Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / cifs / dir.c
diff --git a/kernel/fs/cifs/dir.c b/kernel/fs/cifs/dir.c
new file mode 100644 (file)
index 0000000..c3eb998
--- /dev/null
@@ -0,0 +1,924 @@
+/*
+ *   fs/cifs/dir.c
+ *
+ *   vfs operations that deal with dentries
+ *
+ *   Copyright (C) International Business Machines  Corp., 2002,2009
+ *   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/slab.h>
+#include <linux/namei.h>
+#include <linux/mount.h>
+#include <linux/file.h>
+#include "cifsfs.h"
+#include "cifspdu.h"
+#include "cifsglob.h"
+#include "cifsproto.h"
+#include "cifs_debug.h"
+#include "cifs_fs_sb.h"
+#include "cifs_unicode.h"
+
+static void
+renew_parental_timestamps(struct dentry *direntry)
+{
+       /* BB check if there is a way to get the kernel to do this or if we
+          really need this */
+       do {
+               direntry->d_time = jiffies;
+               direntry = direntry->d_parent;
+       } while (!IS_ROOT(direntry));
+}
+
+char *
+cifs_build_path_to_root(struct smb_vol *vol, struct cifs_sb_info *cifs_sb,
+                       struct cifs_tcon *tcon)
+{
+       int pplen = vol->prepath ? strlen(vol->prepath) + 1 : 0;
+       int dfsplen;
+       char *full_path = NULL;
+
+       /* if no prefix path, simply set path to the root of share to "" */
+       if (pplen == 0) {
+               full_path = kzalloc(1, GFP_KERNEL);
+               return full_path;
+       }
+
+       if (tcon->Flags & SMB_SHARE_IS_IN_DFS)
+               dfsplen = strnlen(tcon->treeName, MAX_TREE_SIZE + 1);
+       else
+               dfsplen = 0;
+
+       full_path = kmalloc(dfsplen + pplen + 1, GFP_KERNEL);
+       if (full_path == NULL)
+               return full_path;
+
+       if (dfsplen)
+               strncpy(full_path, tcon->treeName, dfsplen);
+       full_path[dfsplen] = CIFS_DIR_SEP(cifs_sb);
+       strncpy(full_path + dfsplen + 1, vol->prepath, pplen);
+       convert_delimiter(full_path, CIFS_DIR_SEP(cifs_sb));
+       full_path[dfsplen + pplen] = 0; /* add trailing null */
+       return full_path;
+}
+
+/* Note: caller must free return buffer */
+char *
+build_path_from_dentry(struct dentry *direntry)
+{
+       struct dentry *temp;
+       int namelen;
+       int dfsplen;
+       char *full_path;
+       char dirsep;
+       struct cifs_sb_info *cifs_sb = CIFS_SB(direntry->d_sb);
+       struct cifs_tcon *tcon = cifs_sb_master_tcon(cifs_sb);
+       unsigned seq;
+
+       dirsep = CIFS_DIR_SEP(cifs_sb);
+       if (tcon->Flags & SMB_SHARE_IS_IN_DFS)
+               dfsplen = strnlen(tcon->treeName, MAX_TREE_SIZE + 1);
+       else
+               dfsplen = 0;
+cifs_bp_rename_retry:
+       namelen = dfsplen;
+       seq = read_seqbegin(&rename_lock);
+       rcu_read_lock();
+       for (temp = direntry; !IS_ROOT(temp);) {
+               namelen += (1 + temp->d_name.len);
+               temp = temp->d_parent;
+               if (temp == NULL) {
+                       cifs_dbg(VFS, "corrupt dentry\n");
+                       rcu_read_unlock();
+                       return NULL;
+               }
+       }
+       rcu_read_unlock();
+
+       full_path = kmalloc(namelen+1, GFP_KERNEL);
+       if (full_path == NULL)
+               return full_path;
+       full_path[namelen] = 0; /* trailing null */
+       rcu_read_lock();
+       for (temp = direntry; !IS_ROOT(temp);) {
+               spin_lock(&temp->d_lock);
+               namelen -= 1 + temp->d_name.len;
+               if (namelen < 0) {
+                       spin_unlock(&temp->d_lock);
+                       break;
+               } else {
+                       full_path[namelen] = dirsep;
+                       strncpy(full_path + namelen + 1, temp->d_name.name,
+                               temp->d_name.len);
+                       cifs_dbg(FYI, "name: %s\n", full_path + namelen);
+               }
+               spin_unlock(&temp->d_lock);
+               temp = temp->d_parent;
+               if (temp == NULL) {
+                       cifs_dbg(VFS, "corrupt dentry\n");
+                       rcu_read_unlock();
+                       kfree(full_path);
+                       return NULL;
+               }
+       }
+       rcu_read_unlock();
+       if (namelen != dfsplen || read_seqretry(&rename_lock, seq)) {
+               cifs_dbg(FYI, "did not end path lookup where expected. namelen=%ddfsplen=%d\n",
+                        namelen, dfsplen);
+               /* presumably this is only possible if racing with a rename
+               of one of the parent directories  (we can not lock the dentries
+               above us to prevent this, but retrying should be harmless) */
+               kfree(full_path);
+               goto cifs_bp_rename_retry;
+       }
+       /* DIR_SEP already set for byte  0 / vs \ but not for
+          subsequent slashes in prepath which currently must
+          be entered the right way - not sure if there is an alternative
+          since the '\' is a valid posix character so we can not switch
+          those safely to '/' if any are found in the middle of the prepath */
+       /* BB test paths to Windows with '/' in the midst of prepath */
+
+       if (dfsplen) {
+               strncpy(full_path, tcon->treeName, dfsplen);
+               if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_POSIX_PATHS) {
+                       int i;
+                       for (i = 0; i < dfsplen; i++) {
+                               if (full_path[i] == '\\')
+                                       full_path[i] = '/';
+                       }
+               }
+       }
+       return full_path;
+}
+
+/*
+ * Don't allow the separator character in a path component.
+ * The VFS will not allow "/", but "\" is allowed by posix.
+ */
+static int
+check_name(struct dentry *direntry)
+{
+       struct cifs_sb_info *cifs_sb = CIFS_SB(direntry->d_sb);
+       int i;
+
+       if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_POSIX_PATHS)) {
+               for (i = 0; i < direntry->d_name.len; i++) {
+                       if (direntry->d_name.name[i] == '\\') {
+                               cifs_dbg(FYI, "Invalid file name\n");
+                               return -EINVAL;
+                       }
+               }
+       }
+       return 0;
+}
+
+
+/* Inode operations in similar order to how they appear in Linux file fs.h */
+
+static int
+cifs_do_create(struct inode *inode, struct dentry *direntry, unsigned int xid,
+              struct tcon_link *tlink, unsigned oflags, umode_t mode,
+              __u32 *oplock, struct cifs_fid *fid)
+{
+       int rc = -ENOENT;
+       int create_options = CREATE_NOT_DIR;
+       int desired_access;
+       struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
+       struct cifs_tcon *tcon = tlink_tcon(tlink);
+       char *full_path = NULL;
+       FILE_ALL_INFO *buf = NULL;
+       struct inode *newinode = NULL;
+       int disposition;
+       struct TCP_Server_Info *server = tcon->ses->server;
+       struct cifs_open_parms oparms;
+
+       *oplock = 0;
+       if (tcon->ses->server->oplocks)
+               *oplock = REQ_OPLOCK;
+
+       full_path = build_path_from_dentry(direntry);
+       if (full_path == NULL) {
+               rc = -ENOMEM;
+               goto out;
+       }
+
+       if (tcon->unix_ext && cap_unix(tcon->ses) && !tcon->broken_posix_open &&
+           (CIFS_UNIX_POSIX_PATH_OPS_CAP &
+                       le64_to_cpu(tcon->fsUnixInfo.Capability))) {
+               rc = cifs_posix_open(full_path, &newinode, inode->i_sb, mode,
+                                    oflags, oplock, &fid->netfid, xid);
+               switch (rc) {
+               case 0:
+                       if (newinode == NULL) {
+                               /* query inode info */
+                               goto cifs_create_get_file_info;
+                       }
+
+                       if (!S_ISREG(newinode->i_mode)) {
+                               /*
+                                * The server may allow us to open things like
+                                * FIFOs, but the client isn't set up to deal
+                                * with that. If it's not a regular file, just
+                                * close it and proceed as if it were a normal
+                                * lookup.
+                                */
+                               CIFSSMBClose(xid, tcon, fid->netfid);
+                               goto cifs_create_get_file_info;
+                       }
+                       /* success, no need to query */
+                       goto cifs_create_set_dentry;
+
+               case -ENOENT:
+                       goto cifs_create_get_file_info;
+
+               case -EIO:
+               case -EINVAL:
+                       /*
+                        * EIO could indicate that (posix open) operation is not
+                        * supported, despite what server claimed in capability
+                        * negotiation.
+                        *
+                        * POSIX open in samba versions 3.3.1 and earlier could
+                        * incorrectly fail with invalid parameter.
+                        */
+                       tcon->broken_posix_open = true;
+                       break;
+
+               case -EREMOTE:
+               case -EOPNOTSUPP:
+                       /*
+                        * EREMOTE indicates DFS junction, which is not handled
+                        * in posix open.  If either that or op not supported
+                        * returned, follow the normal lookup.
+                        */
+                       break;
+
+               default:
+                       goto out;
+               }
+               /*
+                * fallthrough to retry, using older open call, this is case
+                * where server does not support this SMB level, and falsely
+                * claims capability (also get here for DFS case which should be
+                * rare for path not covered on files)
+                */
+       }
+
+       desired_access = 0;
+       if (OPEN_FMODE(oflags) & FMODE_READ)
+               desired_access |= GENERIC_READ; /* is this too little? */
+       if (OPEN_FMODE(oflags) & FMODE_WRITE)
+               desired_access |= GENERIC_WRITE;
+
+       disposition = FILE_OVERWRITE_IF;
+       if ((oflags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
+               disposition = FILE_CREATE;
+       else if ((oflags & (O_CREAT | O_TRUNC)) == (O_CREAT | O_TRUNC))
+               disposition = FILE_OVERWRITE_IF;
+       else if ((oflags & O_CREAT) == O_CREAT)
+               disposition = FILE_OPEN_IF;
+       else
+               cifs_dbg(FYI, "Create flag not set in create function\n");
+
+       /*
+        * BB add processing to set equivalent of mode - e.g. via CreateX with
+        * ACLs
+        */
+
+       if (!server->ops->open) {
+               rc = -ENOSYS;
+               goto out;
+       }
+
+       buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
+       if (buf == NULL) {
+               rc = -ENOMEM;
+               goto out;
+       }
+
+       /*
+        * if we're not using unix extensions, see if we need to set
+        * ATTR_READONLY on the create call
+        */
+       if (!tcon->unix_ext && (mode & S_IWUGO) == 0)
+               create_options |= CREATE_OPTION_READONLY;
+
+       if (backup_cred(cifs_sb))
+               create_options |= CREATE_OPEN_BACKUP_INTENT;
+
+       oparms.tcon = tcon;
+       oparms.cifs_sb = cifs_sb;
+       oparms.desired_access = desired_access;
+       oparms.create_options = create_options;
+       oparms.disposition = disposition;
+       oparms.path = full_path;
+       oparms.fid = fid;
+       oparms.reconnect = false;
+
+       rc = server->ops->open(xid, &oparms, oplock, buf);
+       if (rc) {
+               cifs_dbg(FYI, "cifs_create returned 0x%x\n", rc);
+               goto out;
+       }
+
+       /*
+        * If Open reported that we actually created a file then we now have to
+        * set the mode if possible.
+        */
+       if ((tcon->unix_ext) && (*oplock & CIFS_CREATE_ACTION)) {
+               struct cifs_unix_set_info_args args = {
+                               .mode   = mode,
+                               .ctime  = NO_CHANGE_64,
+                               .atime  = NO_CHANGE_64,
+                               .mtime  = NO_CHANGE_64,
+                               .device = 0,
+               };
+
+               if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) {
+                       args.uid = current_fsuid();
+                       if (inode->i_mode & S_ISGID)
+                               args.gid = inode->i_gid;
+                       else
+                               args.gid = current_fsgid();
+               } else {
+                       args.uid = INVALID_UID; /* no change */
+                       args.gid = INVALID_GID; /* no change */
+               }
+               CIFSSMBUnixSetFileInfo(xid, tcon, &args, fid->netfid,
+                                      current->tgid);
+       } else {
+               /*
+                * BB implement mode setting via Windows security
+                * descriptors e.g.
+                */
+               /* CIFSSMBWinSetPerms(xid,tcon,path,mode,-1,-1,nls);*/
+
+               /* Could set r/o dos attribute if mode & 0222 == 0 */
+       }
+
+cifs_create_get_file_info:
+       /* server might mask mode so we have to query for it */
+       if (tcon->unix_ext)
+               rc = cifs_get_inode_info_unix(&newinode, full_path, inode->i_sb,
+                                             xid);
+       else {
+               rc = cifs_get_inode_info(&newinode, full_path, buf, inode->i_sb,
+                                        xid, fid);
+               if (newinode) {
+                       if (server->ops->set_lease_key)
+                               server->ops->set_lease_key(newinode, fid);
+                       if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DYNPERM)
+                               newinode->i_mode = mode;
+                       if ((*oplock & CIFS_CREATE_ACTION) &&
+                           (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID)) {
+                               newinode->i_uid = current_fsuid();
+                               if (inode->i_mode & S_ISGID)
+                                       newinode->i_gid = inode->i_gid;
+                               else
+                                       newinode->i_gid = current_fsgid();
+                       }
+               }
+       }
+
+cifs_create_set_dentry:
+       if (rc != 0) {
+               cifs_dbg(FYI, "Create worked, get_inode_info failed rc = %d\n",
+                        rc);
+               if (server->ops->close)
+                       server->ops->close(xid, tcon, fid);
+               goto out;
+       }
+       d_drop(direntry);
+       d_add(direntry, newinode);
+
+out:
+       kfree(buf);
+       kfree(full_path);
+       return rc;
+}
+
+int
+cifs_atomic_open(struct inode *inode, struct dentry *direntry,
+                struct file *file, unsigned oflags, umode_t mode,
+                int *opened)
+{
+       int rc;
+       unsigned int xid;
+       struct tcon_link *tlink;
+       struct cifs_tcon *tcon;
+       struct TCP_Server_Info *server;
+       struct cifs_fid fid;
+       struct cifs_pending_open open;
+       __u32 oplock;
+       struct cifsFileInfo *file_info;
+
+       /*
+        * Posix open is only called (at lookup time) for file create now. For
+        * opens (rather than creates), because we do not know if it is a file
+        * or directory yet, and current Samba no longer allows us to do posix
+        * open on dirs, we could end up wasting an open call on what turns out
+        * to be a dir. For file opens, we wait to call posix open till
+        * cifs_open.  It could be added to atomic_open in the future but the
+        * performance tradeoff of the extra network request when EISDIR or
+        * EACCES is returned would have to be weighed against the 50% reduction
+        * in network traffic in the other paths.
+        */
+       if (!(oflags & O_CREAT)) {
+               struct dentry *res;
+
+               /*
+                * Check for hashed negative dentry. We have already revalidated
+                * the dentry and it is fine. No need to perform another lookup.
+                */
+               if (!d_unhashed(direntry))
+                       return -ENOENT;
+
+               res = cifs_lookup(inode, direntry, 0);
+               if (IS_ERR(res))
+                       return PTR_ERR(res);
+
+               return finish_no_open(file, res);
+       }
+
+       rc = check_name(direntry);
+       if (rc)
+               return rc;
+
+       xid = get_xid();
+
+       cifs_dbg(FYI, "parent inode = 0x%p name is: %pd and dentry = 0x%p\n",
+                inode, direntry, direntry);
+
+       tlink = cifs_sb_tlink(CIFS_SB(inode->i_sb));
+       if (IS_ERR(tlink)) {
+               rc = PTR_ERR(tlink);
+               goto out_free_xid;
+       }
+
+       tcon = tlink_tcon(tlink);
+       server = tcon->ses->server;
+
+       if (server->ops->new_lease_key)
+               server->ops->new_lease_key(&fid);
+
+       cifs_add_pending_open(&fid, tlink, &open);
+
+       rc = cifs_do_create(inode, direntry, xid, tlink, oflags, mode,
+                           &oplock, &fid);
+
+       if (rc) {
+               cifs_del_pending_open(&open);
+               goto out;
+       }
+
+       if ((oflags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
+               *opened |= FILE_CREATED;
+
+       rc = finish_open(file, direntry, generic_file_open, opened);
+       if (rc) {
+               if (server->ops->close)
+                       server->ops->close(xid, tcon, &fid);
+               cifs_del_pending_open(&open);
+               goto out;
+       }
+
+       if (file->f_flags & O_DIRECT &&
+           CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO) {
+               if (CIFS_SB(inode->i_sb)->mnt_cifs_flags & CIFS_MOUNT_NO_BRL)
+                       file->f_op = &cifs_file_direct_nobrl_ops;
+               else
+                       file->f_op = &cifs_file_direct_ops;
+               }
+
+       file_info = cifs_new_fileinfo(&fid, file, tlink, oplock);
+       if (file_info == NULL) {
+               if (server->ops->close)
+                       server->ops->close(xid, tcon, &fid);
+               cifs_del_pending_open(&open);
+               fput(file);
+               rc = -ENOMEM;
+       }
+
+out:
+       cifs_put_tlink(tlink);
+out_free_xid:
+       free_xid(xid);
+       return rc;
+}
+
+int cifs_create(struct inode *inode, struct dentry *direntry, umode_t mode,
+               bool excl)
+{
+       int rc;
+       unsigned int xid = get_xid();
+       /*
+        * BB below access is probably too much for mknod to request
+        *    but we have to do query and setpathinfo so requesting
+        *    less could fail (unless we want to request getatr and setatr
+        *    permissions (only).  At least for POSIX we do not have to
+        *    request so much.
+        */
+       unsigned oflags = O_EXCL | O_CREAT | O_RDWR;
+       struct tcon_link *tlink;
+       struct cifs_tcon *tcon;
+       struct TCP_Server_Info *server;
+       struct cifs_fid fid;
+       __u32 oplock;
+
+       cifs_dbg(FYI, "cifs_create parent inode = 0x%p name is: %pd and dentry = 0x%p\n",
+                inode, direntry, direntry);
+
+       tlink = cifs_sb_tlink(CIFS_SB(inode->i_sb));
+       rc = PTR_ERR(tlink);
+       if (IS_ERR(tlink))
+               goto out_free_xid;
+
+       tcon = tlink_tcon(tlink);
+       server = tcon->ses->server;
+
+       if (server->ops->new_lease_key)
+               server->ops->new_lease_key(&fid);
+
+       rc = cifs_do_create(inode, direntry, xid, tlink, oflags, mode,
+                           &oplock, &fid);
+       if (!rc && server->ops->close)
+               server->ops->close(xid, tcon, &fid);
+
+       cifs_put_tlink(tlink);
+out_free_xid:
+       free_xid(xid);
+       return rc;
+}
+
+int cifs_mknod(struct inode *inode, struct dentry *direntry, umode_t mode,
+               dev_t device_number)
+{
+       int rc = -EPERM;
+       unsigned int xid;
+       int create_options = CREATE_NOT_DIR | CREATE_OPTION_SPECIAL;
+       struct cifs_sb_info *cifs_sb;
+       struct tcon_link *tlink;
+       struct cifs_tcon *tcon;
+       struct cifs_io_parms io_parms;
+       char *full_path = NULL;
+       struct inode *newinode = NULL;
+       __u32 oplock = 0;
+       struct cifs_fid fid;
+       struct cifs_open_parms oparms;
+       FILE_ALL_INFO *buf = NULL;
+       unsigned int bytes_written;
+       struct win_dev *pdev;
+       struct kvec iov[2];
+
+       if (!old_valid_dev(device_number))
+               return -EINVAL;
+
+       cifs_sb = CIFS_SB(inode->i_sb);
+       tlink = cifs_sb_tlink(cifs_sb);
+       if (IS_ERR(tlink))
+               return PTR_ERR(tlink);
+
+       tcon = tlink_tcon(tlink);
+
+       xid = get_xid();
+
+       full_path = build_path_from_dentry(direntry);
+       if (full_path == NULL) {
+               rc = -ENOMEM;
+               goto mknod_out;
+       }
+
+       if (tcon->unix_ext) {
+               struct cifs_unix_set_info_args args = {
+                       .mode   = mode & ~current_umask(),
+                       .ctime  = NO_CHANGE_64,
+                       .atime  = NO_CHANGE_64,
+                       .mtime  = NO_CHANGE_64,
+                       .device = device_number,
+               };
+               if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) {
+                       args.uid = current_fsuid();
+                       args.gid = current_fsgid();
+               } else {
+                       args.uid = INVALID_UID; /* no change */
+                       args.gid = INVALID_GID; /* no change */
+               }
+               rc = CIFSSMBUnixSetPathInfo(xid, tcon, full_path, &args,
+                                           cifs_sb->local_nls,
+                                           cifs_remap(cifs_sb));
+               if (rc)
+                       goto mknod_out;
+
+               rc = cifs_get_inode_info_unix(&newinode, full_path,
+                                               inode->i_sb, xid);
+
+               if (rc == 0)
+                       d_instantiate(direntry, newinode);
+               goto mknod_out;
+       }
+
+       if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_UNX_EMUL))
+               goto mknod_out;
+
+
+       cifs_dbg(FYI, "sfu compat create special file\n");
+
+       buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
+       if (buf == NULL) {
+               kfree(full_path);
+               rc = -ENOMEM;
+               free_xid(xid);
+               return rc;
+       }
+
+       if (backup_cred(cifs_sb))
+               create_options |= CREATE_OPEN_BACKUP_INTENT;
+
+       oparms.tcon = tcon;
+       oparms.cifs_sb = cifs_sb;
+       oparms.desired_access = GENERIC_WRITE;
+       oparms.create_options = create_options;
+       oparms.disposition = FILE_CREATE;
+       oparms.path = full_path;
+       oparms.fid = &fid;
+       oparms.reconnect = false;
+
+       if (tcon->ses->server->oplocks)
+               oplock = REQ_OPLOCK;
+       else
+               oplock = 0;
+       rc = tcon->ses->server->ops->open(xid, &oparms, &oplock, buf);
+       if (rc)
+               goto mknod_out;
+
+       /*
+        * BB Do not bother to decode buf since no local inode yet to put
+        * timestamps in, but we can reuse it safely.
+        */
+
+       pdev = (struct win_dev *)buf;
+       io_parms.pid = current->tgid;
+       io_parms.tcon = tcon;
+       io_parms.offset = 0;
+       io_parms.length = sizeof(struct win_dev);
+       iov[1].iov_base = buf;
+       iov[1].iov_len = sizeof(struct win_dev);
+       if (S_ISCHR(mode)) {
+               memcpy(pdev->type, "IntxCHR", 8);
+               pdev->major = cpu_to_le64(MAJOR(device_number));
+               pdev->minor = cpu_to_le64(MINOR(device_number));
+               rc = tcon->ses->server->ops->sync_write(xid, &fid, &io_parms,
+                                                       &bytes_written, iov, 1);
+       } else if (S_ISBLK(mode)) {
+               memcpy(pdev->type, "IntxBLK", 8);
+               pdev->major = cpu_to_le64(MAJOR(device_number));
+               pdev->minor = cpu_to_le64(MINOR(device_number));
+               rc = tcon->ses->server->ops->sync_write(xid, &fid, &io_parms,
+                                                       &bytes_written, iov, 1);
+       } /* else if (S_ISFIFO) */
+       tcon->ses->server->ops->close(xid, tcon, &fid);
+       d_drop(direntry);
+
+       /* FIXME: add code here to set EAs */
+
+mknod_out:
+       kfree(full_path);
+       kfree(buf);
+       free_xid(xid);
+       cifs_put_tlink(tlink);
+       return rc;
+}
+
+struct dentry *
+cifs_lookup(struct inode *parent_dir_inode, struct dentry *direntry,
+           unsigned int flags)
+{
+       unsigned int xid;
+       int rc = 0; /* to get around spurious gcc warning, set to zero here */
+       struct cifs_sb_info *cifs_sb;
+       struct tcon_link *tlink;
+       struct cifs_tcon *pTcon;
+       struct inode *newInode = NULL;
+       char *full_path = NULL;
+
+       xid = get_xid();
+
+       cifs_dbg(FYI, "parent inode = 0x%p name is: %pd and dentry = 0x%p\n",
+                parent_dir_inode, direntry, direntry);
+
+       /* check whether path exists */
+
+       cifs_sb = CIFS_SB(parent_dir_inode->i_sb);
+       tlink = cifs_sb_tlink(cifs_sb);
+       if (IS_ERR(tlink)) {
+               free_xid(xid);
+               return (struct dentry *)tlink;
+       }
+       pTcon = tlink_tcon(tlink);
+
+       rc = check_name(direntry);
+       if (rc)
+               goto lookup_out;
+
+       /* can not grab the rename sem here since it would
+       deadlock in the cases (beginning of sys_rename itself)
+       in which we already have the sb rename sem */
+       full_path = build_path_from_dentry(direntry);
+       if (full_path == NULL) {
+               rc = -ENOMEM;
+               goto lookup_out;
+       }
+
+       if (d_really_is_positive(direntry)) {
+               cifs_dbg(FYI, "non-NULL inode in lookup\n");
+       } else {
+               cifs_dbg(FYI, "NULL inode in lookup\n");
+       }
+       cifs_dbg(FYI, "Full path: %s inode = 0x%p\n",
+                full_path, d_inode(direntry));
+
+       if (pTcon->unix_ext) {
+               rc = cifs_get_inode_info_unix(&newInode, full_path,
+                                             parent_dir_inode->i_sb, xid);
+       } else {
+               rc = cifs_get_inode_info(&newInode, full_path, NULL,
+                               parent_dir_inode->i_sb, xid, NULL);
+       }
+
+       if ((rc == 0) && (newInode != NULL)) {
+               d_add(direntry, newInode);
+               /* since paths are not looked up by component - the parent
+                  directories are presumed to be good here */
+               renew_parental_timestamps(direntry);
+
+       } else if (rc == -ENOENT) {
+               rc = 0;
+               direntry->d_time = jiffies;
+               d_add(direntry, NULL);
+       /*      if it was once a directory (but how can we tell?) we could do
+               shrink_dcache_parent(direntry); */
+       } else if (rc != -EACCES) {
+               cifs_dbg(FYI, "Unexpected lookup error %d\n", rc);
+               /* We special case check for Access Denied - since that
+               is a common return code */
+       }
+
+lookup_out:
+       kfree(full_path);
+       cifs_put_tlink(tlink);
+       free_xid(xid);
+       return ERR_PTR(rc);
+}
+
+static int
+cifs_d_revalidate(struct dentry *direntry, unsigned int flags)
+{
+       if (flags & LOOKUP_RCU)
+               return -ECHILD;
+
+       if (d_really_is_positive(direntry)) {
+               if (cifs_revalidate_dentry(direntry))
+                       return 0;
+               else {
+                       /*
+                        * If the inode wasn't known to be a dfs entry when
+                        * the dentry was instantiated, such as when created
+                        * via ->readdir(), it needs to be set now since the
+                        * attributes will have been updated by
+                        * cifs_revalidate_dentry().
+                        */
+                       if (IS_AUTOMOUNT(d_inode(direntry)) &&
+                          !(direntry->d_flags & DCACHE_NEED_AUTOMOUNT)) {
+                               spin_lock(&direntry->d_lock);
+                               direntry->d_flags |= DCACHE_NEED_AUTOMOUNT;
+                               spin_unlock(&direntry->d_lock);
+                       }
+
+                       return 1;
+               }
+       }
+
+       /*
+        * This may be nfsd (or something), anyway, we can't see the
+        * intent of this. So, since this can be for creation, drop it.
+        */
+       if (!flags)
+               return 0;
+
+       /*
+        * Drop the negative dentry, in order to make sure to use the
+        * case sensitive name which is specified by user if this is
+        * for creation.
+        */
+       if (flags & (LOOKUP_CREATE | LOOKUP_RENAME_TARGET))
+               return 0;
+
+       if (time_after(jiffies, direntry->d_time + HZ) || !lookupCacheEnabled)
+               return 0;
+
+       return 1;
+}
+
+/* static int cifs_d_delete(struct dentry *direntry)
+{
+       int rc = 0;
+
+       cifs_dbg(FYI, "In cifs d_delete, name = %pd\n", direntry);
+
+       return rc;
+}     */
+
+const struct dentry_operations cifs_dentry_ops = {
+       .d_revalidate = cifs_d_revalidate,
+       .d_automount = cifs_dfs_d_automount,
+/* d_delete:       cifs_d_delete,      */ /* not needed except for debugging */
+};
+
+static int cifs_ci_hash(const struct dentry *dentry, struct qstr *q)
+{
+       struct nls_table *codepage = CIFS_SB(dentry->d_sb)->local_nls;
+       unsigned long hash;
+       wchar_t c;
+       int i, charlen;
+
+       hash = init_name_hash();
+       for (i = 0; i < q->len; i += charlen) {
+               charlen = codepage->char2uni(&q->name[i], q->len - i, &c);
+               /* error out if we can't convert the character */
+               if (unlikely(charlen < 0))
+                       return charlen;
+               hash = partial_name_hash(cifs_toupper(c), hash);
+       }
+       q->hash = end_name_hash(hash);
+
+       return 0;
+}
+
+static int cifs_ci_compare(const struct dentry *parent, const struct dentry *dentry,
+               unsigned int len, const char *str, const struct qstr *name)
+{
+       struct nls_table *codepage = CIFS_SB(parent->d_sb)->local_nls;
+       wchar_t c1, c2;
+       int i, l1, l2;
+
+       /*
+        * We make the assumption here that uppercase characters in the local
+        * codepage are always the same length as their lowercase counterparts.
+        *
+        * If that's ever not the case, then this will fail to match it.
+        */
+       if (name->len != len)
+               return 1;
+
+       for (i = 0; i < len; i += l1) {
+               /* Convert characters in both strings to UTF-16. */
+               l1 = codepage->char2uni(&str[i], len - i, &c1);
+               l2 = codepage->char2uni(&name->name[i], name->len - i, &c2);
+
+               /*
+                * If we can't convert either character, just declare it to
+                * be 1 byte long and compare the original byte.
+                */
+               if (unlikely(l1 < 0 && l2 < 0)) {
+                       if (str[i] != name->name[i])
+                               return 1;
+                       l1 = 1;
+                       continue;
+               }
+
+               /*
+                * Here, we again ass|u|me that upper/lowercase versions of
+                * a character are the same length in the local NLS.
+                */
+               if (l1 != l2)
+                       return 1;
+
+               /* Now compare uppercase versions of these characters */
+               if (cifs_toupper(c1) != cifs_toupper(c2))
+                       return 1;
+       }
+
+       return 0;
+}
+
+const struct dentry_operations cifs_ci_dentry_ops = {
+       .d_revalidate = cifs_d_revalidate,
+       .d_hash = cifs_ci_hash,
+       .d_compare = cifs_ci_compare,
+       .d_automount = cifs_dfs_d_automount,
+};