2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* fs/cifs/file.c
|
|
|
|
*
|
|
|
|
* vfs operations that deal with files
|
2007-07-09 19:16:18 -06:00
|
|
|
*
|
|
|
|
* Copyright (C) International Business Machines Corp., 2002,2007
|
2005-04-16 16:20:36 -06:00
|
|
|
* Author(s): Steve French (sfrench@us.ibm.com)
|
2006-08-02 15:56:33 -06:00
|
|
|
* Jeremy Allison (jra@samba.org)
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
|
|
|
* 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>
|
2005-10-05 15:50:29 -06:00
|
|
|
#include <linux/backing-dev.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <linux/stat.h>
|
|
|
|
#include <linux/fcntl.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/pagevec.h>
|
2005-10-05 15:50:29 -06:00
|
|
|
#include <linux/writeback.h>
|
2006-12-10 03:19:44 -07:00
|
|
|
#include <linux/task_io_accounting_ops.h>
|
2005-10-20 14:44:56 -06:00
|
|
|
#include <linux/delay.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#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;
|
2007-07-09 19:16:18 -06:00
|
|
|
private_data->pid = current->tgid;
|
2005-04-16 16:20:36 -06:00
|
|
|
init_MUTEX(&private_data->fh_sem);
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_init(&private_data->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
INIT_LIST_HEAD(&private_data->llist);
|
2005-04-16 16:20:36 -06:00
|
|
|
private_data->pfile = file; /* needed for writepage */
|
|
|
|
private_data->pInode = inode;
|
2008-04-28 18:06:05 -06:00
|
|
|
private_data->invalidHandle = false;
|
|
|
|
private_data->closePend = false;
|
2005-10-20 14:44:56 -06:00
|
|
|
/* 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 */
|
2007-07-09 19:16:18 -06:00
|
|
|
atomic_set(&private_data->wrtPending, 0);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2008-05-14 11:21:33 -06:00
|
|
|
return (READ_CONTROL | FILE_WRITE_ATTRIBUTES | FILE_READ_ATTRIBUTES |
|
|
|
|
FILE_WRITE_EA | FILE_APPEND_DATA | FILE_WRITE_DATA |
|
|
|
|
FILE_READ_DATA);
|
|
|
|
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
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;
|
2006-05-30 12:09:31 -06:00
|
|
|
else if ((flags & O_TRUNC) == O_TRUNC)
|
|
|
|
return FILE_OVERWRITE;
|
2005-04-16 16:20:36 -06:00
|
|
|
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
|
2008-09-24 09:32:59 -06:00
|
|
|
list to search for one in write_begin */
|
2005-04-16 16:20:36 -06:00
|
|
|
if ((file->f_flags & O_ACCMODE) == O_WRONLY) {
|
2007-07-09 19:16:18 -06:00
|
|
|
list_add_tail(&pCifsFile->flist,
|
2005-04-16 16:20:36 -06:00
|
|
|
&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));
|
2006-12-08 03:36:48 -07:00
|
|
|
if (timespec_equal(&file->f_path.dentry->d_inode->i_mtime, &temp) &&
|
|
|
|
(file->f_path.dentry->d_inode->i_size ==
|
2005-04-16 16:20:36 -06:00
|
|
|
(loff_t)le64_to_cpu(buf->EndOfFile))) {
|
|
|
|
cFYI(1, ("inode unchanged on server"));
|
|
|
|
} else {
|
2006-12-08 03:36:48 -07:00
|
|
|
if (file->f_path.dentry->d_inode->i_mapping) {
|
2005-04-16 16:20:36 -06:00
|
|
|
/* BB no need to lock inode until after invalidate
|
|
|
|
since namei code should already have it locked? */
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
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;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
cFYI(1, ("invalidating remote inode since open detected it "
|
|
|
|
"changed"));
|
2006-12-08 03:36:48 -07:00
|
|
|
invalidate_remote_inode(file->f_path.dentry->d_inode);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
client_can_cache:
|
2007-07-18 17:21:09 -06:00
|
|
|
if (pTcon->unix_ext)
|
2006-12-08 03:36:48 -07:00
|
|
|
rc = cifs_get_inode_info_unix(&file->f_path.dentry->d_inode,
|
2005-04-16 16:20:36 -06:00
|
|
|
full_path, inode->i_sb, xid);
|
|
|
|
else
|
2006-12-08 03:36:48 -07:00
|
|
|
rc = cifs_get_inode_info(&file->f_path.dentry->d_inode,
|
2008-03-14 16:37:16 -06:00
|
|
|
full_path, buf, inode->i_sb, xid, NULL);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if ((*oplock & 0xF) == OPLOCK_EXCLUSIVE) {
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsInode->clientCanCacheAll = true;
|
|
|
|
pCifsInode->clientCanCacheRead = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
cFYI(1, ("Exclusive Oplock granted on inode %p",
|
2006-12-08 03:36:48 -07:00
|
|
|
file->f_path.dentry->d_inode));
|
2005-04-16 16:20:36 -06:00
|
|
|
} else if ((*oplock & 0xF) == OPLOCK_READ)
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsInode->clientCanCacheRead = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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 */
|
2006-12-08 03:36:48 -07:00
|
|
|
pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
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 "
|
2006-05-31 12:05:34 -06:00
|
|
|
"new file %p", file));
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
full_path = build_path_from_dentry(file->f_path.dentry);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (full_path == NULL) {
|
|
|
|
FreeXid(xid);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2007-07-11 12:30:34 -06:00
|
|
|
cFYI(1, ("inode = 0x%p file flags are 0x%x for %s",
|
2005-04-16 16:20:36 -06:00
|
|
|
inode, file->f_flags, full_path));
|
|
|
|
desiredAccess = cifs_convert_flags(file->f_flags);
|
|
|
|
|
|
|
|
/*********************************************************************
|
|
|
|
* open flag mapping table:
|
2007-07-09 19:16:18 -06:00
|
|
|
*
|
2005-04-16 16:20:36 -06:00
|
|
|
* POSIX Flag CIFS Disposition
|
2007-07-09 19:16:18 -06:00
|
|
|
* ---------- ----------------
|
2005-04-16 16:20:36 -06:00
|
|
|
* 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
|
2007-07-09 19:16:18 -06:00
|
|
|
* FILE_SUPERSEDE (ie create whether or not file exists although
|
2005-04-16 16:20:36 -06:00
|
|
|
* 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)
|
|
|
|
*?
|
2007-07-09 19:16:18 -06:00
|
|
|
*? O_SYNC is a reasonable match to CIFS writethrough flag
|
2005-04-16 16:20:36 -06:00
|
|
|
*? 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
|
2008-04-28 18:06:05 -06:00
|
|
|
oplock = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* 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) */
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
/* 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
|
2005-04-16 16:20:36 -06:00
|
|
|
able to simply do a filemap_fdatawrite/filemap_fdatawait first */
|
|
|
|
buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
|
|
|
|
if (!buf) {
|
|
|
|
rc = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2006-06-06 18:18:43 -06:00
|
|
|
|
|
|
|
if (cifs_sb->tcon->ses->capabilities & CAP_NT_SMBS)
|
2007-07-09 19:16:18 -06:00
|
|
|
rc = CIFSSMBOpen(xid, pTcon, full_path, disposition,
|
2006-06-06 18:18:43 -06:00
|
|
|
desiredAccess, CREATE_NOT_DIR, &netfid, &oplock, buf,
|
2005-04-28 23:41:06 -06:00
|
|
|
cifs_sb->local_nls, cifs_sb->mnt_cifs_flags
|
|
|
|
& CIFS_MOUNT_MAP_SPECIAL_CHR);
|
2006-06-06 18:18:43 -06:00
|
|
|
else
|
|
|
|
rc = -EIO; /* no NT SMB support fall into legacy open below */
|
|
|
|
|
2005-08-25 00:06:05 -06:00
|
|
|
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);
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc) {
|
2006-05-31 12:05:34 -06:00
|
|
|
cFYI(1, ("cifs_open returned 0x%x", rc));
|
2005-04-16 16:20:36 -06:00
|
|
|
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);
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
pCifsInode = CIFS_I(file->f_path.dentry->d_inode);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (pCifsInode) {
|
|
|
|
rc = cifs_open_inode_helper(inode, file, pCifsInode,
|
|
|
|
pCifsFile, pTcon,
|
|
|
|
&oplock, buf, full_path, xid);
|
|
|
|
} else {
|
|
|
|
write_unlock(&GlobalSMBSeslock);
|
|
|
|
}
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
if (oplock & CIFS_CREATE_ACTION) {
|
2005-04-16 16:20:36 -06:00
|
|
|
/* time to set mode which we can not set earlier due to
|
|
|
|
problems creating new read-only files */
|
2007-07-18 17:21:09 -06:00
|
|
|
if (pTcon->unix_ext) {
|
2008-08-02 05:26:12 -06:00
|
|
|
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,
|
2005-04-28 23:41:06 -06:00
|
|
|
cifs_sb->local_nls,
|
2007-07-09 19:16:18 -06:00
|
|
|
cifs_sb->mnt_cifs_flags &
|
2005-04-28 23:41:06 -06:00
|
|
|
CIFS_MOUNT_MAP_SPECIAL_CHR);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
kfree(buf);
|
|
|
|
kfree(full_path);
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2006-06-30 10:23:04 -06:00
|
|
|
/* Try to reacquire byte range locks that were released when session */
|
2005-04-16 16:20:36 -06:00
|
|
|
/* 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;
|
|
|
|
}
|
|
|
|
|
2008-04-28 18:06:05 -06:00
|
|
|
static int cifs_reopen_file(struct file *file, bool can_flush)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int rc = -EACCES;
|
|
|
|
int xid, oplock;
|
|
|
|
struct cifs_sb_info *cifs_sb;
|
|
|
|
struct cifsTconInfo *pTcon;
|
|
|
|
struct cifsFileInfo *pCifsFile;
|
|
|
|
struct cifsInodeInfo *pCifsInode;
|
2007-07-09 19:16:18 -06:00
|
|
|
struct inode *inode;
|
2005-04-16 16:20:36 -06:00
|
|
|
char *full_path = NULL;
|
|
|
|
int desiredAccess;
|
|
|
|
int disposition = FILE_OPEN;
|
|
|
|
__u16 netfid;
|
|
|
|
|
2008-02-07 16:25:02 -07:00
|
|
|
if (file->private_data)
|
2005-04-16 16:20:36 -06:00
|
|
|
pCifsFile = (struct cifsFileInfo *)file->private_data;
|
2008-02-07 16:25:02 -07:00
|
|
|
else
|
2005-04-16 16:20:36 -06:00
|
|
|
return -EBADF;
|
|
|
|
|
|
|
|
xid = GetXid();
|
|
|
|
down(&pCifsFile->fh_sem);
|
2008-04-28 18:06:05 -06:00
|
|
|
if (!pCifsFile->invalidHandle) {
|
2005-04-16 16:20:36 -06:00
|
|
|
up(&pCifsFile->fh_sem);
|
|
|
|
FreeXid(xid);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
if (file->f_path.dentry == NULL) {
|
2007-04-04 11:10:24 -06:00
|
|
|
cERROR(1, ("no valid name if dentry freed"));
|
|
|
|
dump_stack();
|
|
|
|
rc = -EBADF;
|
|
|
|
goto reopen_error_exit;
|
|
|
|
}
|
|
|
|
|
|
|
|
inode = file->f_path.dentry->d_inode;
|
2007-07-09 19:16:18 -06:00
|
|
|
if (inode == NULL) {
|
2007-04-04 11:10:24 -06:00
|
|
|
cERROR(1, ("inode not valid"));
|
|
|
|
dump_stack();
|
|
|
|
rc = -EBADF;
|
|
|
|
goto reopen_error_exit;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
cifs_sb = CIFS_SB(inode->i_sb);
|
|
|
|
pTcon = cifs_sb->tcon;
|
2007-04-04 11:10:24 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/* 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 */
|
2006-12-08 03:36:48 -07:00
|
|
|
full_path = build_path_from_dentry(file->f_path.dentry);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (full_path == NULL) {
|
2007-04-04 11:10:24 -06:00
|
|
|
rc = -ENOMEM;
|
|
|
|
reopen_error_exit:
|
2005-04-16 16:20:36 -06:00
|
|
|
up(&pCifsFile->fh_sem);
|
|
|
|
FreeXid(xid);
|
2007-04-04 11:10:24 -06:00
|
|
|
return rc;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2007-04-04 11:10:24 -06:00
|
|
|
cFYI(1, ("inode = 0x%p file flags 0x%x for %s",
|
2007-07-09 19:16:18 -06:00
|
|
|
inode, file->f_flags, full_path));
|
2005-04-16 16:20:36 -06:00
|
|
|
desiredAccess = cifs_convert_flags(file->f_flags);
|
|
|
|
|
|
|
|
if (oplockEnabled)
|
|
|
|
oplock = REQ_OPLOCK;
|
|
|
|
else
|
2008-04-28 18:06:05 -06:00
|
|
|
oplock = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* Can not refresh inode by passing in file_info buf to be returned
|
2007-07-09 19:16:18 -06:00
|
|
|
by SMBOpen and then calling get_inode_info with returned buf
|
|
|
|
since file might have write behind data that needs to be flushed
|
2005-04-16 16:20:36 -06:00
|
|
|
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,
|
2007-07-09 19:16:18 -06:00
|
|
|
cifs_sb->local_nls, cifs_sb->mnt_cifs_flags &
|
2005-04-28 23:41:06 -06:00
|
|
|
CIFS_MOUNT_MAP_SPECIAL_CHR);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc) {
|
|
|
|
up(&pCifsFile->fh_sem);
|
2006-05-31 12:05:34 -06:00
|
|
|
cFYI(1, ("cifs_open returned 0x%x", rc));
|
|
|
|
cFYI(1, ("oplock: %d", oplock));
|
2005-04-16 16:20:36 -06:00
|
|
|
} else {
|
|
|
|
pCifsFile->netfid = netfid;
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsFile->invalidHandle = false;
|
2005-04-16 16:20:36 -06:00
|
|
|
up(&pCifsFile->fh_sem);
|
|
|
|
pCifsInode = CIFS_I(inode);
|
|
|
|
if (pCifsInode) {
|
|
|
|
if (can_flush) {
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
rc = filemap_write_and_wait(inode->i_mapping);
|
|
|
|
if (rc != 0)
|
|
|
|
CIFS_I(inode)->write_behind_rc = rc;
|
2005-04-16 16:20:36 -06:00
|
|
|
/* temporarily disable caching while we
|
|
|
|
go to server to get inode info */
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsInode->clientCanCacheAll = false;
|
|
|
|
pCifsInode->clientCanCacheRead = false;
|
2007-07-18 17:21:09 -06:00
|
|
|
if (pTcon->unix_ext)
|
2005-04-16 16:20:36 -06:00
|
|
|
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,
|
2008-03-14 16:37:16 -06:00
|
|
|
xid, NULL);
|
2005-04-16 16:20:36 -06:00
|
|
|
} /* 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) {
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsInode->clientCanCacheAll = true;
|
|
|
|
pCifsInode->clientCanCacheRead = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
cFYI(1, ("Exclusive Oplock granted on inode %p",
|
2006-12-08 03:36:48 -07:00
|
|
|
file->f_path.dentry->d_inode));
|
2005-04-16 16:20:36 -06:00
|
|
|
} else if ((oplock & 0xF) == OPLOCK_READ) {
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsInode->clientCanCacheRead = true;
|
|
|
|
pCifsInode->clientCanCacheAll = false;
|
2005-04-16 16:20:36 -06:00
|
|
|
} else {
|
2008-04-28 18:06:05 -06:00
|
|
|
pCifsInode->clientCanCacheRead = false;
|
|
|
|
pCifsInode->clientCanCacheAll = false;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
cifs_relock_file(pCifsFile);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
kfree(full_path);
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
int cifs_close(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
int rc = 0;
|
2007-09-07 16:23:48 -06:00
|
|
|
int xid, timeout;
|
2005-04-16 16:20:36 -06:00
|
|
|
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) {
|
2006-08-02 15:56:33 -06:00
|
|
|
struct cifsLockInfo *li, *tmp;
|
|
|
|
|
2008-04-28 18:06:05 -06:00
|
|
|
pSMBFile->closePend = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
if (pTcon) {
|
|
|
|
/* no sense reconnecting to close a file that is
|
|
|
|
already closed */
|
|
|
|
if (pTcon->tidStatus != CifsNeedReconnect) {
|
2007-09-07 16:23:48 -06:00
|
|
|
timeout = 2;
|
2007-07-09 19:16:18 -06:00
|
|
|
while ((atomic_read(&pSMBFile->wrtPending) != 0)
|
2007-09-07 16:23:48 -06:00
|
|
|
&& (timeout <= 2048)) {
|
2005-10-20 14:44:56 -06:00
|
|
|
/* 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,
|
2007-07-09 19:16:18 -06:00
|
|
|
but this should give enough time to
|
2005-10-20 14:44:56 -06:00
|
|
|
clear the socket */
|
2008-02-12 13:32:36 -07:00
|
|
|
cFYI(DBG2,
|
|
|
|
("close delay, write pending"));
|
2005-10-20 14:44:56 -06:00
|
|
|
msleep(timeout);
|
|
|
|
timeout *= 4;
|
2006-11-07 09:31:16 -07:00
|
|
|
}
|
2007-07-09 19:16:18 -06:00
|
|
|
if (atomic_read(&pSMBFile->wrtPending))
|
2007-07-17 11:34:02 -06:00
|
|
|
cERROR(1,
|
|
|
|
("close with pending writes"));
|
2005-04-16 16:20:36 -06:00
|
|
|
rc = CIFSSMBClose(xid, pTcon,
|
|
|
|
pSMBFile->netfid);
|
|
|
|
}
|
|
|
|
}
|
2006-08-02 15:56:33 -06:00
|
|
|
|
|
|
|
/* Delete any outstanding lock records.
|
|
|
|
We'll lose them when the file is closed anyway. */
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_lock(&pSMBFile->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
list_for_each_entry_safe(li, tmp, &pSMBFile->llist, llist) {
|
|
|
|
list_del(&li->llist);
|
|
|
|
kfree(li);
|
|
|
|
}
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_unlock(&pSMBFile->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
|
2005-04-28 23:41:05 -06:00
|
|
|
write_lock(&GlobalSMBSeslock);
|
2005-04-16 16:20:36 -06:00
|
|
|
list_del(&pSMBFile->flist);
|
|
|
|
list_del(&pSMBFile->tlist);
|
2005-04-28 23:41:05 -06:00
|
|
|
write_unlock(&GlobalSMBSeslock);
|
2007-09-07 16:23:48 -06:00
|
|
|
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;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
kfree(file->private_data);
|
|
|
|
file->private_data = NULL;
|
|
|
|
} else
|
|
|
|
rc = -EBADF;
|
|
|
|
|
2007-09-10 23:50:53 -06:00
|
|
|
read_lock(&GlobalSMBSeslock);
|
2005-04-16 16:20:36 -06:00
|
|
|
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 */
|
2008-04-28 18:06:05 -06:00
|
|
|
CIFS_I(inode)->clientCanCacheRead = false;
|
|
|
|
CIFS_I(inode)->clientCanCacheAll = false;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2007-09-10 23:50:53 -06:00
|
|
|
read_unlock(&GlobalSMBSeslock);
|
2007-07-09 19:16:18 -06:00
|
|
|
if ((rc == 0) && CIFS_I(inode)->write_behind_rc)
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
|
|
|
|
2006-05-31 12:05:34 -06:00
|
|
|
cFYI(1, ("Closedir inode = 0x%p", inode));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
xid = GetXid();
|
|
|
|
|
|
|
|
if (pCFileStruct) {
|
|
|
|
struct cifsTconInfo *pTcon;
|
2007-07-09 19:16:18 -06:00
|
|
|
struct cifs_sb_info *cifs_sb =
|
|
|
|
CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
pTcon = cifs_sb->tcon;
|
|
|
|
|
|
|
|
cFYI(1, ("Freeing private data in close dir"));
|
2008-04-28 18:06:05 -06:00
|
|
|
if (!pCFileStruct->srch_inf.endOfSearch &&
|
|
|
|
!pCFileStruct->invalidHandle) {
|
|
|
|
pCFileStruct->invalidHandle = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
ptmp = pCFileStruct->srch_inf.ntwrk_buf_start;
|
|
|
|
if (ptmp) {
|
2005-12-12 21:53:18 -07:00
|
|
|
cFYI(1, ("closedir free smb buf in srch struct"));
|
2005-04-16 16:20:36 -06:00
|
|
|
pCFileStruct->srch_inf.ntwrk_buf_start = NULL;
|
2007-07-09 19:16:18 -06:00
|
|
|
if (pCFileStruct->srch_inf.smallBuf)
|
2006-02-27 20:45:48 -07:00
|
|
|
cifs_small_buf_release(ptmp);
|
|
|
|
else
|
|
|
|
cifs_buf_release(ptmp);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
kfree(file->private_data);
|
|
|
|
file->private_data = NULL;
|
|
|
|
}
|
|
|
|
/* BB can we lock the filestruct while this is going on? */
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2006-08-02 15:56:33 -06:00
|
|
|
static int store_file_lock(struct cifsFileInfo *fid, __u64 len,
|
|
|
|
__u64 offset, __u8 lockType)
|
|
|
|
{
|
2007-07-09 19:16:18 -06:00
|
|
|
struct cifsLockInfo *li =
|
|
|
|
kmalloc(sizeof(struct cifsLockInfo), GFP_KERNEL);
|
2006-08-02 15:56:33 -06:00
|
|
|
if (li == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
li->offset = offset;
|
|
|
|
li->length = len;
|
|
|
|
li->type = lockType;
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_lock(&fid->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
list_add(&li->llist, &fid->llist);
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_unlock(&fid->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
int cifs_lock(struct file *file, int cmd, struct file_lock *pfLock)
|
|
|
|
{
|
|
|
|
int rc, xid;
|
|
|
|
__u32 numLock = 0;
|
|
|
|
__u32 numUnlock = 0;
|
|
|
|
__u64 length;
|
2008-04-28 18:06:05 -06:00
|
|
|
bool wait_flag = false;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct cifs_sb_info *cifs_sb;
|
|
|
|
struct cifsTconInfo *pTcon;
|
2006-02-28 15:39:25 -07:00
|
|
|
__u16 netfid;
|
|
|
|
__u8 lockType = LOCKING_ANDX_LARGE_FILES;
|
2008-04-28 18:06:05 -06:00
|
|
|
bool posix_locking;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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",
|
2007-07-09 19:16:18 -06:00
|
|
|
cmd, pfLock->fl_flags, pfLock->fl_type, pfLock->fl_start,
|
|
|
|
pfLock->fl_end));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (pfLock->fl_flags & FL_POSIX)
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("Posix"));
|
2005-04-16 16:20:36 -06:00
|
|
|
if (pfLock->fl_flags & FL_FLOCK)
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("Flock"));
|
2005-04-16 16:20:36 -06:00
|
|
|
if (pfLock->fl_flags & FL_SLEEP) {
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("Blocking lock"));
|
2008-04-28 18:06:05 -06:00
|
|
|
wait_flag = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
if (pfLock->fl_flags & FL_ACCESS)
|
|
|
|
cFYI(1, ("Process suspended by mandatory locking - "
|
2006-05-31 12:05:34 -06:00
|
|
|
"not implemented yet"));
|
2005-04-16 16:20:36 -06:00
|
|
|
if (pfLock->fl_flags & FL_LEASE)
|
|
|
|
cFYI(1, ("Lease on file - not implemented yet"));
|
2007-07-09 19:16:18 -06:00
|
|
|
if (pfLock->fl_flags &
|
2005-04-16 16:20:36 -06:00
|
|
|
(~(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) {
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("F_UNLCK"));
|
2005-04-16 16:20:36 -06:00
|
|
|
numUnlock = 1;
|
2006-02-27 20:45:48 -07:00
|
|
|
/* Check if unlock includes more than
|
|
|
|
one lock range */
|
2005-04-16 16:20:36 -06:00
|
|
|
} else if (pfLock->fl_type == F_RDLCK) {
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("F_RDLCK"));
|
2005-04-16 16:20:36 -06:00
|
|
|
lockType |= LOCKING_ANDX_SHARED_LOCK;
|
|
|
|
numLock = 1;
|
|
|
|
} else if (pfLock->fl_type == F_EXLCK) {
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("F_EXLCK"));
|
2005-04-16 16:20:36 -06:00
|
|
|
numLock = 1;
|
|
|
|
} else if (pfLock->fl_type == F_SHLCK) {
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("F_SHLCK"));
|
2005-04-16 16:20:36 -06:00
|
|
|
lockType |= LOCKING_ANDX_SHARED_LOCK;
|
|
|
|
numLock = 1;
|
|
|
|
} else
|
2006-02-27 20:45:48 -07:00
|
|
|
cFYI(1, ("Unknown type of lock"));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
pTcon = cifs_sb->tcon;
|
|
|
|
|
|
|
|
if (file->private_data == NULL) {
|
|
|
|
FreeXid(xid);
|
|
|
|
return -EBADF;
|
|
|
|
}
|
2006-02-28 15:39:25 -07:00
|
|
|
netfid = ((struct cifsFileInfo *)file->private_data)->netfid;
|
|
|
|
|
2006-08-02 15:56:33 -06:00
|
|
|
posix_locking = (cifs_sb->tcon->ses->capabilities & CAP_UNIX) &&
|
|
|
|
(CIFS_UNIX_FCNTL_CAP & le64_to_cpu(cifs_sb->tcon->fsUnixInfo.Capability));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-02-28 15:39:25 -07:00
|
|
|
/* BB add code here to normalize offset and length to
|
|
|
|
account for negative length which we can not accept over the
|
|
|
|
wire */
|
2005-04-16 16:20:36 -06:00
|
|
|
if (IS_GETLK(cmd)) {
|
2007-07-09 19:16:18 -06:00
|
|
|
if (posix_locking) {
|
2006-02-28 15:39:25 -07:00
|
|
|
int posix_lock_type;
|
2007-07-09 19:16:18 -06:00
|
|
|
if (lockType & LOCKING_ANDX_SHARED_LOCK)
|
2006-02-28 15:39:25 -07:00
|
|
|
posix_lock_type = CIFS_RDLCK;
|
|
|
|
else
|
|
|
|
posix_lock_type = CIFS_WRLCK;
|
|
|
|
rc = CIFSSMBPosixLock(xid, pTcon, netfid, 1 /* get */,
|
2006-05-30 12:03:32 -06:00
|
|
|
length, pfLock,
|
2006-02-28 15:39:25 -07:00
|
|
|
posix_lock_type, wait_flag);
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* BB we could chain these into one lock request BB */
|
|
|
|
rc = CIFSSMBLock(xid, pTcon, netfid, length, pfLock->fl_start,
|
|
|
|
0, 1, lockType, 0 /* wait flag */ );
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc == 0) {
|
2007-07-09 19:16:18 -06:00
|
|
|
rc = CIFSSMBLock(xid, pTcon, netfid, length,
|
2005-04-16 16:20:36 -06:00
|
|
|
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 "
|
2006-02-28 15:39:25 -07:00
|
|
|
"range %d during test of lock", rc));
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
|
|
|
}
|
2006-08-02 15:56:33 -06:00
|
|
|
|
|
|
|
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) {
|
2006-02-28 15:39:25 -07:00
|
|
|
int posix_lock_type;
|
2007-07-09 19:16:18 -06:00
|
|
|
if (lockType & LOCKING_ANDX_SHARED_LOCK)
|
2006-02-28 15:39:25 -07:00
|
|
|
posix_lock_type = CIFS_RDLCK;
|
|
|
|
else
|
|
|
|
posix_lock_type = CIFS_WRLCK;
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
if (numUnlock == 1)
|
2006-03-03 16:36:34 -07:00
|
|
|
posix_lock_type = CIFS_UNLCK;
|
2006-08-02 15:56:33 -06:00
|
|
|
|
2006-02-28 15:39:25 -07:00
|
|
|
rc = CIFSSMBPosixLock(xid, pTcon, netfid, 0 /* set */,
|
2006-05-30 12:03:32 -06:00
|
|
|
length, pfLock,
|
2006-02-28 15:39:25 -07:00
|
|
|
posix_lock_type, wait_flag);
|
2006-08-02 15:56:33 -06:00
|
|
|
} else {
|
2007-07-09 19:16:18 -06:00
|
|
|
struct cifsFileInfo *fid =
|
|
|
|
(struct cifsFileInfo *)file->private_data;
|
2006-08-02 15:56:33 -06:00
|
|
|
|
|
|
|
if (numLock) {
|
2007-07-09 19:16:18 -06:00
|
|
|
rc = CIFSSMBLock(xid, pTcon, netfid, length,
|
|
|
|
pfLock->fl_start,
|
2006-08-02 15:56:33 -06:00
|
|
|
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;
|
|
|
|
|
2006-09-21 01:35:29 -06:00
|
|
|
rc = 0;
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_lock(&fid->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
list_for_each_entry_safe(li, tmp, &fid->llist, llist) {
|
|
|
|
if (pfLock->fl_start <= li->offset &&
|
2007-08-23 21:22:48 -06:00
|
|
|
(pfLock->fl_start + length) >=
|
2007-08-23 21:16:51 -06:00
|
|
|
(li->offset + li->length)) {
|
2007-07-09 19:16:18 -06:00
|
|
|
stored_rc = CIFSSMBLock(xid, pTcon,
|
|
|
|
netfid,
|
2006-08-02 15:56:33 -06:00
|
|
|
li->length, li->offset,
|
2008-04-28 18:06:05 -06:00
|
|
|
1, 0, li->type, false);
|
2006-08-02 15:56:33 -06:00
|
|
|
if (stored_rc)
|
|
|
|
rc = stored_rc;
|
|
|
|
|
|
|
|
list_del(&li->llist);
|
|
|
|
kfree(li);
|
|
|
|
}
|
|
|
|
}
|
2007-05-02 22:33:45 -06:00
|
|
|
mutex_unlock(&fid->lock_mutex);
|
2006-08-02 15:56:33 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-08-26 13:42:59 -06:00
|
|
|
if (pfLock->fl_flags & FL_POSIX)
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
pTcon = cifs_sb->tcon;
|
|
|
|
|
|
|
|
/* cFYI(1,
|
|
|
|
(" write %d bytes to offset %lld of %s", write_size,
|
2006-12-08 03:36:48 -07:00
|
|
|
*poffset, file->f_path.dentry->d_name.name)); */
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (file->private_data == NULL)
|
|
|
|
return -EBADF;
|
2007-04-02 12:47:20 -06:00
|
|
|
open_file = (struct cifsFileInfo *) file->private_data;
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2008-08-28 05:54:59 -06:00
|
|
|
rc = generic_write_checks(file, poffset, &write_size, 0);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
xid = GetXid();
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
if (*poffset > file->f_path.dentry->d_inode->i_size)
|
2007-11-13 15:41:37 -07:00
|
|
|
long_op = CIFS_VLONG_OP; /* writes past EOF take long time */
|
2005-04-16 16:20:36 -06:00
|
|
|
else
|
2007-11-13 15:41:37 -07:00
|
|
|
long_op = CIFS_LONG_OP;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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;
|
2007-07-09 19:16:18 -06:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
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 */
|
2008-04-28 18:06:05 -06:00
|
|
|
rc = cifs_reopen_file(file, false);
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
2007-11-13 15:41:37 -07:00
|
|
|
long_op = CIFS_STD_OP; /* subsequent writes fast -
|
2005-04-16 16:20:36 -06:00
|
|
|
15 seconds is plenty */
|
|
|
|
}
|
|
|
|
|
2005-08-24 14:59:35 -06:00
|
|
|
cifs_stats_bytes_written(pTcon, total_written);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* since the write may have blocked check these pointers again */
|
2007-02-26 09:46:11 -07:00
|
|
|
if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
|
|
|
|
struct inode *inode = file->f_path.dentry->d_inode;
|
2007-07-09 19:16:18 -06:00
|
|
|
/* Do not update local mtime - server will set its actual value on write
|
|
|
|
* inode->i_ctime = inode->i_mtime =
|
2007-02-26 09:46:11 -07:00
|
|
|
* 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,
|
2005-04-16 16:20:36 -06:00
|
|
|
*poffset);
|
2007-02-26 09:46:11 -07:00
|
|
|
spin_unlock(&inode->i_lock);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2007-07-09 19:16:18 -06:00
|
|
|
mark_inode_dirty_sync(file->f_path.dentry->d_inode);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
FreeXid(xid);
|
|
|
|
return total_written;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t cifs_write(struct file *file, const char *write_data,
|
2008-09-24 09:32:59 -06:00
|
|
|
size_t write_size, loff_t *poffset)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
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;
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
pTcon = cifs_sb->tcon;
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
cFYI(1, ("write %zd bytes to offset %lld of %s", write_size,
|
2006-12-08 03:36:48 -07:00
|
|
|
*poffset, file->f_path.dentry->d_name.name));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (file->private_data == NULL)
|
|
|
|
return -EBADF;
|
2007-04-02 12:47:20 -06:00
|
|
|
open_file = (struct cifsFileInfo *)file->private_data;
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
xid = GetXid();
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
if (*poffset > file->f_path.dentry->d_inode->i_size)
|
2007-11-13 15:41:37 -07:00
|
|
|
long_op = CIFS_VLONG_OP; /* writes past EOF can be slow */
|
2005-04-16 16:20:36 -06:00
|
|
|
else
|
2007-11-13 15:41:37 -07:00
|
|
|
long_op = CIFS_LONG_OP;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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
|
2007-07-09 19:16:18 -06:00
|
|
|
while we blocked so return what we managed to
|
2005-04-16 16:20:36 -06:00
|
|
|
write */
|
|
|
|
return total_written;
|
2007-07-09 19:16:18 -06:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
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
|
2007-07-09 19:16:18 -06:00
|
|
|
reopen_file not to flush data to
|
2005-04-16 16:20:36 -06:00
|
|
|
server now */
|
2008-04-28 18:06:05 -06:00
|
|
|
rc = cifs_reopen_file(file, false);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc != 0)
|
|
|
|
break;
|
|
|
|
}
|
2007-07-09 19:16:18 -06:00
|
|
|
if (experimEnabled || (pTcon->ses->server &&
|
|
|
|
((pTcon->ses->server->secMode &
|
2006-05-30 12:08:26 -06:00
|
|
|
(SECMODE_SIGN_REQUIRED | SECMODE_SIGN_ENABLED))
|
2006-05-30 12:05:10 -06:00
|
|
|
== 0))) {
|
2005-10-03 14:37:24 -06:00
|
|
|
struct kvec iov[2];
|
|
|
|
unsigned int len;
|
|
|
|
|
2005-10-10 11:57:19 -06:00
|
|
|
len = min((size_t)cifs_sb->wsize,
|
2005-10-03 14:37:24 -06:00
|
|
|
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;
|
2005-06-13 12:24:43 -06:00
|
|
|
rc = CIFSSMBWrite2(xid, pTcon,
|
2005-10-03 14:37:24 -06:00
|
|
|
open_file->netfid, len,
|
2005-06-13 12:24:43 -06:00
|
|
|
*poffset, &bytes_written,
|
2005-10-03 14:37:24 -06:00
|
|
|
iov, 1, long_op);
|
2005-06-13 12:24:43 -06:00
|
|
|
} else
|
2006-04-22 09:53:05 -06:00
|
|
|
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);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
if (rc || (bytes_written == 0)) {
|
|
|
|
if (total_written)
|
|
|
|
break;
|
|
|
|
else {
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
} else
|
|
|
|
*poffset += bytes_written;
|
2007-11-13 15:41:37 -07:00
|
|
|
long_op = CIFS_STD_OP; /* subsequent writes fast -
|
2005-04-16 16:20:36 -06:00
|
|
|
15 seconds is plenty */
|
|
|
|
}
|
|
|
|
|
2005-08-24 14:59:35 -06:00
|
|
|
cifs_stats_bytes_written(pTcon, total_written);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* since the write may have blocked check these pointers again */
|
2007-02-26 09:46:11 -07:00
|
|
|
if ((file->f_path.dentry) && (file->f_path.dentry->d_inode)) {
|
2007-02-16 21:34:13 -07:00
|
|
|
/*BB We could make this contingent on superblock ATIME flag too */
|
2007-02-26 09:46:11 -07:00
|
|
|
/* 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);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2007-02-26 09:46:11 -07:00
|
|
|
mark_inode_dirty_sync(file->f_path.dentry->d_inode);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
FreeXid(xid);
|
|
|
|
return total_written;
|
|
|
|
}
|
|
|
|
|
2007-10-25 15:17:17 -06:00
|
|
|
#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
|
|
|
|
|
2005-10-05 20:32:49 -06:00
|
|
|
struct cifsFileInfo *find_writable_file(struct cifsInodeInfo *cifs_inode)
|
2005-10-05 13:23:19 -06:00
|
|
|
{
|
|
|
|
struct cifsFileInfo *open_file;
|
2008-09-22 19:33:33 -06:00
|
|
|
bool any_available = false;
|
2005-10-05 20:32:49 -06:00
|
|
|
int rc;
|
2005-10-05 13:23:19 -06:00
|
|
|
|
2006-04-22 09:53:05 -06:00
|
|
|
/* 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 */
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
if (cifs_inode == NULL) {
|
|
|
|
cERROR(1, ("Null inode passed to cifs_writeable_file"));
|
2006-04-22 09:53:05 -06:00
|
|
|
dump_stack();
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-10-05 13:23:19 -06:00
|
|
|
read_lock(&GlobalSMBSeslock);
|
2007-10-01 19:11:08 -06:00
|
|
|
refind_writable:
|
2005-10-05 13:23:19 -06:00
|
|
|
list_for_each_entry(open_file, &cifs_inode->openFileList, flist) {
|
2008-09-22 19:33:33 -06:00
|
|
|
if (open_file->closePend ||
|
|
|
|
(!any_available && open_file->pid != current->tgid))
|
2005-10-05 13:23:19 -06:00
|
|
|
continue;
|
2008-09-22 19:33:33 -06:00
|
|
|
|
2005-10-05 13:23:19 -06:00
|
|
|
if (open_file->pfile &&
|
|
|
|
((open_file->pfile->f_flags & O_RDWR) ||
|
|
|
|
(open_file->pfile->f_flags & O_WRONLY))) {
|
2005-10-20 14:44:56 -06:00
|
|
|
atomic_inc(&open_file->wrtPending);
|
2007-10-01 19:11:08 -06:00
|
|
|
|
|
|
|
if (!open_file->invalidHandle) {
|
|
|
|
/* found a good writable file */
|
|
|
|
read_unlock(&GlobalSMBSeslock);
|
|
|
|
return open_file;
|
|
|
|
}
|
2007-11-16 16:05:52 -07:00
|
|
|
|
2005-10-05 13:23:19 -06:00
|
|
|
read_unlock(&GlobalSMBSeslock);
|
2007-10-01 19:11:08 -06:00
|
|
|
/* Had to unlock since following call can block */
|
2008-04-28 18:06:05 -06:00
|
|
|
rc = cifs_reopen_file(open_file->pfile, false);
|
2007-11-16 16:05:52 -07:00
|
|
|
if (!rc) {
|
2007-10-01 19:11:08 -06:00
|
|
|
if (!open_file->closePend)
|
|
|
|
return open_file;
|
|
|
|
else { /* start over in case this was deleted */
|
|
|
|
/* since the list could be modified */
|
2005-10-05 15:50:29 -06:00
|
|
|
read_lock(&GlobalSMBSeslock);
|
2007-09-07 16:23:48 -06:00
|
|
|
atomic_dec(&open_file->wrtPending);
|
2007-10-01 19:11:08 -06:00
|
|
|
goto refind_writable;
|
2005-10-05 15:50:29 -06:00
|
|
|
}
|
|
|
|
}
|
2007-10-01 19:11:08 -06:00
|
|
|
|
|
|
|
/* 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);
|
2007-11-16 16:05:52 -07:00
|
|
|
|
2007-10-01 19:11:08 -06:00
|
|
|
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. */
|
2005-10-05 13:23:19 -06:00
|
|
|
}
|
|
|
|
}
|
2008-09-22 19:33:33 -06:00
|
|
|
/* couldn't find useable FH with same pid, try any available */
|
|
|
|
if (!any_available) {
|
|
|
|
any_available = true;
|
|
|
|
goto refind_writable;
|
|
|
|
}
|
2005-10-05 13:23:19 -06:00
|
|
|
read_unlock(&GlobalSMBSeslock);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
2005-10-05 13:23:19 -06:00
|
|
|
struct cifsFileInfo *open_file;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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)
|
2007-07-09 19:16:18 -06:00
|
|
|
to = (unsigned)(mapping->host->i_size - offset);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2005-10-05 13:23:19 -06:00
|
|
|
open_file = find_writable_file(CIFS_I(mapping->host));
|
|
|
|
if (open_file) {
|
|
|
|
bytes_written = cifs_write(open_file->pfile, write_data,
|
|
|
|
to-from, &offset);
|
2005-10-20 14:44:56 -06:00
|
|
|
atomic_dec(&open_file->wrtPending);
|
2005-04-16 16:20:36 -06:00
|
|
|
/* Does mm or vfs already set times? */
|
2005-10-05 13:23:19 -06:00
|
|
|
inode->i_atime = inode->i_mtime = current_fs_time(inode->i_sb);
|
2007-12-30 21:21:29 -07:00
|
|
|
if ((bytes_written > 0) && (offset))
|
2005-10-05 13:23:19 -06:00
|
|
|
rc = 0;
|
2007-12-30 21:21:29 -07:00
|
|
|
else if (bytes_written < 0)
|
|
|
|
rc = bytes_written;
|
2005-10-05 13:23:19 -06:00
|
|
|
} else {
|
2005-04-16 16:20:36 -06:00
|
|
|
cFYI(1, ("No writeable filehandles for inode"));
|
|
|
|
rc = -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
kunmap(page);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cifs_writepages(struct address_space *mapping,
|
2005-10-05 15:50:29 -06:00
|
|
|
struct writeback_control *wbc)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2005-10-05 15:50:29 -06:00
|
|
|
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;
|
[PATCH] writeback: fix range handling
When a writeback_control's `start' and `end' fields are used to
indicate a one-byte-range starting at file offset zero, the required
values of .start=0,.end=0 mean that the ->writepages() implementation
has no way of telling that it is being asked to perform a range
request. Because we're currently overloading (start == 0 && end == 0)
to mean "this is not a write-a-range request".
To make all this sane, the patch changes range of writeback_control.
So caller does: If it is calling ->writepages() to write pages, it
sets range (range_start/end or range_cyclic) always.
And if range_cyclic is true, ->writepages() thinks the range is
cyclic, otherwise it just uses range_start and range_end.
This patch does,
- Add LLONG_MAX, LLONG_MIN, ULLONG_MAX to include/linux/kernel.h
-1 is usually ok for range_end (type is long long). But, if someone did,
range_end += val; range_end is "val - 1"
u64val = range_end >> bits; u64val is "~(0ULL)"
or something, they are wrong. So, this adds LLONG_MAX to avoid nasty
things, and uses LLONG_MAX for range_end.
- All callers of ->writepages() sets range_start/end or range_cyclic.
- Fix updates of ->writeback_index. It seems already bit strange.
If it starts at 0 and ended by check of nr_to_write, this last
index may reduce chance to scan end of file. So, this updates
->writeback_index only if range_cyclic is true or whole-file is
scanned.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: "Vladimir V. Saveliev" <vs@namesys.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 03:03:26 -06:00
|
|
|
pgoff_t end;
|
2005-10-05 15:50:29 -06:00
|
|
|
pgoff_t index;
|
2007-07-09 19:16:18 -06:00
|
|
|
int range_whole = 0;
|
|
|
|
struct kvec *iov;
|
2005-10-12 16:32:05 -06:00
|
|
|
int len;
|
2005-10-05 15:50:29 -06:00
|
|
|
int n_iov = 0;
|
|
|
|
pgoff_t next;
|
|
|
|
int nr_pages;
|
|
|
|
__u64 offset = 0;
|
2005-10-20 14:44:56 -06:00
|
|
|
struct cifsFileInfo *open_file;
|
2005-10-05 15:50:29 -06:00
|
|
|
struct page *page;
|
|
|
|
struct pagevec pvec;
|
|
|
|
int rc = 0;
|
|
|
|
int scanned = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
int xid;
|
|
|
|
|
2005-10-05 15:50:29 -06:00
|
|
|
cifs_sb = CIFS_SB(mapping->host->i_sb);
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2005-10-05 15:50:29 -06:00
|
|
|
/*
|
|
|
|
* 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);
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
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)
|
2006-04-22 09:53:05 -06:00
|
|
|
return generic_writepages(mapping, wbc);
|
2005-10-05 16:14:33 -06:00
|
|
|
|
2007-02-01 21:21:57 -07:00
|
|
|
iov = kmalloc(32 * sizeof(struct kvec), GFP_KERNEL);
|
2007-07-09 19:16:18 -06:00
|
|
|
if (iov == NULL)
|
2007-02-01 21:21:57 -07:00
|
|
|
return generic_writepages(mapping, wbc);
|
|
|
|
|
|
|
|
|
2005-10-05 15:50:29 -06:00
|
|
|
/*
|
|
|
|
* 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;
|
2007-02-01 21:21:57 -07:00
|
|
|
kfree(iov);
|
2005-10-05 15:50:29 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
xid = GetXid();
|
|
|
|
|
2005-10-05 15:50:29 -06:00
|
|
|
pagevec_init(&pvec, 0);
|
[PATCH] writeback: fix range handling
When a writeback_control's `start' and `end' fields are used to
indicate a one-byte-range starting at file offset zero, the required
values of .start=0,.end=0 mean that the ->writepages() implementation
has no way of telling that it is being asked to perform a range
request. Because we're currently overloading (start == 0 && end == 0)
to mean "this is not a write-a-range request".
To make all this sane, the patch changes range of writeback_control.
So caller does: If it is calling ->writepages() to write pages, it
sets range (range_start/end or range_cyclic) always.
And if range_cyclic is true, ->writepages() thinks the range is
cyclic, otherwise it just uses range_start and range_end.
This patch does,
- Add LLONG_MAX, LLONG_MIN, ULLONG_MAX to include/linux/kernel.h
-1 is usually ok for range_end (type is long long). But, if someone did,
range_end += val; range_end is "val - 1"
u64val = range_end >> bits; u64val is "~(0ULL)"
or something, they are wrong. So, this adds LLONG_MAX to avoid nasty
things, and uses LLONG_MAX for range_end.
- All callers of ->writepages() sets range_start/end or range_cyclic.
- Fix updates of ->writeback_index. It seems already bit strange.
If it starts at 0 and ended by check of nr_to_write, this last
index may reduce chance to scan end of file. So, this updates
->writeback_index only if range_cyclic is true or whole-file is
scanned.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: "Vladimir V. Saveliev" <vs@namesys.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 03:03:26 -06:00
|
|
|
if (wbc->range_cyclic) {
|
2005-10-05 15:50:29 -06:00
|
|
|
index = mapping->writeback_index; /* Start from prev offset */
|
[PATCH] writeback: fix range handling
When a writeback_control's `start' and `end' fields are used to
indicate a one-byte-range starting at file offset zero, the required
values of .start=0,.end=0 mean that the ->writepages() implementation
has no way of telling that it is being asked to perform a range
request. Because we're currently overloading (start == 0 && end == 0)
to mean "this is not a write-a-range request".
To make all this sane, the patch changes range of writeback_control.
So caller does: If it is calling ->writepages() to write pages, it
sets range (range_start/end or range_cyclic) always.
And if range_cyclic is true, ->writepages() thinks the range is
cyclic, otherwise it just uses range_start and range_end.
This patch does,
- Add LLONG_MAX, LLONG_MIN, ULLONG_MAX to include/linux/kernel.h
-1 is usually ok for range_end (type is long long). But, if someone did,
range_end += val; range_end is "val - 1"
u64val = range_end >> bits; u64val is "~(0ULL)"
or something, they are wrong. So, this adds LLONG_MAX to avoid nasty
things, and uses LLONG_MAX for range_end.
- All callers of ->writepages() sets range_start/end or range_cyclic.
- Fix updates of ->writeback_index. It seems already bit strange.
If it starts at 0 and ended by check of nr_to_write, this last
index may reduce chance to scan end of file. So, this updates
->writeback_index only if range_cyclic is true or whole-file is
scanned.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: "Vladimir V. Saveliev" <vs@namesys.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 03:03:26 -06:00
|
|
|
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;
|
2005-10-05 15:50:29 -06:00
|
|
|
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);
|
2008-08-02 04:01:03 -06:00
|
|
|
else if (!trylock_page(page))
|
2005-10-05 15:50:29 -06:00
|
|
|
break;
|
|
|
|
|
|
|
|
if (unlikely(page->mapping != mapping)) {
|
|
|
|
unlock_page(page);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
[PATCH] writeback: fix range handling
When a writeback_control's `start' and `end' fields are used to
indicate a one-byte-range starting at file offset zero, the required
values of .start=0,.end=0 mean that the ->writepages() implementation
has no way of telling that it is being asked to perform a range
request. Because we're currently overloading (start == 0 && end == 0)
to mean "this is not a write-a-range request".
To make all this sane, the patch changes range of writeback_control.
So caller does: If it is calling ->writepages() to write pages, it
sets range (range_start/end or range_cyclic) always.
And if range_cyclic is true, ->writepages() thinks the range is
cyclic, otherwise it just uses range_start and range_end.
This patch does,
- Add LLONG_MAX, LLONG_MIN, ULLONG_MAX to include/linux/kernel.h
-1 is usually ok for range_end (type is long long). But, if someone did,
range_end += val; range_end is "val - 1"
u64val = range_end >> bits; u64val is "~(0ULL)"
or something, they are wrong. So, this adds LLONG_MAX to avoid nasty
things, and uses LLONG_MAX for range_end.
- All callers of ->writepages() sets range_start/end or range_cyclic.
- Fix updates of ->writeback_index. It seems already bit strange.
If it starts at 0 and ended by check of nr_to_write, this last
index may reduce chance to scan end of file. So, this updates
->writeback_index only if range_cyclic is true or whole-file is
scanned.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: "Vladimir V. Saveliev" <vs@namesys.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 03:03:26 -06:00
|
|
|
if (!wbc->range_cyclic && page->index > end) {
|
2005-10-05 15:50:29 -06:00
|
|
|
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) ||
|
2006-12-23 17:19:07 -07:00
|
|
|
!clear_page_dirty_for_io(page)) {
|
2005-10-05 15:50:29 -06:00
|
|
|
unlock_page(page);
|
|
|
|
break;
|
|
|
|
}
|
2005-10-12 16:32:05 -06:00
|
|
|
|
2006-12-23 17:19:07 -07:00
|
|
|
/*
|
|
|
|
* This actually clears the dirty bit in the radix tree.
|
|
|
|
* See cifs_writepage() for more commentary.
|
|
|
|
*/
|
|
|
|
set_page_writeback(page);
|
|
|
|
|
2005-10-12 16:32:05 -06:00
|
|
|
if (page_offset(page) >= mapping->host->i_size) {
|
|
|
|
done = 1;
|
|
|
|
unlock_page(page);
|
2006-12-23 17:19:07 -07:00
|
|
|
end_page_writeback(page);
|
2005-10-12 16:32:05 -06:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2005-10-05 15:50:29 -06:00
|
|
|
/*
|
|
|
|
* BB can we get rid of this? pages are held by pvec
|
|
|
|
*/
|
|
|
|
page_cache_get(page);
|
|
|
|
|
2005-10-12 16:32:05 -06:00
|
|
|
len = min(mapping->host->i_size - page_offset(page),
|
|
|
|
(loff_t)PAGE_CACHE_SIZE);
|
|
|
|
|
2005-10-05 15:50:29 -06:00
|
|
|
/* reserve iov[0] for the smb header */
|
|
|
|
n_iov++;
|
|
|
|
iov[n_iov].iov_base = kmap(page);
|
2005-10-12 16:32:05 -06:00
|
|
|
iov[n_iov].iov_len = len;
|
|
|
|
bytes_to_write += len;
|
2005-10-05 15:50:29 -06:00
|
|
|
|
|
|
|
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) {
|
2005-10-20 14:44:56 -06:00
|
|
|
/* 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;
|
2005-10-11 20:58:06 -06:00
|
|
|
} else {
|
2005-10-20 14:44:56 -06:00
|
|
|
rc = CIFSSMBWrite2(xid, cifs_sb->tcon,
|
|
|
|
open_file->netfid,
|
|
|
|
bytes_to_write, offset,
|
|
|
|
&bytes_written, iov, n_iov,
|
2007-11-13 15:41:37 -07:00
|
|
|
CIFS_LONG_OP);
|
2005-10-20 14:44:56 -06:00
|
|
|
atomic_dec(&open_file->wrtPending);
|
|
|
|
if (rc || bytes_written < bytes_to_write) {
|
2007-07-17 11:34:02 -06:00
|
|
|
cERROR(1, ("Write2 ret %d, wrote %d",
|
2005-10-20 14:44:56 -06:00
|
|
|
rc, bytes_written));
|
|
|
|
/* BB what if continued retry is
|
|
|
|
requested via mount flags? */
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
if (rc == -ENOSPC)
|
|
|
|
set_bit(AS_ENOSPC, &mapping->flags);
|
|
|
|
else
|
|
|
|
set_bit(AS_EIO, &mapping->flags);
|
2005-10-20 14:44:56 -06:00
|
|
|
} else {
|
|
|
|
cifs_stats_bytes_written(cifs_sb->tcon,
|
|
|
|
bytes_written);
|
|
|
|
}
|
2005-10-05 15:50:29 -06:00
|
|
|
}
|
|
|
|
for (i = 0; i < n_iov; i++) {
|
|
|
|
page = pvec.pages[first + i];
|
2006-01-27 16:11:47 -07:00
|
|
|
/* 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
|
2007-07-09 19:16:18 -06:00
|
|
|
when page marked as error */
|
|
|
|
if (rc)
|
2006-01-27 16:11:47 -07:00
|
|
|
SetPageError(page);
|
2005-10-05 15:50:29 -06:00
|
|
|
kunmap(page);
|
|
|
|
unlock_page(page);
|
2006-12-23 17:19:07 -07:00
|
|
|
end_page_writeback(page);
|
2005-10-05 15:50:29 -06:00
|
|
|
page_cache_release(page);
|
|
|
|
}
|
|
|
|
if ((wbc->nr_to_write -= n_iov) <= 0)
|
|
|
|
done = 1;
|
|
|
|
index = next;
|
|
|
|
}
|
|
|
|
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;
|
|
|
|
}
|
[PATCH] writeback: fix range handling
When a writeback_control's `start' and `end' fields are used to
indicate a one-byte-range starting at file offset zero, the required
values of .start=0,.end=0 mean that the ->writepages() implementation
has no way of telling that it is being asked to perform a range
request. Because we're currently overloading (start == 0 && end == 0)
to mean "this is not a write-a-range request".
To make all this sane, the patch changes range of writeback_control.
So caller does: If it is calling ->writepages() to write pages, it
sets range (range_start/end or range_cyclic) always.
And if range_cyclic is true, ->writepages() thinks the range is
cyclic, otherwise it just uses range_start and range_end.
This patch does,
- Add LLONG_MAX, LLONG_MIN, ULLONG_MAX to include/linux/kernel.h
-1 is usually ok for range_end (type is long long). But, if someone did,
range_end += val; range_end is "val - 1"
u64val = range_end >> bits; u64val is "~(0ULL)"
or something, they are wrong. So, this adds LLONG_MAX to avoid nasty
things, and uses LLONG_MAX for range_end.
- All callers of ->writepages() sets range_start/end or range_cyclic.
- Fix updates of ->writeback_index. It seems already bit strange.
If it starts at 0 and ended by check of nr_to_write, this last
index may reduce chance to scan end of file. So, this updates
->writeback_index only if range_cyclic is true or whole-file is
scanned.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Cc: Steven French <sfrench@us.ibm.com>
Cc: "Vladimir V. Saveliev" <vs@namesys.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 03:03:26 -06:00
|
|
|
if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
|
2005-10-05 15:50:29 -06:00
|
|
|
mapping->writeback_index = index;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
FreeXid(xid);
|
2007-02-01 21:21:57 -07:00
|
|
|
kfree(iov);
|
2005-04-16 16:20:36 -06:00
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
static int cifs_writepage(struct page *page, struct writeback_control *wbc)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
int rc = -EFAULT;
|
|
|
|
int xid;
|
|
|
|
|
|
|
|
xid = GetXid();
|
|
|
|
/* BB add check for wbc flags */
|
|
|
|
page_cache_get(page);
|
2008-02-07 16:25:02 -07:00
|
|
|
if (!PageUptodate(page))
|
2005-04-16 16:20:36 -06:00
|
|
|
cFYI(1, ("ppw - page not up to date"));
|
2006-12-23 17:19:07 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
2007-07-09 19:16:18 -06:00
|
|
|
set_page_writeback(page);
|
2005-04-16 16:20:36 -06:00
|
|
|
rc = cifs_partialpagewrite(page, 0, PAGE_CACHE_SIZE);
|
|
|
|
SetPageUptodate(page); /* BB add check for error and Clearuptodate? */
|
|
|
|
unlock_page(page);
|
2006-12-23 17:19:07 -07:00
|
|
|
end_page_writeback(page);
|
|
|
|
page_cache_release(page);
|
2005-04-16 16:20:36 -06:00
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2008-09-24 09:32:59 -06:00
|
|
|
static int cifs_write_end(struct file *file, struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned copied,
|
|
|
|
struct page *page, void *fsdata)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-09-24 09:32:59 -06:00
|
|
|
int rc;
|
|
|
|
struct inode *inode = mapping->host;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-09-24 09:32:59 -06:00
|
|
|
cFYI(1, ("write_end for page %p from pos %lld with %d bytes",
|
|
|
|
page, pos, copied));
|
|
|
|
|
|
|
|
if (!PageUptodate(page) && copied == PAGE_CACHE_SIZE)
|
|
|
|
SetPageUptodate(page);
|
2008-02-07 16:25:02 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!PageUptodate(page)) {
|
2008-09-24 09:32:59 -06:00
|
|
|
char *page_data;
|
|
|
|
unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
|
|
|
|
int xid;
|
|
|
|
|
|
|
|
xid = GetXid();
|
2005-04-16 16:20:36 -06:00
|
|
|
/* 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);
|
2008-09-24 09:32:59 -06:00
|
|
|
rc = cifs_write(file, page_data + offset, copied, &pos);
|
|
|
|
/* if (rc < 0) should we set writebehind rc? */
|
2005-04-16 16:20:36 -06:00
|
|
|
kunmap(page);
|
2008-09-24 09:32:59 -06:00
|
|
|
|
|
|
|
FreeXid(xid);
|
2007-07-09 19:16:18 -06:00
|
|
|
} else {
|
2008-09-24 09:32:59 -06:00
|
|
|
rc = copied;
|
|
|
|
pos += copied;
|
2005-04-16 16:20:36 -06:00
|
|
|
set_page_dirty(page);
|
|
|
|
}
|
|
|
|
|
2008-09-24 09:32:59 -06:00
|
|
|
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);
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
int cifs_fsync(struct file *file, struct dentry *dentry, int datasync)
|
|
|
|
{
|
|
|
|
int xid;
|
|
|
|
int rc = 0;
|
2006-12-08 03:36:48 -07:00
|
|
|
struct inode *inode = file->f_path.dentry->d_inode;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
xid = GetXid();
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
cFYI(1, ("Sync file - name: %s datasync: 0x%x",
|
2005-04-16 16:20:36 -06:00
|
|
|
dentry->d_name.name, datasync));
|
2007-07-12 18:33:32 -06:00
|
|
|
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
rc = filemap_write_and_wait(inode->i_mapping);
|
|
|
|
if (rc == 0) {
|
|
|
|
rc = CIFS_I(inode)->write_behind_rc;
|
2005-04-16 16:20:36 -06:00
|
|
|
CIFS_I(inode)->write_behind_rc = 0;
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2006-03-26 02:37:17 -07:00
|
|
|
/* static void cifs_sync_page(struct page *page)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
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)
|
2006-03-26 02:37:17 -07:00
|
|
|
return; */
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
/* fill in rpages then
|
2005-04-16 16:20:36 -06:00
|
|
|
result = cifs_pagein_inode(inode, index, rpages); */ /* BB finish */
|
|
|
|
|
2006-05-31 12:05:34 -06:00
|
|
|
/* cFYI(1, ("rpages is %d for sync page of Index %ld", rpages, index));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-03-26 02:37:17 -07:00
|
|
|
#if 0
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc < 0)
|
|
|
|
return rc;
|
|
|
|
return 0;
|
2006-03-26 02:37:17 -07:00
|
|
|
#endif
|
2005-04-16 16:20:36 -06:00
|
|
|
} */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* As file closes, flush all cached write data for this inode checking
|
|
|
|
* for write behind errors.
|
|
|
|
*/
|
2006-06-23 03:05:12 -06:00
|
|
|
int cifs_flush(struct file *file, fl_owner_t id)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2007-07-09 19:16:18 -06:00
|
|
|
struct inode *inode = file->f_path.dentry->d_inode;
|
2005-04-16 16:20:36 -06:00
|
|
|
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);
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
/* reset wb rc if we were able to write out dirty pages */
|
|
|
|
if (!rc) {
|
|
|
|
rc = CIFS_I(inode)->write_behind_rc;
|
2005-04-16 16:20:36 -06:00
|
|
|
CIFS_I(inode)->write_behind_rc = 0;
|
[CIFS] Fix potential data corruption when writing out cached dirty pages
Fix RedHat bug 329431
The idea here is separate "conscious" from "unconscious" flushes.
Conscious flushes are those due to a fsync() or close(). Unconscious
ones are flushes that occur as a side effect of some other operation or
due to memory pressure.
Currently, when an error occurs during an unconscious flush (ENOSPC or
EIO), we toss out the page and don't preserve that error to report to
the user when a conscious flush occurs. If after the unconscious flush,
there are no more dirty pages for the inode, the conscious flush will
simply return success even though there were previous errors when writing
out pages. This can lead to data corruption.
The easiest way to reproduce this is to mount up a CIFS share that's
very close to being full or where the user is very close to quota. mv
a file to the share that's slightly larger than the quota allows. The
writes will all succeed (since they go to pagecache). The mv will do a
setattr to set the new file's attributes. This calls
filemap_write_and_wait,
which will return an error since all of the pages can't be written out.
Then later, when the flush and release ops occur, there are no more
dirty pages in pagecache for the file and those operations return 0. mv
then assumes that the file was written out correctly and deletes the
original.
CIFS already has a write_behind_rc variable where it stores the results
from earlier flushes, but that value is only reported in cifs_close.
Since the VFS ignores the return value from the release operation, this
isn't helpful. We should be reporting this error during the flush
operation.
This patch does the following:
1) changes cifs_fsync to use filemap_write_and_wait and cifs_flush and also
sync to check its return code. If it returns successful, they then check
the value of write_behind_rc to see if an earlier flush had reported any
errors. If so, they return that error and clear write_behind_rc.
2) sets write_behind_rc in a few other places where pages are written
out as a side effect of other operations and the code waits on them.
3) changes cifs_setattr to only call filemap_write_and_wait for
ATTR_SIZE changes.
4) makes cifs_writepages accurately distinguish between EIO and ENOSPC
errors when writing out pages.
Some simple testing indicates that the patch works as expected and that
it fixes the reproduceable known problem.
Acked-by: Dave Kleikamp <shaggy@austin.rr.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2007-11-20 16:19:03 -07:00
|
|
|
}
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
cFYI(1, ("Flush inode %p file %p rc %d", inode, file, rc));
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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();
|
2006-12-08 03:36:48 -07:00
|
|
|
cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
pTcon = cifs_sb->tcon;
|
|
|
|
|
|
|
|
if (file->private_data == NULL) {
|
|
|
|
FreeXid(xid);
|
|
|
|
return -EBADF;
|
|
|
|
}
|
|
|
|
open_file = (struct cifsFileInfo *)file->private_data;
|
|
|
|
|
2008-02-07 16:25:02 -07:00
|
|
|
if ((file->f_flags & O_ACCMODE) == O_WRONLY)
|
2005-04-16 16:20:36 -06:00
|
|
|
cFYI(1, ("attempting read on write only file instance"));
|
2008-02-07 16:25:02 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
for (total_read = 0, current_offset = read_data;
|
|
|
|
read_size > total_read;
|
|
|
|
total_read += bytes_read, current_offset += bytes_read) {
|
2007-07-09 19:16:18 -06:00
|
|
|
current_read_size = min_t(const int, read_size - total_read,
|
2005-04-16 16:20:36 -06:00
|
|
|
cifs_sb->rsize);
|
|
|
|
rc = -EAGAIN;
|
|
|
|
smb_read_data = NULL;
|
|
|
|
while (rc == -EAGAIN) {
|
2005-12-12 21:53:18 -07:00
|
|
|
int buf_type = CIFS_NO_BUFFER;
|
2007-07-09 19:16:18 -06:00
|
|
|
if ((open_file->invalidHandle) &&
|
2005-04-16 16:20:36 -06:00
|
|
|
(!open_file->closePend)) {
|
2008-04-28 18:06:05 -06:00
|
|
|
rc = cifs_reopen_file(file, true);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc != 0)
|
|
|
|
break;
|
|
|
|
}
|
2005-08-31 22:50:37 -06:00
|
|
|
rc = CIFSSMBRead(xid, pTcon,
|
2005-12-12 21:53:18 -07:00
|
|
|
open_file->netfid,
|
|
|
|
current_read_size, *poffset,
|
|
|
|
&bytes_read, &smb_read_data,
|
|
|
|
&buf_type);
|
2005-04-16 16:20:36 -06:00
|
|
|
pSMBr = (struct smb_com_read_rsp *)smb_read_data;
|
|
|
|
if (smb_read_data) {
|
2006-02-14 21:30:52 -07:00
|
|
|
if (copy_to_user(current_offset,
|
|
|
|
smb_read_data +
|
|
|
|
4 /* RFC1001 length field */ +
|
|
|
|
le16_to_cpu(pSMBr->DataOffset),
|
2008-02-07 16:25:02 -07:00
|
|
|
bytes_read))
|
2006-02-14 21:30:52 -07:00
|
|
|
rc = -EFAULT;
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
if (buf_type == CIFS_SMALL_BUFFER)
|
2005-12-12 21:53:18 -07:00
|
|
|
cifs_small_buf_release(smb_read_data);
|
2007-07-09 19:16:18 -06:00
|
|
|
else if (buf_type == CIFS_LARGE_BUFFER)
|
2005-12-12 21:53:18 -07:00
|
|
|
cifs_buf_release(smb_read_data);
|
2005-04-16 16:20:36 -06:00
|
|
|
smb_read_data = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (rc || (bytes_read == 0)) {
|
|
|
|
if (total_read) {
|
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
} else {
|
2005-08-24 14:59:35 -06:00
|
|
|
cifs_stats_bytes_read(pTcon, bytes_read);
|
2005-04-16 16:20:36 -06:00
|
|
|
*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;
|
2005-12-12 21:53:18 -07:00
|
|
|
int buf_type = CIFS_NO_BUFFER;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
xid = GetXid();
|
2006-12-08 03:36:48 -07:00
|
|
|
cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
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"));
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
for (total_read = 0, current_offset = read_data;
|
2005-04-16 16:20:36 -06:00
|
|
|
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);
|
2005-09-16 00:06:38 -06:00
|
|
|
/* For windows me and 9x we do not want to request more
|
|
|
|
than it negotiated since it will refuse the read then */
|
2007-07-09 19:16:18 -06:00
|
|
|
if ((pTcon->ses) &&
|
2005-09-16 00:06:38 -06:00
|
|
|
!(pTcon->ses->capabilities & CAP_LARGE_FILES)) {
|
|
|
|
current_read_size = min_t(const int, current_read_size,
|
|
|
|
pTcon->ses->server->maxBuf - 128);
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
rc = -EAGAIN;
|
|
|
|
while (rc == -EAGAIN) {
|
2007-07-09 19:16:18 -06:00
|
|
|
if ((open_file->invalidHandle) &&
|
2005-04-16 16:20:36 -06:00
|
|
|
(!open_file->closePend)) {
|
2008-04-28 18:06:05 -06:00
|
|
|
rc = cifs_reopen_file(file, true);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc != 0)
|
|
|
|
break;
|
|
|
|
}
|
2005-08-31 22:50:37 -06:00
|
|
|
rc = CIFSSMBRead(xid, pTcon,
|
2005-12-12 21:53:18 -07:00
|
|
|
open_file->netfid,
|
|
|
|
current_read_size, *poffset,
|
|
|
|
&bytes_read, ¤t_offset,
|
|
|
|
&buf_type);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
if (rc || (bytes_read == 0)) {
|
|
|
|
if (total_read) {
|
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
} else {
|
2005-08-24 14:59:35 -06:00
|
|
|
cifs_stats_bytes_read(pTcon, total_read);
|
2005-04-16 16:20:36 -06:00
|
|
|
*poffset += bytes_read;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
FreeXid(xid);
|
|
|
|
return total_read;
|
|
|
|
}
|
|
|
|
|
|
|
|
int cifs_file_mmap(struct file *file, struct vm_area_struct *vma)
|
|
|
|
{
|
2006-12-08 03:36:48 -07:00
|
|
|
struct dentry *dentry = file->f_path.dentry;
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
static void cifs_copy_cache_pages(struct address_space *mapping,
|
2005-04-16 16:20:36 -06:00
|
|
|
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"));
|
2005-06-09 15:44:07 -06:00
|
|
|
data += PAGE_CACHE_SIZE;
|
|
|
|
bytes_read -= PAGE_CACHE_SIZE;
|
2005-04-16 16:20:36 -06:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
target = kmap_atomic(page, KM_USER0);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
if (PAGE_CACHE_SIZE > bytes_read) {
|
|
|
|
memcpy(target, data, bytes_read);
|
|
|
|
/* zero the tail end of this partial page */
|
2007-07-09 19:16:18 -06:00
|
|
|
memset(target + bytes_read, 0,
|
2005-04-16 16:20:36 -06:00
|
|
|
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))
|
|
|
|
__pagevec_lru_add(plru_pvec);
|
|
|
|
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;
|
2007-10-12 13:10:28 -06:00
|
|
|
unsigned int bytes_read = 0;
|
2007-07-09 19:16:18 -06:00
|
|
|
unsigned int read_size, i;
|
2005-04-16 16:20:36 -06:00
|
|
|
char *smb_read_data = NULL;
|
|
|
|
struct smb_com_read_rsp *pSMBr;
|
|
|
|
struct pagevec lru_pvec;
|
|
|
|
struct cifsFileInfo *open_file;
|
2005-12-12 21:53:18 -07:00
|
|
|
int buf_type = CIFS_NO_BUFFER;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
xid = GetXid();
|
|
|
|
if (file->private_data == NULL) {
|
|
|
|
FreeXid(xid);
|
|
|
|
return -EBADF;
|
|
|
|
}
|
|
|
|
open_file = (struct cifsFileInfo *)file->private_data;
|
2006-12-08 03:36:48 -07:00
|
|
|
cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
|
2005-04-16 16:20:36 -06:00
|
|
|
pTcon = cifs_sb->tcon;
|
2005-08-31 22:50:37 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
pagevec_init(&lru_pvec, 0);
|
2008-02-12 13:32:36 -07:00
|
|
|
cFYI(DBG2, ("rpages: num pages %d", num_pages));
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
2007-07-09 19:16:18 -06:00
|
|
|
expected_index =
|
2005-04-16 16:20:36 -06:00
|
|
|
list_entry(page_list->prev, struct page, lru)->index;
|
2007-07-09 19:16:18 -06:00
|
|
|
list_for_each_entry_reverse(tmp_page, page_list, lru) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (tmp_page->index == expected_index) {
|
|
|
|
contig_pages++;
|
|
|
|
expected_index++;
|
|
|
|
} else
|
2007-07-09 19:16:18 -06:00
|
|
|
break;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
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);
|
2008-02-12 13:32:36 -07:00
|
|
|
cFYI(DBG2, ("rpages: read size 0x%x contiguous pages %d",
|
2007-06-24 12:30:48 -06:00
|
|
|
read_size, contig_pages));
|
2005-04-16 16:20:36 -06:00
|
|
|
rc = -EAGAIN;
|
|
|
|
while (rc == -EAGAIN) {
|
2007-07-09 19:16:18 -06:00
|
|
|
if ((open_file->invalidHandle) &&
|
2005-04-16 16:20:36 -06:00
|
|
|
(!open_file->closePend)) {
|
2008-04-28 18:06:05 -06:00
|
|
|
rc = cifs_reopen_file(file, true);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc != 0)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2005-08-31 22:50:37 -06:00
|
|
|
rc = CIFSSMBRead(xid, pTcon,
|
2005-12-12 21:53:18 -07:00
|
|
|
open_file->netfid,
|
|
|
|
read_size, offset,
|
|
|
|
&bytes_read, &smb_read_data,
|
|
|
|
&buf_type);
|
2005-08-25 00:06:05 -06:00
|
|
|
/* BB more RC checks ? */
|
2007-07-09 19:16:18 -06:00
|
|
|
if (rc == -EAGAIN) {
|
2005-04-16 16:20:36 -06:00
|
|
|
if (smb_read_data) {
|
2007-07-09 19:16:18 -06:00
|
|
|
if (buf_type == CIFS_SMALL_BUFFER)
|
2005-12-12 21:53:18 -07:00
|
|
|
cifs_small_buf_release(smb_read_data);
|
2007-07-09 19:16:18 -06:00
|
|
|
else if (buf_type == CIFS_LARGE_BUFFER)
|
2005-12-12 21:53:18 -07:00
|
|
|
cifs_buf_release(smb_read_data);
|
2005-04-16 16:20:36 -06:00
|
|
|
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) {
|
2006-12-10 03:19:44 -07:00
|
|
|
task_io_account_read(bytes_read);
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
2005-08-24 14:59:35 -06:00
|
|
|
cifs_stats_bytes_read(pTcon, bytes_read);
|
2007-10-12 13:10:28 -06:00
|
|
|
if ((bytes_read & PAGE_CACHE_MASK) != bytes_read) {
|
2005-04-16 16:20:36 -06:00
|
|
|
i++; /* account for partial page */
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
/* server copy of file can have smaller size
|
2005-04-16 16:20:36 -06:00
|
|
|
than client */
|
2007-07-09 19:16:18 -06:00
|
|
|
/* BB do we need to verify this common case ?
|
|
|
|
this case is ok - if we are at server EOF
|
2005-04-16 16:20:36 -06:00
|
|
|
we will hit it on next read */
|
|
|
|
|
2006-11-02 23:07:08 -07:00
|
|
|
/* break; */
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
cFYI(1, ("No bytes read (%d) at offset %lld . "
|
|
|
|
"Cleaning remaining pages from readahead list",
|
|
|
|
bytes_read, offset));
|
2007-07-09 19:16:18 -06:00
|
|
|
/* BB turn off caching and do new lookup on
|
2005-04-16 16:20:36 -06:00
|
|
|
file size at server? */
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (smb_read_data) {
|
2007-07-09 19:16:18 -06:00
|
|
|
if (buf_type == CIFS_SMALL_BUFFER)
|
2005-12-12 21:53:18 -07:00
|
|
|
cifs_small_buf_release(smb_read_data);
|
2007-07-09 19:16:18 -06:00
|
|
|
else if (buf_type == CIFS_LARGE_BUFFER)
|
2005-12-12 21:53:18 -07:00
|
|
|
cifs_buf_release(smb_read_data);
|
2005-04-16 16:20:36 -06:00
|
|
|
smb_read_data = NULL;
|
|
|
|
}
|
|
|
|
bytes_read = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
pagevec_lru_add(&lru_pvec);
|
|
|
|
|
|
|
|
/* need to free smb_read_data buf before exit */
|
|
|
|
if (smb_read_data) {
|
2007-07-09 19:16:18 -06:00
|
|
|
if (buf_type == CIFS_SMALL_BUFFER)
|
2006-01-18 15:20:39 -07:00
|
|
|
cifs_small_buf_release(smb_read_data);
|
2007-07-09 19:16:18 -06:00
|
|
|
else if (buf_type == CIFS_LARGE_BUFFER)
|
2006-01-18 15:20:39 -07:00
|
|
|
cifs_buf_release(smb_read_data);
|
2005-04-16 16:20:36 -06:00
|
|
|
smb_read_data = NULL;
|
2007-07-09 19:16:18 -06:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
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 */
|
2007-07-09 19:16:18 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
rc = cifs_read(file, read_data, PAGE_CACHE_SIZE, poffset);
|
2007-07-09 19:16:18 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (rc < 0)
|
|
|
|
goto io_error;
|
|
|
|
else
|
2007-07-09 19:16:18 -06:00
|
|
|
cFYI(1, ("Bytes read %d", rc));
|
|
|
|
|
2006-12-08 03:36:48 -07:00
|
|
|
file->f_path.dentry->d_inode->i_atime =
|
|
|
|
current_fs_time(file->f_path.dentry->d_inode->i_sb);
|
2007-07-09 19:16:18 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
if (PAGE_CACHE_SIZE > rc)
|
|
|
|
memset(read_data + rc, 0, PAGE_CACHE_SIZE - rc);
|
|
|
|
|
|
|
|
flush_dcache_page(page);
|
|
|
|
SetPageUptodate(page);
|
|
|
|
rc = 0;
|
2007-07-09 19:16:18 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
io_error:
|
2007-07-09 19:16:18 -06:00
|
|
|
kunmap(page);
|
2005-04-16 16:20:36 -06:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
cFYI(1, ("readpage %p at offset %d 0x%x\n",
|
2005-04-16 16:20:36 -06:00
|
|
|
page, (int)offset, (int)offset));
|
|
|
|
|
|
|
|
rc = cifs_readpage_worker(file, page, &offset);
|
|
|
|
|
|
|
|
unlock_page(page);
|
|
|
|
|
|
|
|
FreeXid(xid);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2007-07-26 09:54:16 -06:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/* 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
|
2007-07-09 19:16:18 -06:00
|
|
|
refreshing the inode only on increases in the file size
|
2005-04-16 16:20:36 -06:00
|
|
|
but this is tricky to do without racing with writebehind
|
|
|
|
page caching in the current Linux kernel design */
|
2008-04-28 18:06:05 -06:00
|
|
|
bool is_size_safe_to_change(struct cifsInodeInfo *cifsInode, __u64 end_of_file)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2007-07-26 09:54:16 -06:00
|
|
|
if (!cifsInode)
|
2008-04-28 18:06:05 -06:00
|
|
|
return true;
|
2007-07-12 18:33:32 -06:00
|
|
|
|
2007-07-26 09:54:16 -06:00
|
|
|
if (is_inode_writable(cifsInode)) {
|
|
|
|
/* This inode is open for write at least once */
|
2006-01-12 15:41:28 -07:00
|
|
|
struct cifs_sb_info *cifs_sb;
|
|
|
|
|
|
|
|
cifs_sb = CIFS_SB(cifsInode->vfs_inode.i_sb);
|
2008-02-07 16:25:02 -07:00
|
|
|
if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DIRECT_IO) {
|
2007-07-09 19:16:18 -06:00
|
|
|
/* since no page cache to corrupt on directio
|
2006-01-12 15:41:28 -07:00
|
|
|
we can change size safely */
|
2008-04-28 18:06:05 -06:00
|
|
|
return true;
|
2006-01-12 15:41:28 -07:00
|
|
|
}
|
|
|
|
|
2007-07-09 19:16:18 -06:00
|
|
|
if (i_size_read(&cifsInode->vfs_inode) < end_of_file)
|
2008-04-28 18:06:05 -06:00
|
|
|
return true;
|
2007-02-08 11:14:13 -07:00
|
|
|
|
2008-04-28 18:06:05 -06:00
|
|
|
return false;
|
2005-10-20 14:44:56 -06:00
|
|
|
} else
|
2008-04-28 18:06:05 -06:00
|
|
|
return true;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2008-09-24 09:32:59 -06:00
|
|
|
static int cifs_write_begin(struct file *file, struct address_space *mapping,
|
|
|
|
loff_t pos, unsigned len, unsigned flags,
|
|
|
|
struct page **pagep, void **fsdata)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2008-09-24 09:32:59 -06:00
|
|
|
pgoff_t index = pos >> PAGE_CACHE_SHIFT;
|
|
|
|
loff_t offset = pos & (PAGE_CACHE_SIZE - 1);
|
|
|
|
|
|
|
|
cFYI(1, ("write_begin from %lld len %d", (long long)pos, len));
|
|
|
|
|
|
|
|
*pagep = __grab_cache_page(mapping, index);
|
|
|
|
if (!*pagep)
|
|
|
|
return -ENOMEM;
|
2007-03-05 17:31:00 -07:00
|
|
|
|
2008-09-24 09:32:59 -06:00
|
|
|
if (PageUptodate(*pagep))
|
2007-03-05 17:31:00 -07:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* If we are writing a full page it will be up to date,
|
|
|
|
no need to read from the server */
|
2008-09-24 09:32:59 -06:00
|
|
|
if (len == PAGE_CACHE_SIZE && flags & AOP_FLAG_UNINTERRUPTIBLE)
|
2007-03-05 17:31:00 -07:00
|
|
|
return 0;
|
|
|
|
|
2008-09-24 09:32:59 -06:00
|
|
|
if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
|
|
|
|
int rc;
|
2007-03-05 17:31:00 -07:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/* might as well read a page, it is fast enough */
|
2008-09-24 09:32:59 -06:00
|
|
|
rc = cifs_readpage_worker(file, *pagep, &offset);
|
|
|
|
|
|
|
|
/* we do not need to pass errors back
|
|
|
|
e.g. if we do not have read access to the file
|
|
|
|
because cifs_write_end will attempt synchronous writes
|
|
|
|
-- shaggy */
|
2007-03-05 17:31:00 -07:00
|
|
|
} 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
|
2008-09-24 09:32:59 -06:00
|
|
|
this will be written out by write_end so is fine */
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-06-28 05:26:44 -06:00
|
|
|
const struct address_space_operations cifs_addr_ops = {
|
2005-04-16 16:20:36 -06:00
|
|
|
.readpage = cifs_readpage,
|
|
|
|
.readpages = cifs_readpages,
|
|
|
|
.writepage = cifs_writepage,
|
2005-10-05 15:50:29 -06:00
|
|
|
.writepages = cifs_writepages,
|
2008-09-24 09:32:59 -06:00
|
|
|
.write_begin = cifs_write_begin,
|
|
|
|
.write_end = cifs_write_end,
|
2005-04-16 16:20:36 -06:00
|
|
|
.set_page_dirty = __set_page_dirty_nobuffers,
|
|
|
|
/* .sync_page = cifs_sync_page, */
|
|
|
|
/* .direct_IO = */
|
|
|
|
};
|
2006-06-01 13:41:23 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
2006-06-28 05:26:44 -06:00
|
|
|
const struct address_space_operations cifs_addr_ops_smallbuf = {
|
2006-06-01 13:41:23 -06:00
|
|
|
.readpage = cifs_readpage,
|
|
|
|
.writepage = cifs_writepage,
|
|
|
|
.writepages = cifs_writepages,
|
2008-09-24 09:32:59 -06:00
|
|
|
.write_begin = cifs_write_begin,
|
|
|
|
.write_end = cifs_write_end,
|
2006-06-01 13:41:23 -06:00
|
|
|
.set_page_dirty = __set_page_dirty_nobuffers,
|
|
|
|
/* .sync_page = cifs_sync_page, */
|
|
|
|
/* .direct_IO = */
|
|
|
|
};
|