
	temp = cifs_NTtimeToUnix(le64_to_cpu(buf->LastWriteTime));
129 130
	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
131 132 133
			    (loff_t)le64_to_cpu(buf->EndOfFile))) {
		cFYI(1, ("inode unchanged on server"));
	} else {
134
		if (file->f_path.dentry->d_inode->i_mapping) {
L
Linus Torvalds 已提交
135 136
		/* BB no need to lock inode until after invalidate
		   since namei code should already have it locked? */
137 138 139
			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
140 141 142
		}
		cFYI(1, ("invalidating remote inode since open detected it "
			 "changed"));
143
		invalidate_remote_inode(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
144 145 146
	}

client_can_cache:
147
	if (pTcon->unix_ext)
148
		rc = cifs_get_inode_info_unix(&file->f_path.dentry->d_inode,
L
Linus Torvalds 已提交
149 150
			full_path, inode->i_sb, xid);
	else
151
		rc = cifs_get_inode_info(&file->f_path.dentry->d_inode,
152
			full_path, buf, inode->i_sb, xid, NULL);
L
Linus Torvalds 已提交
153 154

	if ((*oplock & 0xF) == OPLOCK_EXCLUSIVE) {
155 156
		pCifsInode->clientCanCacheAll = true;
		pCifsInode->clientCanCacheRead = true;
L
Linus Torvalds 已提交
157
		cFYI(1, ("Exclusive Oplock granted on inode %p",
158
			 file->f_path.dentry->d_inode));
L
Linus Torvalds 已提交
159
	} else if ((*oplock & 0xF) == OPLOCK_READ)
160
		pCifsInode->clientCanCacheRead = true;
L
Linus Torvalds 已提交
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

	return rc;
}

int cifs_open(struct inode *inode, struct file *file)
{
	int rc = -EACCES;
	int xid, oplock;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	struct cifsFileInfo *pCifsFile;
	struct cifsInodeInfo *pCifsInode;
	struct list_head *tmp;
	char *full_path = NULL;
	int desiredAccess;
	int disposition;
	__u16 netfid;
	FILE_ALL_INFO *buf = NULL;

	xid = GetXid();

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

	if (file->f_flags & O_CREAT) {
		/* search inode for this file and fill in file->private_data */
187
		pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
188 189 190 191 192 193 194 195 196 197
		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;
198

L
Linus Torvalds 已提交
199 200 201 202 203 204 205 206 207 208 209 210
				file->private_data = pCifsFile;
				break;
			}
		}
		read_unlock(&GlobalSMBSeslock);
		if (file->private_data != NULL) {
			rc = 0;
			FreeXid(xid);
			return rc;
		} else {
			if (file->f_flags & O_EXCL)
				cERROR(1, ("could not find file instance for "
211
					   "new file %p", file));
L
Linus Torvalds 已提交
212 213 214
		}
	}

215
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
216 217 218 219 220
	if (full_path == NULL) {
		FreeXid(xid);
		return -ENOMEM;
	}

221
	cFYI(1, ("inode = 0x%p file flags are 0x%x for %s",
L
Linus Torvalds 已提交
222 223 224 225 226
		 inode, file->f_flags, full_path));
	desiredAccess = cifs_convert_flags(file->f_flags);

/*********************************************************************
 *  open flag mapping table:
S
Steve French 已提交
227
 *
L
Linus Torvalds 已提交
228
 *	POSIX Flag            CIFS Disposition
S
Steve French 已提交
229
 *	----------            ----------------
L
Linus Torvalds 已提交
230 231 232 233 234 235 236
 *	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
237
 *	FILE_SUPERSEDE (ie create whether or not file exists although
L
Linus Torvalds 已提交
238 239 240 241
 *	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
242
 *?  O_SYNC is a reasonable match to CIFS writethrough flag
L
Linus Torvalds 已提交
243 244 245 246 247 248 249 250 251 252 253
 *?  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);

	if (oplockEnabled)
		oplock = REQ_OPLOCK;
	else
254
		oplock = 0;
L
Linus Torvalds 已提交
255 256 257 258 259 260 261

	/* 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
262 263
	/* 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
264 265 266 267 268 269
	   able to simply do a filemap_fdatawrite/filemap_fdatawait first */
	buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
	if (!buf) {
		rc = -ENOMEM;
		goto out;
	}
270 271

	if (cifs_sb->tcon->ses->capabilities & CAP_NT_SMBS)
S
Steve French 已提交
272
		rc = CIFSSMBOpen(xid, pTcon, full_path, disposition,
273
			 desiredAccess, CREATE_NOT_DIR, &netfid, &oplock, buf,
274 275
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags
				 & CIFS_MOUNT_MAP_SPECIAL_CHR);
276 277 278
	else
		rc = -EIO; /* no NT SMB support fall into legacy open below */

279 280 281 282 283 284 285
	if (rc == -EIO) {
		/* Old server, try legacy style OpenX */
		rc = SMBLegacyOpen(xid, pTcon, full_path, disposition,
			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
286
	if (rc) {
287
		cFYI(1, ("cifs_open returned 0x%x", rc));
L
Linus Torvalds 已提交
288 289 290 291 292 293 294 295 296 297 298 299
		goto out;
	}
	file->private_data =
		kmalloc(sizeof(struct cifsFileInfo), GFP_KERNEL);
	if (file->private_data == NULL) {
		rc = -ENOMEM;
		goto out;
	}
	pCifsFile = cifs_init_private(file->private_data, inode, file, netfid);
	write_lock(&GlobalSMBSeslock);
	list_add(&pCifsFile->tlist, &pTcon->openFileList);

300
	pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
301 302 303 304 305 306 307 308
	if (pCifsInode) {
		rc = cifs_open_inode_helper(inode, file, pCifsInode,
					    pCifsFile, pTcon,
					    &oplock, buf, full_path, xid);
	} else {
		write_unlock(&GlobalSMBSeslock);
	}

S
Steve French 已提交
309
	if (oplock & CIFS_CREATE_ACTION) {
L
Linus Torvalds 已提交
310 311
		/* time to set mode which we can not set earlier due to
		   problems creating new read-only files */
312
		if (pTcon->unix_ext) {
313 314 315 316 317 318 319 320 321 322
			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,
			};
			CIFSSMBUnixSetInfo(xid, pTcon, full_path, &args,
323
					    cifs_sb->local_nls,
S
Steve French 已提交
324
					    cifs_sb->mnt_cifs_flags &
325
						CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
326 327 328 329 330 331 332 333 334 335
		}
	}

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

A
Adrian Bunk 已提交
336
/* Try to reacquire byte range locks that were released when session */
L
Linus Torvalds 已提交
337 338 339 340 341 342 343 344 345 346
/* 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;
}

347
static int cifs_reopen_file(struct file *file, bool can_flush)
L
Linus Torvalds 已提交
348 349 350 351 352 353 354
{
	int rc = -EACCES;
	int xid, oplock;
	struct cifs_sb_info *cifs_sb;
	struct cifsTconInfo *pTcon;
	struct cifsFileInfo *pCifsFile;
	struct cifsInodeInfo *pCifsInode;
S
Steve French 已提交
355
	struct inode *inode;
L
Linus Torvalds 已提交
356 357 358 359 360
	char *full_path = NULL;
	int desiredAccess;
	int disposition = FILE_OPEN;
	__u16 netfid;

S
Steve French 已提交
361
	if (file->private_data)
L
Linus Torvalds 已提交
362
		pCifsFile = (struct cifsFileInfo *)file->private_data;
S
Steve French 已提交
363
	else
L
Linus Torvalds 已提交
364 365 366 367
		return -EBADF;

	xid = GetXid();
	down(&pCifsFile->fh_sem);
368
	if (!pCifsFile->invalidHandle) {
L
Linus Torvalds 已提交
369 370 371 372 373
		up(&pCifsFile->fh_sem);
		FreeXid(xid);
		return 0;
	}

374
	if (file->f_path.dentry == NULL) {
375 376 377 378 379 380 381
		cERROR(1, ("no valid name if dentry freed"));
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
	}

	inode = file->f_path.dentry->d_inode;
S
Steve French 已提交
382
	if (inode == NULL) {
383 384 385 386
		cERROR(1, ("inode not valid"));
		dump_stack();
		rc = -EBADF;
		goto reopen_error_exit;
L
Linus Torvalds 已提交
387
	}
388

L
Linus Torvalds 已提交
389 390
	cifs_sb = CIFS_SB(inode->i_sb);
	pTcon = cifs_sb->tcon;
391

L
Linus Torvalds 已提交
392 393 394 395
/* 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 */
396
	full_path = build_path_from_dentry(file->f_path.dentry);
L
Linus Torvalds 已提交
397
	if (full_path == NULL) {
398 399
		rc = -ENOMEM;
reopen_error_exit:
L
Linus Torvalds 已提交
400 401
		up(&pCifsFile->fh_sem);
		FreeXid(xid);
402
		return rc;
L
Linus Torvalds 已提交
403 404
	}

405
	cFYI(1, ("inode = 0x%p file flags 0x%x for %s",
S
Steve French 已提交
406
		 inode, file->f_flags, full_path));
L
Linus Torvalds 已提交
407 408 409 410 411
	desiredAccess = cifs_convert_flags(file->f_flags);

	if (oplockEnabled)
		oplock = REQ_OPLOCK;
	else
412
		oplock = 0;
L
Linus Torvalds 已提交
413 414

	/* Can not refresh inode by passing in file_info buf to be returned
S
Steve French 已提交
415 416
	   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
417 418 419 420 421
	   and server version of file size can be stale. If we knew for sure
	   that inode was not dirty locally we could do this */

	rc = CIFSSMBOpen(xid, pTcon, full_path, disposition, desiredAccess,
			 CREATE_NOT_DIR, &netfid, &oplock, NULL,
S
Steve French 已提交
422
			 cifs_sb->local_nls, cifs_sb->mnt_cifs_flags &
423
				CIFS_MOUNT_MAP_SPECIAL_CHR);
L
Linus Torvalds 已提交
424 425
	if (rc) {
		up(&pCifsFile->fh_sem);
426 427
		cFYI(1, ("cifs_open returned 0x%x", rc));
		cFYI(1, ("oplock: %d", oplock));
L
Linus Torvalds 已提交
428 429
	} else {
		pCifsFile->netfid = netfid;
430
		pCifsFile->invalidHandle = false;
L
Linus Torvalds 已提交
431 432 433 434
		up(&pCifsFile->fh_sem);
		pCifsInode = CIFS_I(inode);
		if (pCifsInode) {
			if (can_flush) {
435 436 437
				rc = filemap_write_and_wait(inode->i_mapping);
				if (rc != 0)
					CIFS_I(inode)->write_behind_rc = rc;
L
Linus Torvalds 已提交
438 439
			/* temporarily disable caching while we
			   go to server to get inode info */
440 441
				pCifsInode->clientCanCacheAll = false;
				pCifsInode->clientCanCacheRead = false;
442
				if (pTcon->unix_ext)
L
Linus Torvalds 已提交
443 444 445 446 447
					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,
448
						xid, NULL);
L
Linus Torvalds 已提交
449 450 451 452 453 454 455
			} /* 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) {
456 457
				pCifsInode->clientCanCacheAll = true;
				pCifsInode->clientCanCacheRead = true;
L
Linus Torvalds 已提交
458
				cFYI(1, ("Exclusive Oplock granted on inode %p",
459
					 file->f_path.dentry->d_inode));
L
Linus Torvalds 已提交
460
			} else if ((oplock & 0xF) == OPLOCK_READ) {
461 462
				pCifsInode->clientCanCacheRead = true;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
463
			} else {
464 465
				pCifsInode->clientCanCacheRead = false;
				pCifsInode->clientCanCacheAll = false;
L
Linus Torvalds 已提交
466 467 468 469 470 471 472 473 474 475 476 477 478
			}
			cifs_relock_file(pCifsFile);
		}
	}

	kfree(full_path);
	FreeXid(xid);
	return rc;
}

int cifs_close(struct inode *inode, struct file *file)
{
	int rc = 0;
479
	int xid, timeout;
L
Linus Torvalds 已提交
480 481 482 483 484 485 486 487 488 489
	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
490
		struct cifsLockInfo *li, *tmp;
491
		write_lock(&GlobalSMBSeslock);
492
		pSMBFile->closePend = true;
L
Linus Torvalds 已提交
493 494 495
		if (pTcon) {
			/* no sense reconnecting to close a file that is
			   already closed */
S
Steve French 已提交
496
			if (!pTcon->need_reconnect) {
497
				write_unlock(&GlobalSMBSeslock);
498
				timeout = 2;
S
Steve French 已提交
499
				while ((atomic_read(&pSMBFile->wrtPending) != 0)
500
					&& (timeout <= 2048)) {
501 502 503 504 505
					/* 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
506
					but this should give enough time to
507
					clear the socket */
508 509
					cFYI(DBG2,
						("close delay, write pending"));
510 511
					msleep(timeout);
					timeout *= 4;
512
				}
S
Steve French 已提交
513
				if (atomic_read(&pSMBFile->wrtPending))
514 515 516 517
					cERROR(1, ("close with pending write"));
				if (!pTcon->need_reconnect &&
				    !pSMBFile->invalidHandle)
					rc = CIFSSMBClose(xid, pTcon,
L
Linus Torvalds 已提交
518
						  pSMBFile->netfid);
519 520 521 522
			} else
				write_unlock(&GlobalSMBSeslock);
		} else
			write_unlock(&GlobalSMBSeslock);
J
[CIFS]  
Jeremy Allison 已提交
523 524 525

		/* Delete any outstanding lock records.
		   We'll lose them when the file is closed anyway. */
526
		mutex_lock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
527 528 529 530
		list_for_each_entry_safe(li, tmp, &pSMBFile->llist, llist) {
			list_del(&li->llist);
			kfree(li);
		}
531
		mutex_unlock(&pSMBFile->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
532

533
		write_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
534 535
		list_del(&pSMBFile->flist);
		list_del(&pSMBFile->tlist);
536
		write_unlock(&GlobalSMBSeslock);
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553
		timeout = 10;
		/* We waited above to give the SMBWrite a chance to issue
		   on the wire (so we do not get SMBWrite returning EBADF
		   if writepages is racing with close.  Note that writepages
		   does not specify a file handle, so it is possible for a file
		   to be opened twice, and the application close the "wrong"
		   file handle - in these cases we delay long enough to allow
		   the SMBWrite to get on the wire before the SMB Close.
		   We allow total wait here over 45 seconds, more than
		   oplock break time, and more than enough to allow any write
		   to complete on the server, or to time out on the client */
		while ((atomic_read(&pSMBFile->wrtPending) != 0)
				&& (timeout <= 50000)) {
			cERROR(1, ("writes pending, delay free of handle"));
			msleep(timeout);
			timeout *= 8;
		}
L
Linus Torvalds 已提交
554 555 556 557 558
		kfree(file->private_data);
		file->private_data = NULL;
	} else
		rc = -EBADF;

559
	read_lock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
560 561 562 563
	if (list_empty(&(CIFS_I(inode)->openFileList))) {
		cFYI(1, ("closing last open instance for inode %p", inode));
		/* 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 */
564 565
		CIFS_I(inode)->clientCanCacheRead = false;
		CIFS_I(inode)->clientCanCacheAll  = false;
L
Linus Torvalds 已提交
566
	}
567
	read_unlock(&GlobalSMBSeslock);
S
Steve French 已提交
568
	if ((rc == 0) && CIFS_I(inode)->write_behind_rc)
L
Linus Torvalds 已提交
569 570 571 572 573 574 575 576 577 578 579 580 581
		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;

582
	cFYI(1, ("Closedir inode = 0x%p", inode));
L
Linus Torvalds 已提交
583 584 585 586 587

	xid = GetXid();

	if (pCFileStruct) {
		struct cifsTconInfo *pTcon;
S
Steve French 已提交
588 589
		struct cifs_sb_info *cifs_sb =
			CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
590 591 592 593

		pTcon = cifs_sb->tcon;

		cFYI(1, ("Freeing private data in close dir"));
594
		write_lock(&GlobalSMBSeslock);
595 596 597
		if (!pCFileStruct->srch_inf.endOfSearch &&
		    !pCFileStruct->invalidHandle) {
			pCFileStruct->invalidHandle = true;
598
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
599 600 601 602 603
			rc = CIFSFindClose(xid, pTcon, pCFileStruct->netfid);
			cFYI(1, ("Closing uncompleted readdir with rc %d",
				 rc));
			/* not much we can do if it fails anyway, ignore rc */
			rc = 0;
604 605
		} else
			write_unlock(&GlobalSMBSeslock);
L
Linus Torvalds 已提交
606 607
		ptmp = pCFileStruct->srch_inf.ntwrk_buf_start;
		if (ptmp) {
608
			cFYI(1, ("closedir free smb buf in srch struct"));
L
Linus Torvalds 已提交
609
			pCFileStruct->srch_inf.ntwrk_buf_start = NULL;
S
Steve French 已提交
610
			if (pCFileStruct->srch_inf.smallBuf)
611 612 613
				cifs_small_buf_release(ptmp);
			else
				cifs_buf_release(ptmp);
L
Linus Torvalds 已提交
614 615 616 617 618 619 620 621 622
		}
		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
623 624 625
static int store_file_lock(struct cifsFileInfo *fid, __u64 len,
				__u64 offset, __u8 lockType)
{
S
Steve French 已提交
626 627
	struct cifsLockInfo *li =
		kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
J
[CIFS]  
Jeremy Allison 已提交
628 629 630 631 632
	if (li == NULL)
		return -ENOMEM;
	li->offset = offset;
	li->length = len;
	li->type = lockType;
633
	mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
634
	list_add(&li->llist, &fid->llist);
635
	mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
636 637 638
	return 0;
}

L
Linus Torvalds 已提交
639 640 641 642 643 644
int cifs_lock(struct file *file, int cmd, struct file_lock *pfLock)
{
	int rc, xid;
	__u32 numLock = 0;
	__u32 numUnlock = 0;
	__u64 length;
645
	bool wait_flag = false;
L
Linus Torvalds 已提交
646
	struct cifs_sb_info *cifs_sb;
647
	struct cifsTconInfo *tcon;
648 649
	__u16 netfid;
	__u8 lockType = LOCKING_ANDX_LARGE_FILES;
650
	bool posix_locking = 0;
L
Linus Torvalds 已提交
651 652 653 654 655 656 657

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

	cFYI(1, ("Lock parm: 0x%x flockflags: "
		 "0x%x flocktype: 0x%x start: %lld end: %lld",
S
Steve French 已提交
658 659
		cmd, pfLock->fl_flags, pfLock->fl_type, pfLock->fl_start,
		pfLock->fl_end));
L
Linus Torvalds 已提交
660 661

	if (pfLock->fl_flags & FL_POSIX)
662
		cFYI(1, ("Posix"));
L
Linus Torvalds 已提交
663
	if (pfLock->fl_flags & FL_FLOCK)
664
		cFYI(1, ("Flock"));
L
Linus Torvalds 已提交
665
	if (pfLock->fl_flags & FL_SLEEP) {
666
		cFYI(1, ("Blocking lock"));
667
		wait_flag = true;
L
Linus Torvalds 已提交
668 669 670
	}
	if (pfLock->fl_flags & FL_ACCESS)
		cFYI(1, ("Process suspended by mandatory locking - "
671
			 "not implemented yet"));
L
Linus Torvalds 已提交
672 673
	if (pfLock->fl_flags & FL_LEASE)
		cFYI(1, ("Lease on file - not implemented yet"));
S
Steve French 已提交
674
	if (pfLock->fl_flags &
L
Linus Torvalds 已提交
675 676 677 678 679 680 681
	    (~(FL_POSIX | FL_FLOCK | FL_SLEEP | FL_ACCESS | FL_LEASE)))
		cFYI(1, ("Unknown lock flags 0x%x", pfLock->fl_flags));

	if (pfLock->fl_type == F_WRLCK) {
		cFYI(1, ("F_WRLCK "));
		numLock = 1;
	} else if (pfLock->fl_type == F_UNLCK) {
682
		cFYI(1, ("F_UNLCK"));
L
Linus Torvalds 已提交
683
		numUnlock = 1;
684 685
		/* Check if unlock includes more than
		one lock range */
L
Linus Torvalds 已提交
686
	} else if (pfLock->fl_type == F_RDLCK) {
687
		cFYI(1, ("F_RDLCK"));
L
Linus Torvalds 已提交
688 689 690
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else if (pfLock->fl_type == F_EXLCK) {
691
		cFYI(1, ("F_EXLCK"));
L
Linus Torvalds 已提交
692 693
		numLock = 1;
	} else if (pfLock->fl_type == F_SHLCK) {
694
		cFYI(1, ("F_SHLCK"));
L
Linus Torvalds 已提交
695 696 697
		lockType |= LOCKING_ANDX_SHARED_LOCK;
		numLock = 1;
	} else
698
		cFYI(1, ("Unknown type of lock"));
L
Linus Torvalds 已提交
699

700
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
701
	tcon = cifs_sb->tcon;
L
Linus Torvalds 已提交
702 703 704 705 706

	if (file->private_data == NULL) {
		FreeXid(xid);
		return -EBADF;
	}
707 708
	netfid = ((struct cifsFileInfo *)file->private_data)->netfid;

709 710 711 712
	if ((tcon->ses->capabilities & CAP_UNIX) &&
	    (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
	    (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL == 0))
		posix_locking = 1;
713 714 715
	/* 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
716
	if (IS_GETLK(cmd)) {
S
Steve French 已提交
717
		if (posix_locking) {
718
			int posix_lock_type;
S
Steve French 已提交
719
			if (lockType & LOCKING_ANDX_SHARED_LOCK)
720 721 722
				posix_lock_type = CIFS_RDLCK;
			else
				posix_lock_type = CIFS_WRLCK;
723
			rc = CIFSSMBPosixLock(xid, tcon, netfid, 1 /* get */,
724
					length,	pfLock,
725 726 727 728 729 730
					posix_lock_type, wait_flag);
			FreeXid(xid);
			return rc;
		}

		/* BB we could chain these into one lock request BB */
731
		rc = CIFSSMBLock(xid, tcon, netfid, length, pfLock->fl_start,
732
				 0, 1, lockType, 0 /* wait flag */ );
L
Linus Torvalds 已提交
733
		if (rc == 0) {
734
			rc = CIFSSMBLock(xid, tcon, netfid, length,
L
Linus Torvalds 已提交
735 736 737 738 739 740
					 pfLock->fl_start, 1 /* numUnlock */ ,
					 0 /* numLock */ , lockType,
					 0 /* wait flag */ );
			pfLock->fl_type = F_UNLCK;
			if (rc != 0)
				cERROR(1, ("Error unlocking previously locked "
741
					   "range %d during test of lock", rc));
L
Linus Torvalds 已提交
742 743 744 745 746 747 748 749 750 751 752
			rc = 0;

		} else {
			/* if rc == ERR_SHARING_VIOLATION ? */
			rc = 0;	/* do not change lock type to unlock
				   since range in use */
		}

		FreeXid(xid);
		return rc;
	}
J
[CIFS]  
Jeremy Allison 已提交
753 754 755 756 757 758 759 760 761

	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) {
762
		int posix_lock_type;
S
Steve French 已提交
763
		if (lockType & LOCKING_ANDX_SHARED_LOCK)
764 765 766
			posix_lock_type = CIFS_RDLCK;
		else
			posix_lock_type = CIFS_WRLCK;
767

S
Steve French 已提交
768
		if (numUnlock == 1)
769
			posix_lock_type = CIFS_UNLCK;
J
[CIFS]  
Jeremy Allison 已提交
770

771
		rc = CIFSSMBPosixLock(xid, tcon, netfid, 0 /* set */,
772
				      length, pfLock,
773
				      posix_lock_type, wait_flag);
J
[CIFS]  
Jeremy Allison 已提交
774
	} else {
S
Steve French 已提交
775 776
		struct cifsFileInfo *fid =
			(struct cifsFileInfo *)file->private_data;
J
[CIFS]  
Jeremy Allison 已提交
777 778

		if (numLock) {
779
			rc = CIFSSMBLock(xid, tcon, netfid, length,
S
Steve French 已提交
780
					pfLock->fl_start,
J
[CIFS]  
Jeremy Allison 已提交
781 782 783 784 785 786 787 788 789 790 791 792 793
					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;

794
			rc = 0;
795
			mutex_lock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
796 797
			list_for_each_entry_safe(li, tmp, &fid->llist, llist) {
				if (pfLock->fl_start <= li->offset &&
S
Steve French 已提交
798
						(pfLock->fl_start + length) >=
799
						(li->offset + li->length)) {
800
					stored_rc = CIFSSMBLock(xid, tcon,
S
Steve French 已提交
801
							netfid,
J
[CIFS]  
Jeremy Allison 已提交
802
							li->length, li->offset,
803
							1, 0, li->type, false);
J
[CIFS]  
Jeremy Allison 已提交
804 805 806 807 808 809 810
					if (stored_rc)
						rc = stored_rc;

					list_del(&li->llist);
					kfree(li);
				}
			}
811
			mutex_unlock(&fid->lock_mutex);
J
[CIFS]  
Jeremy Allison 已提交
812 813 814
		}
	}

815
	if (pfLock->fl_flags & FL_POSIX)
L
Linus Torvalds 已提交
816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831
		posix_lock_file_wait(file, pfLock);
	FreeXid(xid);
	return rc;
}

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;

832
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
833 834 835 836 837

	pTcon = cifs_sb->tcon;

	/* cFYI(1,
	   (" write %d bytes to offset %lld of %s", write_size,
838
	   *poffset, file->f_path.dentry->d_name.name)); */
L
Linus Torvalds 已提交
839 840 841

	if (file->private_data == NULL)
		return -EBADF;
842
	open_file = (struct cifsFileInfo *) file->private_data;
843

844 845 846 847
	rc = generic_write_checks(file, poffset, &write_size, 0);
	if (rc)
		return rc;

L
Linus Torvalds 已提交
848 849
	xid = GetXid();

850
	if (*poffset > file->f_path.dentry->d_inode->i_size)
851
		long_op = CIFS_VLONG_OP; /* writes past EOF take long time */
L
Linus Torvalds 已提交
852
	else
853
		long_op = CIFS_LONG_OP;
L
Linus Torvalds 已提交
854 855 856 857 858 859 860 861 862 863 864 865

	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
866
			}
L
Linus Torvalds 已提交
867 868 869 870 871 872 873 874 875 876 877 878
			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 */
879
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
				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;
			}
		} else
			*poffset += bytes_written;
900
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
901 902 903
				    15 seconds is plenty */
	}

904
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
905 906

	/* since the write may have blocked check these pointers again */
907 908
	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
909 910
/* Do not update local mtime - server will set its actual value on write
 *		inode->i_ctime = inode->i_mtime =
911 912 913 914 915
 * 			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
916
					*poffset);
917
			spin_unlock(&inode->i_lock);
L
Linus Torvalds 已提交
918
		}
S
Steve French 已提交
919
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
920 921 922 923 924 925
	}
	FreeXid(xid);
	return total_written;
}

static ssize_t cifs_write(struct file *file, const char *write_data,
N
Nick Piggin 已提交
926
			  size_t write_size, loff_t *poffset)
L
Linus Torvalds 已提交
927 928 929 930 931 932 933 934 935
{
	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;

936
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
937 938 939

	pTcon = cifs_sb->tcon;

S
Steve French 已提交
940
	cFYI(1, ("write %zd bytes to offset %lld of %s", write_size,
941
	   *poffset, file->f_path.dentry->d_name.name));
L
Linus Torvalds 已提交
942 943 944

	if (file->private_data == NULL)
		return -EBADF;
945
	open_file = (struct cifsFileInfo *)file->private_data;
946

L
Linus Torvalds 已提交
947 948
	xid = GetXid();

949
	if (*poffset > file->f_path.dentry->d_inode->i_size)
950
		long_op = CIFS_VLONG_OP; /* writes past EOF can be slow */
L
Linus Torvalds 已提交
951
	else
952
		long_op = CIFS_LONG_OP;
L
Linus Torvalds 已提交
953 954 955 956 957 958 959 960 961 962

	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
963
			   while we blocked so return what we managed to
L
Linus Torvalds 已提交
964 965
			   write */
				return total_written;
S
Steve French 已提交
966
			}
L
Linus Torvalds 已提交
967 968 969 970 971 972 973 974 975 976
			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
977
				   reopen_file not to flush data to
L
Linus Torvalds 已提交
978
				   server now */
979
				rc = cifs_reopen_file(file, false);
L
Linus Torvalds 已提交
980 981 982
				if (rc != 0)
					break;
			}
S
Steve French 已提交
983 984
			if (experimEnabled || (pTcon->ses->server &&
				((pTcon->ses->server->secMode &
S
Steve French 已提交
985
				(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED))
S
Steve French 已提交
986
				== 0))) {
987 988 989
				struct kvec iov[2];
				unsigned int len;

990
				len = min((size_t)cifs_sb->wsize,
991 992 993 994 995
					  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;
996
				rc = CIFSSMBWrite2(xid, pTcon,
997
						open_file->netfid, len,
998
						*poffset, &bytes_written,
999
						iov, 1, long_op);
1000
			} else
1001 1002 1003 1004 1005 1006 1007
				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
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
		}
		if (rc || (bytes_written == 0)) {
			if (total_written)
				break;
			else {
				FreeXid(xid);
				return rc;
			}
		} else
			*poffset += bytes_written;
1018
		long_op = CIFS_STD_OP; /* subsequent writes fast -
L
Linus Torvalds 已提交
1019 1020 1021
				    15 seconds is plenty */
	}

1022
	cifs_stats_bytes_written(pTcon, total_written);
L
Linus Torvalds 已提交
1023 1024

	/* since the write may have blocked check these pointers again */
1025
	if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
1026
/*BB We could make this contingent on superblock ATIME flag too */
1027 1028 1029 1030 1031 1032 1033 1034
/*		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
1035
		}
1036
		mark_inode_dirty_sync(file->f_path.dentry->d_inode);
L
Linus Torvalds 已提交
1037 1038 1039 1040 1041
	}
	FreeXid(xid);
	return total_written;
}

S
Steve French 已提交
1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
#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 */
				atomic_inc(&open_file->wrtPending);
				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

1073
struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode)
1074 1075
{
	struct cifsFileInfo *open_file;
1076
	bool any_available = false;
1077
	int rc;
1078

1079 1080 1081 1082
	/* 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
1083 1084
	if (cifs_inode == NULL) {
		cERROR(1, ("Null inode passed to cifs_writeable_file"));
1085 1086 1087 1088
		dump_stack();
		return NULL;
	}

1089
	read_lock(&GlobalSMBSeslock);
1090
refind_writable:
1091
	list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
1092 1093
		if (open_file->closePend ||
		    (!any_available && open_file->pid != current->tgid))
1094
			continue;
1095

1096 1097 1098
		if (open_file->pfile &&
		    ((open_file->pfile->f_flags & O_RDWR) ||
		     (open_file->pfile->f_flags & O_WRONLY))) {
1099
			atomic_inc(&open_file->wrtPending);
1100 1101 1102 1103 1104 1105

			if (!open_file->invalidHandle) {
				/* found a good writable file */
				read_unlock(&GlobalSMBSeslock);
				return open_file;
			}
S
Steve French 已提交
1106

1107
			read_unlock(&GlobalSMBSeslock);
1108
			/* Had to unlock since following call can block */
1109
			rc = cifs_reopen_file(open_file->pfile, false);
S
Steve French 已提交
1110
			if (!rc) {
1111 1112 1113 1114
				if (!open_file->closePend)
					return open_file;
				else { /* start over in case this was deleted */
				       /* since the list could be modified */
1115
					read_lock(&GlobalSMBSeslock);
1116
					atomic_dec(&open_file->wrtPending);
1117
					goto refind_writable;
1118 1119
				}
			}
1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132

			/* 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 */
			cFYI(1, ("wp failed on reopen file"));
			read_lock(&GlobalSMBSeslock);
			/* can not use this handle, no write
			   pending on this one after all */
			atomic_dec(&open_file->wrtPending);
S
Steve French 已提交
1133

1134 1135 1136 1137 1138 1139 1140 1141 1142
			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. */
1143 1144
		}
	}
1145 1146 1147 1148 1149
	/* couldn't find useable FH with same pid, try any available */
	if (!any_available) {
		any_available = true;
		goto refind_writable;
	}
1150 1151 1152 1153
	read_unlock(&GlobalSMBSeslock);
	return NULL;
}

L
Linus Torvalds 已提交
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163
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;
1164
	struct cifsFileInfo *open_file;
L
Linus Torvalds 已提交
1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189

	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
1190
		to = (unsigned)(mapping->host->i_size - offset);
L
Linus Torvalds 已提交
1191

1192 1193 1194 1195
	open_file = find_writable_file(CIFS_I(mapping->host));
	if (open_file) {
		bytes_written = cifs_write(open_file->pfile, write_data,
					   to-from, &offset);
1196
		atomic_dec(&open_file->wrtPending);
L
Linus Torvalds 已提交
1197
		/* Does mm or vfs already set times? */
1198
		inode->i_atime = inode->i_mtime = current_fs_time(inode->i_sb);
1199
		if ((bytes_written > 0) && (offset))
1200
			rc = 0;
1201 1202
		else if (bytes_written < 0)
			rc = bytes_written;
1203
	} else {
L
Linus Torvalds 已提交
1204 1205 1206 1207 1208 1209 1210 1211 1212
		cFYI(1, ("No writeable filehandles for inode"));
		rc = -EIO;
	}

	kunmap(page);
	return rc;
}

static int cifs_writepages(struct address_space *mapping,
1213
			   struct writeback_control *wbc)
L
Linus Torvalds 已提交
1214
{
1215 1216 1217 1218 1219
	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;
1220
	pgoff_t end;
1221
	pgoff_t index;
S
Steve French 已提交
1222 1223
	int range_whole = 0;
	struct kvec *iov;
1224
	int len;
1225 1226 1227 1228
	int n_iov = 0;
	pgoff_t next;
	int nr_pages;
	__u64 offset = 0;
1229
	struct cifsFileInfo *open_file;
1230 1231 1232 1233
	struct page *page;
	struct pagevec pvec;
	int rc = 0;
	int scanned = 0;
L
Linus Torvalds 已提交
1234 1235
	int xid;

1236
	cifs_sb = CIFS_SB(mapping->host->i_sb);
1237

1238 1239 1240 1241 1242 1243 1244
	/*
	 * 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
1245 1246 1247 1248
	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)
1249
				return generic_writepages(mapping, wbc);
1250

1251
	iov = kmalloc(32 * sizeof(struct kvec), GFP_KERNEL);
S
Steve French 已提交
1252
	if (iov == NULL)
1253 1254 1255
		return generic_writepages(mapping, wbc);


1256 1257 1258 1259 1260 1261
	/*
	 * 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;
1262
		kfree(iov);
1263 1264 1265
		return 0;
	}

L
Linus Torvalds 已提交
1266 1267
	xid = GetXid();

1268
	pagevec_init(&pvec, 0);
1269
	if (wbc->range_cyclic) {
1270
		index = mapping->writeback_index; /* Start from prev offset */
1271 1272 1273 1274 1275 1276
		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;
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
		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
1304
			else if (!trylock_page(page))
1305 1306 1307 1308 1309 1310 1311
				break;

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

1312
			if (!wbc->range_cyclic && page->index > end) {
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
				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) ||
1328
					!clear_page_dirty_for_io(page)) {
1329 1330 1331
				unlock_page(page);
				break;
			}
1332

1333 1334 1335 1336 1337 1338
			/*
			 * This actually clears the dirty bit in the radix tree.
			 * See cifs_writepage() for more commentary.
			 */
			set_page_writeback(page);

1339 1340 1341
			if (page_offset(page) >= mapping->host->i_size) {
				done = 1;
				unlock_page(page);
1342
				end_page_writeback(page);
1343 1344 1345
				break;
			}

1346 1347 1348 1349 1350
			/*
			 * BB can we get rid of this?  pages are held by pvec
			 */
			page_cache_get(page);

1351 1352 1353
			len = min(mapping->host->i_size - page_offset(page),
				  (loff_t)PAGE_CACHE_SIZE);

1354 1355 1356
			/* reserve iov[0] for the smb header */
			n_iov++;
			iov[n_iov].iov_base = kmap(page);
1357 1358
			iov[n_iov].iov_len = len;
			bytes_to_write += len;
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368

			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) {
1369 1370 1371 1372 1373 1374 1375 1376
			/* 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) {
				cERROR(1, ("No writable handles for inode"));
				rc = -EBADF;
S
Steve French 已提交
1377
			} else {
1378 1379 1380 1381
				rc = CIFSSMBWrite2(xid, cifs_sb->tcon,
						   open_file->netfid,
						   bytes_to_write, offset,
						   &bytes_written, iov, n_iov,
1382
						   CIFS_LONG_OP);
1383 1384
				atomic_dec(&open_file->wrtPending);
				if (rc || bytes_written < bytes_to_write) {
1385
					cERROR(1, ("Write2 ret %d, wrote %d",
1386 1387 1388
						  rc, bytes_written));
					/* BB what if continued retry is
					   requested via mount flags? */
1389 1390 1391 1392
					if (rc == -ENOSPC)
						set_bit(AS_ENOSPC, &mapping->flags);
					else
						set_bit(AS_EIO, &mapping->flags);
1393 1394 1395 1396
				} else {
					cifs_stats_bytes_written(cifs_sb->tcon,
								 bytes_written);
				}
1397 1398 1399
			}
			for (i = 0; i < n_iov; i++) {
				page = pvec.pages[first + i];
1400 1401 1402 1403
				/* 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
1404 1405
				when page marked as error */
				if (rc)
1406
					SetPageError(page);
1407 1408
				kunmap(page);
				unlock_page(page);
1409
				end_page_writeback(page);
1410 1411 1412 1413 1414
				page_cache_release(page);
			}
			if ((wbc->nr_to_write -= n_iov) <= 0)
				done = 1;
			index = next;
1415 1416 1417 1418
		} else
			/* Need to re-find the pages we skipped */
			index = pvec.pages[0]->index + 1;

1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429
		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;
	}
1430
	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1431 1432
		mapping->writeback_index = index;

L
Linus Torvalds 已提交
1433
	FreeXid(xid);
1434
	kfree(iov);
L
Linus Torvalds 已提交
1435 1436 1437
	return rc;
}

S
Steve French 已提交
1438
static int cifs_writepage(struct page *page, struct writeback_control *wbc)
L
Linus Torvalds 已提交
1439 1440 1441 1442 1443 1444 1445
{
	int rc = -EFAULT;
	int xid;

	xid = GetXid();
/* BB add check for wbc flags */
	page_cache_get(page);
S
Steve French 已提交
1446
	if (!PageUptodate(page))
L
Linus Torvalds 已提交
1447
		cFYI(1, ("ppw - page not up to date"));
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458

	/*
	 * 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
1459
	set_page_writeback(page);
L
Linus Torvalds 已提交
1460 1461 1462
	rc = cifs_partialpagewrite(page, 0, PAGE_CACHE_SIZE);
	SetPageUptodate(page); /* BB add check for error and Clearuptodate? */
	unlock_page(page);
1463 1464
	end_page_writeback(page);
	page_cache_release(page);
L
Linus Torvalds 已提交
1465 1466 1467 1468
	FreeXid(xid);
	return rc;
}

N
Nick Piggin 已提交
1469 1470 1471
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
1472
{
N
Nick Piggin 已提交
1473 1474
	int rc;
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
1475

N
Nick Piggin 已提交
1476 1477 1478
	cFYI(1, ("write_end for page %p from pos %lld with %d bytes",
		 page, pos, copied));

1479 1480 1481 1482 1483
	if (PageChecked(page)) {
		if (copied == len)
			SetPageUptodate(page);
		ClearPageChecked(page);
	} else if (!PageUptodate(page) && copied == PAGE_CACHE_SIZE)
N
Nick Piggin 已提交
1484
		SetPageUptodate(page);
S
Steve French 已提交
1485

L
Linus Torvalds 已提交
1486
	if (!PageUptodate(page)) {
N
Nick Piggin 已提交
1487 1488 1489 1490 1491
		char *page_data;
		unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
		int xid;

		xid = GetXid();
L
Linus Torvalds 已提交
1492 1493 1494 1495 1496 1497
		/* 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
1498 1499
		rc = cifs_write(file, page_data + offset, copied, &pos);
		/* if (rc < 0) should we set writebehind rc? */
L
Linus Torvalds 已提交
1500
		kunmap(page);
N
Nick Piggin 已提交
1501 1502

		FreeXid(xid);
S
Steve French 已提交
1503
	} else {
N
Nick Piggin 已提交
1504 1505
		rc = copied;
		pos += copied;
L
Linus Torvalds 已提交
1506 1507 1508
		set_page_dirty(page);
	}

N
Nick Piggin 已提交
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
	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
1519 1520 1521 1522 1523 1524 1525
	return rc;
}

int cifs_fsync(struct file *file, struct dentry *dentry, int datasync)
{
	int xid;
	int rc = 0;
1526
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1527 1528 1529

	xid = GetXid();

S
Steve French 已提交
1530
	cFYI(1, ("Sync file - name: %s datasync: 0x%x",
L
Linus Torvalds 已提交
1531
		dentry->d_name.name, datasync));
1532

1533 1534 1535
	rc = filemap_write_and_wait(inode->i_mapping);
	if (rc == 0) {
		rc = CIFS_I(inode)->write_behind_rc;
L
Linus Torvalds 已提交
1536
		CIFS_I(inode)->write_behind_rc = 0;
1537
	}
L
Linus Torvalds 已提交
1538 1539 1540 1541
	FreeXid(xid);
	return rc;
}

1542
/* static void cifs_sync_page(struct page *page)
L
Linus Torvalds 已提交
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
{
	struct address_space *mapping;
	struct inode *inode;
	unsigned long index = page->index;
	unsigned int rpages = 0;
	int rc = 0;

	cFYI(1, ("sync page %p",page));
	mapping = page->mapping;
	if (!mapping)
		return 0;
	inode = mapping->host;
	if (!inode)
1556
		return; */
L
Linus Torvalds 已提交
1557

S
Steve French 已提交
1558
/*	fill in rpages then
L
Linus Torvalds 已提交
1559 1560
	result = cifs_pagein_inode(inode, index, rpages); */ /* BB finish */

1561
/*	cFYI(1, ("rpages is %d for sync page of Index %ld", rpages, index));
L
Linus Torvalds 已提交
1562

1563
#if 0
L
Linus Torvalds 已提交
1564 1565 1566
	if (rc < 0)
		return rc;
	return 0;
1567
#endif
L
Linus Torvalds 已提交
1568 1569 1570 1571 1572 1573
} */

/*
 * As file closes, flush all cached write data for this inode checking
 * for write behind errors.
 */
1574
int cifs_flush(struct file *file, fl_owner_t id)
L
Linus Torvalds 已提交
1575
{
S
Steve French 已提交
1576
	struct inode *inode = file->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
	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);
1589 1590 1591
	/* 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
1592
		CIFS_I(inode)->write_behind_rc = 0;
1593
	}
1594

S
Steve French 已提交
1595
	cFYI(1, ("Flush inode %p file %p rc %d", inode, file, rc));
L
Linus Torvalds 已提交
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615

	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();
1616
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1617 1618 1619 1620 1621 1622 1623 1624
	pTcon = cifs_sb->tcon;

	if (file->private_data == NULL) {
		FreeXid(xid);
		return -EBADF;
	}
	open_file = (struct cifsFileInfo *)file->private_data;

S
Steve French 已提交
1625
	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
L
Linus Torvalds 已提交
1626
		cFYI(1, ("attempting read on write only file instance"));
S
Steve French 已提交
1627

L
Linus Torvalds 已提交
1628 1629 1630
	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
1631
		current_read_size = min_t(const int, read_size - total_read,
L
Linus Torvalds 已提交
1632 1633 1634 1635
					  cifs_sb->rsize);
		rc = -EAGAIN;
		smb_read_data = NULL;
		while (rc == -EAGAIN) {
1636
			int buf_type = CIFS_NO_BUFFER;
S
Steve French 已提交
1637
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1638
			    (!open_file->closePend)) {
1639
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1640 1641 1642
				if (rc != 0)
					break;
			}
1643
			rc = CIFSSMBRead(xid, pTcon,
1644 1645 1646 1647
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
L
Linus Torvalds 已提交
1648 1649
			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
			if (smb_read_data) {
1650 1651 1652 1653
				if (copy_to_user(current_offset,
						smb_read_data +
						4 /* RFC1001 length field */ +
						le16_to_cpu(pSMBr->DataOffset),
S
Steve French 已提交
1654
						bytes_read))
1655 1656
					rc = -EFAULT;

S
Steve French 已提交
1657
				if (buf_type == CIFS_SMALL_BUFFER)
1658
					cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1659
				else if (buf_type == CIFS_LARGE_BUFFER)
1660
					cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
				smb_read_data = NULL;
			}
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1672
			cifs_stats_bytes_read(pTcon, bytes_read);
L
Linus Torvalds 已提交
1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
			*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;
1693
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1694 1695

	xid = GetXid();
1696
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
	pTcon = cifs_sb->tcon;

	if (file->private_data == NULL) {
		FreeXid(xid);
		return -EBADF;
	}
	open_file = (struct cifsFileInfo *)file->private_data;

	if ((file->f_flags & O_ACCMODE) == O_WRONLY)
		cFYI(1, ("attempting read on write only file instance"));

S
Steve French 已提交
1708
	for (total_read = 0, current_offset = read_data;
L
Linus Torvalds 已提交
1709 1710 1711 1712
	     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);
1713 1714
		/* 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
1715
		if ((pTcon->ses) &&
1716 1717 1718 1719
			!(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
1720 1721
		rc = -EAGAIN;
		while (rc == -EAGAIN) {
S
Steve French 已提交
1722
			if ((open_file->invalidHandle) &&
L
Linus Torvalds 已提交
1723
			    (!open_file->closePend)) {
1724
				rc = cifs_reopen_file(file, true);
L
Linus Torvalds 已提交
1725 1726 1727
				if (rc != 0)
					break;
			}
1728
			rc = CIFSSMBRead(xid, pTcon,
1729 1730 1731 1732
					 open_file->netfid,
					 current_read_size, *poffset,
					 &bytes_read, &current_offset,
					 &buf_type);
L
Linus Torvalds 已提交
1733 1734 1735 1736 1737 1738 1739 1740 1741
		}
		if (rc || (bytes_read == 0)) {
			if (total_read) {
				break;
			} else {
				FreeXid(xid);
				return rc;
			}
		} else {
1742
			cifs_stats_bytes_read(pTcon, total_read);
L
Linus Torvalds 已提交
1743 1744 1745 1746 1747 1748 1749 1750 1751
			*poffset += bytes_read;
		}
	}
	FreeXid(xid);
	return total_read;
}

int cifs_file_mmap(struct file *file, struct vm_area_struct *vma)
{
1752
	struct dentry *dentry = file->f_path.dentry;
L
Linus Torvalds 已提交
1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767
	int rc, xid;

	xid = GetXid();
	rc = cifs_revalidate(dentry);
	if (rc) {
		cFYI(1, ("Validation prior to mmap failed, error=%d", rc));
		FreeXid(xid);
		return rc;
	}
	rc = generic_file_mmap(file, vma);
	FreeXid(xid);
	return rc;
}


S
Steve French 已提交
1768
static void cifs_copy_cache_pages(struct address_space *mapping,
L
Linus Torvalds 已提交
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785
	struct list_head *pages, int bytes_read, char *data,
	struct pagevec *plru_pvec)
{
	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);

		if (add_to_page_cache(page, mapping, page->index,
				      GFP_KERNEL)) {
			page_cache_release(page);
			cFYI(1, ("Add page cache failed"));
1786 1787
			data += PAGE_CACHE_SIZE;
			bytes_read -= PAGE_CACHE_SIZE;
L
Linus Torvalds 已提交
1788 1789 1790
			continue;
		}

S
Steve French 已提交
1791
		target = kmap_atomic(page, KM_USER0);
L
Linus Torvalds 已提交
1792 1793 1794 1795

		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
1796
			memset(target + bytes_read, 0,
L
Linus Torvalds 已提交
1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808
			       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);
		if (!pagevec_add(plru_pvec, page))
1809
			__pagevec_lru_add_file(plru_pvec);
L
Linus Torvalds 已提交
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
		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
1824
	unsigned int bytes_read = 0;
S
Steve French 已提交
1825
	unsigned int read_size, i;
L
Linus Torvalds 已提交
1826 1827 1828 1829
	char *smb_read_data = NULL;
	struct smb_com_read_rsp *pSMBr;
	struct pagevec lru_pvec;
	struct cifsFileInfo *open_file;
1830
	int buf_type = CIFS_NO_BUFFER;
L
Linus Torvalds 已提交
1831 1832 1833 1834 1835 1836 1837

	xid = GetXid();
	if (file->private_data == NULL) {
		FreeXid(xid);
		return -EBADF;
	}
	open_file = (struct cifsFileInfo *)file->private_data;
1838
	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
L
Linus Torvalds 已提交
1839
	pTcon = cifs_sb->tcon;
1840

L
Linus Torvalds 已提交
1841
	pagevec_init(&lru_pvec, 0);
S
Steve French 已提交
1842
	cFYI(DBG2, ("rpages: num pages %d", num_pages));
L
Linus Torvalds 已提交
1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
	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
1856
		expected_index =
L
Linus Torvalds 已提交
1857
			list_entry(page_list->prev, struct page, lru)->index;
S
Steve French 已提交
1858
		list_for_each_entry_reverse(tmp_page, page_list, lru) {
L
Linus Torvalds 已提交
1859 1860 1861 1862
			if (tmp_page->index == expected_index) {
				contig_pages++;
				expected_index++;
			} else
S
Steve French 已提交
1863
				break;
L
Linus Torvalds 已提交
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
		}
		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);
1875
		cFYI(DBG2, ("rpages: read size 0x%x  contiguous pages %d",
1876
				read_size, contig_pages));
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 1885
				if (rc != 0)
					break;
			}

1886
			rc = CIFSSMBRead(xid, pTcon,
1887 1888 1889 1890
					 open_file->netfid,
					 read_size, offset,
					 &bytes_read, &smb_read_data,
					 &buf_type);
1891
			/* BB more RC checks ? */
S
Steve French 已提交
1892
			if (rc == -EAGAIN) {
L
Linus Torvalds 已提交
1893
				if (smb_read_data) {
S
Steve French 已提交
1894
					if (buf_type == CIFS_SMALL_BUFFER)
1895
						cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1896
					else if (buf_type == CIFS_LARGE_BUFFER)
1897
						cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1898 1899 1900 1901 1902 1903 1904 1905
					smb_read_data = NULL;
				}
			}
		}
		if ((rc < 0) || (smb_read_data == NULL)) {
			cFYI(1, ("Read error in readpages: %d", rc));
			break;
		} else if (bytes_read > 0) {
1906
			task_io_account_read(bytes_read);
L
Linus Torvalds 已提交
1907 1908 1909 1910 1911 1912
			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
			cifs_copy_cache_pages(mapping, page_list, bytes_read,
				smb_read_data + 4 /* RFC1001 hdr */ +
				le16_to_cpu(pSMBr->DataOffset), &lru_pvec);

			i +=  bytes_read >> PAGE_CACHE_SHIFT;
1913
			cifs_stats_bytes_read(pTcon, bytes_read);
S
Steve French 已提交
1914
			if ((bytes_read & PAGE_CACHE_MASK) != bytes_read) {
L
Linus Torvalds 已提交
1915 1916
				i++; /* account for partial page */

S
Steve French 已提交
1917
				/* server copy of file can have smaller size
L
Linus Torvalds 已提交
1918
				   than client */
S
Steve French 已提交
1919 1920
				/* 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
1921 1922
				   we will hit it on next read */

O
OGAWA Hirofumi 已提交
1923
				/* break; */
L
Linus Torvalds 已提交
1924 1925 1926 1927 1928
			}
		} else {
			cFYI(1, ("No bytes read (%d) at offset %lld . "
				 "Cleaning remaining pages from readahead list",
				 bytes_read, offset));
S
Steve French 已提交
1929
			/* BB turn off caching and do new lookup on
L
Linus Torvalds 已提交
1930 1931 1932 1933
			   file size at server? */
			break;
		}
		if (smb_read_data) {
S
Steve French 已提交
1934
			if (buf_type == CIFS_SMALL_BUFFER)
1935
				cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1936
			else if (buf_type == CIFS_LARGE_BUFFER)
1937
				cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1938 1939 1940 1941 1942
			smb_read_data = NULL;
		}
		bytes_read = 0;
	}

1943
	pagevec_lru_add_file(&lru_pvec);
L
Linus Torvalds 已提交
1944 1945 1946

/* need to free smb_read_data buf before exit */
	if (smb_read_data) {
S
Steve French 已提交
1947
		if (buf_type == CIFS_SMALL_BUFFER)
1948
			cifs_small_buf_release(smb_read_data);
S
Steve French 已提交
1949
		else if (buf_type == CIFS_LARGE_BUFFER)
1950
			cifs_buf_release(smb_read_data);
L
Linus Torvalds 已提交
1951
		smb_read_data = NULL;
S
Steve French 已提交
1952
	}
L
Linus Torvalds 已提交
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966

	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
1967

L
Linus Torvalds 已提交
1968
	rc = cifs_read(file, read_data, PAGE_CACHE_SIZE, poffset);
S
Steve French 已提交
1969

L
Linus Torvalds 已提交
1970 1971 1972
	if (rc < 0)
		goto io_error;
	else
S
Steve French 已提交
1973 1974
		cFYI(1, ("Bytes read %d", rc));

1975 1976
	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
1977

L
Linus Torvalds 已提交
1978 1979 1980 1981 1982 1983
	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
1984

L
Linus Torvalds 已提交
1985
io_error:
S
Steve French 已提交
1986
	kunmap(page);
L
Linus Torvalds 已提交
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
	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) {
		FreeXid(xid);
		return -EBADF;
	}

S
Steve French 已提交
2004
	cFYI(1, ("readpage %p at offset %d 0x%x\n",
L
Linus Torvalds 已提交
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
		 page, (int)offset, (int)offset));

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

	unlock_page(page);

	FreeXid(xid);
	return rc;
}

2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033
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
2034 2035 2036
/* 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
2037
   refreshing the inode only on increases in the file size
L
Linus Torvalds 已提交
2038 2039
   but this is tricky to do without racing with writebehind
   page caching in the current Linux kernel design */
2040
bool is_size_safe_to_change(struct cifsInodeInfo *cifsInode, __u64 end_of_file)
L
Linus Torvalds 已提交
2041
{
2042
	if (!cifsInode)
2043
		return true;
2044

2045 2046
	if (is_inode_writable(cifsInode)) {
		/* This inode is open for write at least once */
2047 2048 2049
		struct cifs_sb_info *cifs_sb;

		cifs_sb = CIFS_SB(cifsInode->vfs_inode.i_sb);
S
Steve French 已提交
2050
		if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DIRECT_IO) {
S
Steve French 已提交
2051
			/* since no page cache to corrupt on directio
2052
			we can change size safely */
2053
			return true;
2054 2055
		}

S
Steve French 已提交
2056
		if (i_size_read(&cifsInode->vfs_inode) < end_of_file)
2057
			return true;
2058

2059
		return false;
2060
	} else
2061
		return true;
L
Linus Torvalds 已提交
2062 2063
}

N
Nick Piggin 已提交
2064 2065 2066
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
2067
{
N
Nick Piggin 已提交
2068 2069
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	loff_t offset = pos & (PAGE_CACHE_SIZE - 1);
2070 2071 2072 2073
	loff_t page_start = pos & PAGE_MASK;
	loff_t i_size;
	struct page *page;
	int rc = 0;
N
Nick Piggin 已提交
2074 2075 2076

	cFYI(1, ("write_begin from %lld len %d", (long long)pos, len));

2077 2078 2079 2080 2081
	page = __grab_cache_page(mapping, index);
	if (!page) {
		rc = -ENOMEM;
		goto out;
	}
2082

2083 2084
	if (PageUptodate(page))
		goto out;
2085

2086 2087 2088 2089 2090 2091 2092
	/*
	 * 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;
2093

2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
	/*
	 * 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
2117

2118 2119 2120 2121 2122 2123 2124
	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);
2125 2126 2127 2128
	} 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
2129
		   this will be written out by write_end so is fine */
L
Linus Torvalds 已提交
2130
	}
2131 2132 2133
out:
	*pagep = page;
	return rc;
L
Linus Torvalds 已提交
2134 2135
}

2136
const struct address_space_operations cifs_addr_ops = {
L
Linus Torvalds 已提交
2137 2138 2139
	.readpage = cifs_readpage,
	.readpages = cifs_readpages,
	.writepage = cifs_writepage,
2140
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2141 2142
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
L
Linus Torvalds 已提交
2143 2144 2145 2146
	.set_page_dirty = __set_page_dirty_nobuffers,
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};
D
Dave Kleikamp 已提交
2147 2148 2149 2150 2151 2152

/*
 * 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.
 */
2153
const struct address_space_operations cifs_addr_ops_smallbuf = {
D
Dave Kleikamp 已提交
2154 2155 2156
	.readpage = cifs_readpage,
	.writepage = cifs_writepage,
	.writepages = cifs_writepages,
N
Nick Piggin 已提交
2157 2158
	.write_begin = cifs_write_begin,
	.write_end = cifs_write_end,
D
Dave Kleikamp 已提交
2159 2160 2161 2162
	.set_page_dirty = __set_page_dirty_nobuffers,
	/* .sync_page = cifs_sync_page, */
	/* .direct_IO = */
};