kernel_optimize_test/fs/cifs/link.c
Steve French 8b1327f6ed [CIFS] file create with acl support enabled is slow
Shirish Pargaonkar noted:
With cifsacl mount option, when a file is created on the Windows server,
exclusive oplock is broken right away because the get cifs acl code
again opens the file to obtain security descriptor.
The client does not have the newly created file handle or inode in any
of its lists yet so it does not respond to oplock break and server waits for
its duration and then responds to the second open. This slows down file
creation signficantly.  The fix is to pass the file descriptor to the get
cifsacl code wherever available so that get cifs acl code does not send
second open (NT Create ANDX) and oplock is not broken.

CC: Shirish Pargaonkar <shirishp@us.ibm.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
2008-03-14 22:37:16 +00:00

362 lines
9.5 KiB
C

/*
* fs/cifs/link.c
*
* Copyright (C) International Business Machines Corp., 2002,2008
* Author(s): Steve French (sfrench@us.ibm.com)
*
* 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>
#include <linux/stat.h>
#include <linux/namei.h>
#include "cifsfs.h"
#include "cifspdu.h"
#include "cifsglob.h"
#include "cifsproto.h"
#include "cifs_debug.h"
#include "cifs_fs_sb.h"
int
cifs_hardlink(struct dentry *old_file, struct inode *inode,
struct dentry *direntry)
{
int rc = -EACCES;
int xid;
char *fromName = NULL;
char *toName = NULL;
struct cifs_sb_info *cifs_sb_target;
struct cifsTconInfo *pTcon;
struct cifsInodeInfo *cifsInode;
xid = GetXid();
cifs_sb_target = CIFS_SB(inode->i_sb);
pTcon = cifs_sb_target->tcon;
/* No need to check for cross device links since server will do that
BB note DFS case in future though (when we may have to check) */
fromName = build_path_from_dentry(old_file);
toName = build_path_from_dentry(direntry);
if ((fromName == NULL) || (toName == NULL)) {
rc = -ENOMEM;
goto cifs_hl_exit;
}
/* if (cifs_sb_target->tcon->ses->capabilities & CAP_UNIX)*/
if (pTcon->unix_ext)
rc = CIFSUnixCreateHardLink(xid, pTcon, fromName, toName,
cifs_sb_target->local_nls,
cifs_sb_target->mnt_cifs_flags &
CIFS_MOUNT_MAP_SPECIAL_CHR);
else {
rc = CIFSCreateHardLink(xid, pTcon, fromName, toName,
cifs_sb_target->local_nls,
cifs_sb_target->mnt_cifs_flags &
CIFS_MOUNT_MAP_SPECIAL_CHR);
if ((rc == -EIO) || (rc == -EINVAL))
rc = -EOPNOTSUPP;
}
d_drop(direntry); /* force new lookup from server of target */
/* if source file is cached (oplocked) revalidate will not go to server
until the file is closed or oplock broken so update nlinks locally */
if (old_file->d_inode) {
cifsInode = CIFS_I(old_file->d_inode);
if (rc == 0) {
old_file->d_inode->i_nlink++;
/* BB should we make this contingent on superblock flag NOATIME? */
/* old_file->d_inode->i_ctime = CURRENT_TIME;*/
/* parent dir timestamps will update from srv
within a second, would it really be worth it
to set the parent dir cifs inode time to zero
to force revalidate (faster) for it too? */
}
/* if not oplocked will force revalidate to get info
on source file from srv */
cifsInode->time = 0;
/* Will update parent dir timestamps from srv within a second.
Would it really be worth it to set the parent dir (cifs
inode) time field to zero to force revalidate on parent
directory faster ie
CIFS_I(inode)->time = 0; */
}
cifs_hl_exit:
kfree(fromName);
kfree(toName);
FreeXid(xid);
return rc;
}
void *
cifs_follow_link(struct dentry *direntry, struct nameidata *nd)
{
struct inode *inode = direntry->d_inode;
int rc = -EACCES;
int xid;
char *full_path = NULL;
char *target_path = ERR_PTR(-ENOMEM);
struct cifs_sb_info *cifs_sb;
struct cifsTconInfo *pTcon;
xid = GetXid();
full_path = build_path_from_dentry(direntry);
if (!full_path)
goto out_no_free;
cFYI(1, ("Full path: %s inode = 0x%p", full_path, inode));
cifs_sb = CIFS_SB(inode->i_sb);
pTcon = cifs_sb->tcon;
target_path = kmalloc(PATH_MAX, GFP_KERNEL);
if (!target_path) {
target_path = ERR_PTR(-ENOMEM);
goto out;
}
/* We could change this to:
if (pTcon->unix_ext)
but there does not seem any point in refusing to
get symlink info if we can, even if unix extensions
turned off for this mount */
if (pTcon->ses->capabilities & CAP_UNIX)
rc = CIFSSMBUnixQuerySymLink(xid, pTcon, full_path,
target_path,
PATH_MAX-1,
cifs_sb->local_nls);
else {
/* BB add read reparse point symlink code here */
/* rc = CIFSSMBQueryReparseLinkInfo */
/* BB Add code to Query ReparsePoint info */
/* BB Add MAC style xsymlink check here if enabled */
}
if (rc == 0) {
/* BB Add special case check for Samba DFS symlinks */
target_path[PATH_MAX-1] = 0;
} else {
kfree(target_path);
target_path = ERR_PTR(rc);
}
out:
kfree(full_path);
out_no_free:
FreeXid(xid);
nd_set_link(nd, target_path);
return NULL; /* No cookie */
}
int
cifs_symlink(struct inode *inode, struct dentry *direntry, const char *symname)
{
int rc = -EOPNOTSUPP;
int xid;
struct cifs_sb_info *cifs_sb;
struct cifsTconInfo *pTcon;
char *full_path = NULL;
struct inode *newinode = NULL;
xid = GetXid();
cifs_sb = CIFS_SB(inode->i_sb);
pTcon = cifs_sb->tcon;
full_path = build_path_from_dentry(direntry);
if (full_path == NULL) {
FreeXid(xid);
return -ENOMEM;
}
cFYI(1, ("Full path: %s", full_path));
cFYI(1, ("symname is %s", symname));
/* BB what if DFS and this volume is on different share? BB */
if (pTcon->unix_ext)
rc = CIFSUnixCreateSymLink(xid, pTcon, full_path, symname,
cifs_sb->local_nls);
/* else
rc = CIFSCreateReparseSymLink(xid, pTcon, fromName, toName,
cifs_sb_target->local_nls); */
if (rc == 0) {
if (pTcon->unix_ext)
rc = cifs_get_inode_info_unix(&newinode, full_path,
inode->i_sb, xid);
else
rc = cifs_get_inode_info(&newinode, full_path, NULL,
inode->i_sb, xid, NULL);
if (rc != 0) {
cFYI(1, ("Create symlink ok, getinodeinfo fail rc = %d",
rc));
} else {
if (pTcon->nocase)
direntry->d_op = &cifs_ci_dentry_ops;
else
direntry->d_op = &cifs_dentry_ops;
d_instantiate(direntry, newinode);
}
}
kfree(full_path);
FreeXid(xid);
return rc;
}
int
cifs_readlink(struct dentry *direntry, char __user *pBuffer, int buflen)
{
struct inode *inode = direntry->d_inode;
int rc = -EACCES;
int xid;
int oplock = FALSE;
struct cifs_sb_info *cifs_sb;
struct cifsTconInfo *pTcon;
char *full_path = NULL;
char *tmp_path = NULL;
char *tmpbuffer;
int len;
__u16 fid;
xid = GetXid();
cifs_sb = CIFS_SB(inode->i_sb);
pTcon = cifs_sb->tcon;
/* BB would it be safe against deadlock to grab this sem
even though rename itself grabs the sem and calls lookup? */
/* mutex_lock(&inode->i_sb->s_vfs_rename_mutex);*/
full_path = build_path_from_dentry(direntry);
/* mutex_unlock(&inode->i_sb->s_vfs_rename_mutex);*/
if (full_path == NULL) {
FreeXid(xid);
return -ENOMEM;
}
cFYI(1,
("Full path: %s inode = 0x%p pBuffer = 0x%p buflen = %d",
full_path, inode, pBuffer, buflen));
if (buflen > PATH_MAX)
len = PATH_MAX;
else
len = buflen;
tmpbuffer = kmalloc(len, GFP_KERNEL);
if (tmpbuffer == NULL) {
kfree(full_path);
FreeXid(xid);
return -ENOMEM;
}
/* BB add read reparse point symlink code and
Unix extensions symlink code here BB */
/* We could disable this based on pTcon->unix_ext flag instead ... but why? */
if (cifs_sb->tcon->ses->capabilities & CAP_UNIX)
rc = CIFSSMBUnixQuerySymLink(xid, pTcon, full_path,
tmpbuffer,
len - 1,
cifs_sb->local_nls);
else if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_UNX_EMUL) {
cERROR(1, ("SFU style symlinks not implemented yet"));
/* add open and read as in fs/cifs/inode.c */
} else {
rc = CIFSSMBOpen(xid, pTcon, full_path, FILE_OPEN, GENERIC_READ,
OPEN_REPARSE_POINT, &fid, &oplock, NULL,
cifs_sb->local_nls,
cifs_sb->mnt_cifs_flags &
CIFS_MOUNT_MAP_SPECIAL_CHR);
if (!rc) {
rc = CIFSSMBQueryReparseLinkInfo(xid, pTcon, full_path,
tmpbuffer,
len - 1,
fid,
cifs_sb->local_nls);
if (CIFSSMBClose(xid, pTcon, fid)) {
cFYI(1, ("Error closing junction point "
"(open for ioctl)"));
}
/* BB unwind this long, nested function, or remove BB */
if (rc == -EIO) {
/* Query if DFS Junction */
unsigned int num_referrals = 0;
struct dfs_info3_param *refs = NULL;
tmp_path =
kmalloc(MAX_TREE_SIZE + MAX_PATHCONF + 1,
GFP_KERNEL);
if (tmp_path) {
strncpy(tmp_path, pTcon->treeName,
MAX_TREE_SIZE);
strncat(tmp_path, full_path,
MAX_PATHCONF);
rc = get_dfs_path(xid, pTcon->ses,
tmp_path,
cifs_sb->local_nls,
&num_referrals, &refs,
cifs_sb->mnt_cifs_flags &
CIFS_MOUNT_MAP_SPECIAL_CHR);
cFYI(1, ("Get DFS for %s rc = %d ",
tmp_path, rc));
if ((num_referrals == 0) && (rc == 0))
rc = -EACCES;
else {
cFYI(1, ("num referral: %d",
num_referrals));
if (refs && refs->path_name) {
strncpy(tmpbuffer,
refs->path_name,
len-1);
}
}
kfree(refs);
kfree(tmp_path);
}
/* BB add code like else decode referrals
then memcpy to tmpbuffer and free referrals
string array BB */
}
}
}
/* BB Anything else to do to handle recursive links? */
/* BB Should we be using page ops here? */
/* BB null terminate returned string in pBuffer? BB */
if (rc == 0) {
rc = vfs_readlink(direntry, pBuffer, len, tmpbuffer);
cFYI(1,
("vfs_readlink called from cifs_readlink returned %d",
rc));
}
kfree(tmpbuffer);
kfree(full_path);
FreeXid(xid);
return rc;
}
void cifs_put_link(struct dentry *direntry, struct nameidata *nd, void *cookie)
{
char *p = nd_get_link(nd);
if (!IS_ERR(p))
kfree(p);
}