Merge branch 'tracing/ftrace'; commit 'v2.6.29-rc6' into tracing/core

12 files changed, 538 insertions, 194 deletions

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 3e8023efaff7..872f104576e5 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1222,7 +1222,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
         /*
          * ok we haven't committed the transaction yet, lets do a commit
          */
-        if (file->private_data)
+        if (file && file->private_data)
                 btrfs_ioctl_trans_end(file);
 
         trans = btrfs_start_transaction(root, 1);
@@ -1231,7 +1231,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
                 goto out;
         }
 
-        ret = btrfs_log_dentry_safe(trans, root, file->f_dentry);
+        ret = btrfs_log_dentry_safe(trans, root, dentry);
         if (ret < 0)
                 goto out;
 
@@ -1245,7 +1245,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
          * file again, but that will end up using the synchronization
          * inside btrfs_sync_log to keep things safe.
          */
-        mutex_unlock(&file->f_dentry->d_inode->i_mutex);
+        mutex_unlock(&dentry->d_inode->i_mutex);
 
         if (ret > 0) {
                 ret = btrfs_commit_transaction(trans, root);
@@ -1253,7 +1253,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
                 btrfs_sync_log(trans, root);
                 ret = btrfs_end_transaction(trans, root);
         }
-        mutex_lock(&file->f_dentry->d_inode->i_mutex);
+        mutex_lock(&dentry->d_inode->i_mutex);
 out:
         return ret > 0 ? EIO : ret;
 }
diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES
index 73ac7ebd1dfc..851388fafc73 100644
--- a/fs/cifs/CHANGES
+++ b/fs/cifs/CHANGES
@@ -1,3 +1,13 @@
+Version 1.57
+------------
+Improve support for multiple security contexts to the same server. We
+used to use the same "vcnumber" for all connections which could cause
+the server to treat subsequent connections, especially those that
+are authenticated as guest, as reconnections, invalidating the earlier
+user's smb session.  This fix allows cifs to mount multiple times to the
+same server with different userids without risking invalidating earlier
+established security contexts.
+
 Version 1.56
 ------------
 Add "forcemandatorylock" mount option to allow user to use mandatory
@@ -7,7 +17,10 @@ specified and user does not have access to query information about the
 top of the share.  Fix problem in 2.6.28 resolving DFS paths to
 Samba servers (worked to Windows).  Fix rmdir so that pending search
 (readdir) requests do not get invalid results which include the now
-removed directory.
+removed directory.  Fix oops in cifs_dfs_ref.c when prefixpath is not reachable
+when using DFS.  Add better file create support to servers which support
+the CIFS POSIX protocol extensions (this adds support for new flags
+on create, and improves semantics for write of locked ranges).
 
 Version 1.55
 ------------
diff --git a/fs/cifs/cifsfs.h b/fs/cifs/cifsfs.h
index 7ac481841f87..2b1d28a9ee28 100644
--- a/fs/cifs/cifsfs.h
+++ b/fs/cifs/cifsfs.h
@@ -100,5 +100,5 @@ extern long cifs_ioctl(struct file *filep, unsigned int cmd, unsigned long arg);
 extern const struct export_operations cifs_export_ops;
 #endif /* EXPERIMENTAL */
 
-#define CIFS_VERSION   "1.56"
+#define CIFS_VERSION   "1.57"
 #endif                          /* _CIFSFS_H */
diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
index 94c1ca0ec953..e004f6db5fc8 100644
--- a/fs/cifs/cifsglob.h
+++ b/fs/cifs/cifsglob.h
@@ -164,9 +164,12 @@ struct TCP_Server_Info {
         /* multiplexed reads or writes */
         unsigned int maxBuf;    /* maxBuf specifies the maximum */
         /* message size the server can send or receive for non-raw SMBs */
-        unsigned int maxRw;     /* maxRw specifies the maximum */
+        unsigned int max_rw;    /* maxRw specifies the maximum */
         /* message size the server can send or receive for */
         /* SMB_COM_WRITE_RAW or SMB_COM_READ_RAW. */
+        unsigned int max_vcs;   /* maximum number of smb sessions, at least
+                                   those that can be specified uniquely with
+                                   vcnumbers */
         char sessid[4];         /* unique token id for this session */
         /* (returned on Negotiate */
         int capabilities; /* allow selective disabling of caps by smb sess */
@@ -210,6 +213,7 @@ struct cifsSesInfo {
         unsigned overrideSecFlg;  /* if non-zero override global sec flags */
         __u16 ipc_tid;          /* special tid for connection to IPC share */
         __u16 flags;
+        __u16 vcnum;
         char *serverOS;         /* name of operating system underlying server */
         char *serverNOS;        /* name of network operating system of server */
         char *serverDomain;     /* security realm of server */
diff --git a/fs/cifs/cifsproto.h b/fs/cifs/cifsproto.h
index 382ba6298809..083dfc57c7a3 100644
--- a/fs/cifs/cifsproto.h
+++ b/fs/cifs/cifsproto.h
@@ -42,6 +42,7 @@ extern void _FreeXid(unsigned int);
 #define GetXid() (int)_GetXid(); cFYI(1,("CIFS VFS: in %s as Xid: %d with uid: %d",__func__, xid,current_fsuid()));
 #define FreeXid(curr_xid) {_FreeXid(curr_xid); cFYI(1,("CIFS VFS: leaving %s (xid = %d) rc = %d",__func__,curr_xid,(int)rc));}
 extern char *build_path_from_dentry(struct dentry *);
+extern char *cifs_build_path_to_root(struct cifs_sb_info *cifs_sb);
 extern char *build_wildcard_path_from_dentry(struct dentry *direntry);
 /* extern void renew_parental_timestamps(struct dentry *direntry);*/
 extern int SendReceive(const unsigned int /* xid */ , struct cifsSesInfo *,
@@ -91,6 +92,9 @@ extern u64 cifs_UnixTimeToNT(struct timespec);
 extern __le64 cnvrtDosCifsTm(__u16 date, __u16 time);
 extern struct timespec cnvrtDosUnixTm(__u16 date, __u16 time);
 
+extern void posix_fill_in_inode(struct inode *tmp_inode,
+                                FILE_UNIX_BASIC_INFO *pData, int isNewInode);
+extern struct inode *cifs_new_inode(struct super_block *sb, __u64 *inum);
 extern int cifs_get_inode_info(struct inode **pinode,
                         const unsigned char *search_path,
                         FILE_ALL_INFO *pfile_info,
diff --git a/fs/cifs/cifssmb.c b/fs/cifs/cifssmb.c
index 552642a507c4..939e2f76b959 100644
--- a/fs/cifs/cifssmb.c
+++ b/fs/cifs/cifssmb.c
@@ -528,14 +528,15 @@ CIFSSMBNegotiate(unsigned int xid, struct cifsSesInfo *ses)
                 server->maxReq = le16_to_cpu(rsp->MaxMpxCount);
                 server->maxBuf = min((__u32)le16_to_cpu(rsp->MaxBufSize),
                                 (__u32)CIFSMaxBufSize + MAX_CIFS_HDR_SIZE);
+                server->max_vcs = le16_to_cpu(rsp->MaxNumberVcs);
                 GETU32(server->sessid) = le32_to_cpu(rsp->SessionKey);
                 /* even though we do not use raw we might as well set this
                 accurately, in case we ever find a need for it */
                 if ((le16_to_cpu(rsp->RawMode) & RAW_ENABLE) == RAW_ENABLE) {
-                        server->maxRw = 0xFF00;
+                        server->max_rw = 0xFF00;
                         server->capabilities = CAP_MPX_MODE | CAP_RAW_MODE;
                 } else {
-                        server->maxRw = 0;/* we do not need to use raw anyway */
+                        server->max_rw = 0;/* do not need to use raw anyway */
                         server->capabilities = CAP_MPX_MODE;
                 }
                 tmp = (__s16)le16_to_cpu(rsp->ServerTimeZone);
@@ -638,7 +639,7 @@ CIFSSMBNegotiate(unsigned int xid, struct cifsSesInfo *ses)
         /* probably no need to store and check maxvcs */
         server->maxBuf = min(le32_to_cpu(pSMBr->MaxBufferSize),
                         (__u32) CIFSMaxBufSize + MAX_CIFS_HDR_SIZE);
-        server->maxRw = le32_to_cpu(pSMBr->MaxRawSize);
+        server->max_rw = le32_to_cpu(pSMBr->MaxRawSize);
         cFYI(DBG2, ("Max buf = %d", ses->server->maxBuf));
         GETU32(ses->server->sessid) = le32_to_cpu(pSMBr->SessionKey);
         server->capabilities = le32_to_cpu(pSMBr->Capabilities);
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 2209be943051..da0f4ffa0613 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -23,7 +23,6 @@
 #include <linux/string.h>
 #include <linux/list.h>
 #include <linux/wait.h>
-#include <linux/ipv6.h>
 #include <linux/pagemap.h>
 #include <linux/ctype.h>
 #include <linux/utsname.h>
@@ -35,6 +34,7 @@
 #include <linux/freezer.h>
 #include <asm/uaccess.h>
 #include <asm/processor.h>
+#include <net/ipv6.h>
 #include "cifspdu.h"
 #include "cifsglob.h"
 #include "cifsproto.h"
@@ -1379,8 +1379,8 @@ cifs_find_tcp_session(struct sockaddr_storage *addr)
                      server->addr.sockAddr.sin_addr.s_addr))
                         continue;
                 else if (addr->ss_family == AF_INET6 &&
-                         memcmp(&server->addr.sockAddr6.sin6_addr,
-                                &addr6->sin6_addr, sizeof(addr6->sin6_addr)))
+                         !ipv6_addr_equal(&server->addr.sockAddr6.sin6_addr,
+                                          &addr6->sin6_addr))
                         continue;
 
                 ++server->srv_count;
@@ -2180,6 +2180,33 @@ static void setup_cifs_sb(struct smb_vol *pvolume_info,
                            "mount option supported"));
 }
 
+static int
+is_path_accessible(int xid, struct cifsTconInfo *tcon,
+                   struct cifs_sb_info *cifs_sb, const char *full_path)
+{
+        int rc;
+        __u64 inode_num;
+        FILE_ALL_INFO *pfile_info;
+
+        rc = CIFSGetSrvInodeNumber(xid, tcon, full_path, &inode_num,
+                                   cifs_sb->local_nls,
+                                   cifs_sb->mnt_cifs_flags &
+                                                CIFS_MOUNT_MAP_SPECIAL_CHR);
+        if (rc != -EOPNOTSUPP)
+                return rc;
+
+        pfile_info = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
+        if (pfile_info == NULL)
+                return -ENOMEM;
+
+        rc = CIFSSMBQPathInfo(xid, tcon, full_path, pfile_info,
+                              0 /* not legacy */, cifs_sb->local_nls,
+                              cifs_sb->mnt_cifs_flags &
+                                CIFS_MOUNT_MAP_SPECIAL_CHR);
+        kfree(pfile_info);
+        return rc;
+}
+
 int
 cifs_mount(struct super_block *sb, struct cifs_sb_info *cifs_sb,
            char *mount_data, const char *devname)
@@ -2190,6 +2217,7 @@ cifs_mount(struct super_block *sb, struct cifs_sb_info *cifs_sb,
         struct cifsSesInfo *pSesInfo = NULL;
         struct cifsTconInfo *tcon = NULL;
         struct TCP_Server_Info *srvTcp = NULL;
+        char   *full_path;
 
         xid = GetXid();
 
@@ -2426,6 +2454,23 @@ mount_fail_check:
                 cifs_sb->rsize = min(cifs_sb->rsize,
                                (tcon->ses->server->maxBuf - MAX_CIFS_HDR_SIZE));
 
+        if (!rc && cifs_sb->prepathlen) {
+                /* build_path_to_root works only when we have a valid tcon */
+                full_path = cifs_build_path_to_root(cifs_sb);
+                if (full_path == NULL) {
+                        rc = -ENOMEM;
+                        goto mount_fail_check;
+                }
+                rc = is_path_accessible(xid, tcon, cifs_sb, full_path);
+                if (rc) {
+                        cERROR(1, ("Path %s in not accessible: %d",
+                                                full_path, rc));
+                        kfree(full_path);
+                        goto mount_fail_check;
+                }
+                kfree(full_path);
+        }
+
         /* volume_info->password is freed above when existing session found
         (in which case it is not needed anymore) but when new sesion is created
         the password ptr is put in the new session structure (in which case the
diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index 964aad03c5ad..89fb72832652 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -3,7 +3,7 @@
  *
  *   vfs operations that deal with dentries
  *
- *   Copyright (C) International Business Machines  Corp., 2002,2008
+ *   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
@@ -129,6 +129,78 @@ cifs_bp_rename_retry:
         return full_path;
 }
 
+static int cifs_posix_open(char *full_path, struct inode **pinode,
+                    struct super_block *sb, int mode, int oflags,
+                    int *poplock, __u16 *pnetfid, int xid)
+{
+        int rc;
+        __u32 oplock;
+        FILE_UNIX_BASIC_INFO *presp_data;
+        __u32 posix_flags = 0;
+        struct cifs_sb_info *cifs_sb = CIFS_SB(sb);
+
+        cFYI(1, ("posix open %s", full_path));
+
+        presp_data = kzalloc(sizeof(FILE_UNIX_BASIC_INFO), GFP_KERNEL);
+        if (presp_data == NULL)
+                return -ENOMEM;
+
+/* So far cifs posix extensions can only map the following flags.
+   There are other valid fmode oflags such as FMODE_LSEEK, FMODE_PREAD, but
+   so far we do not seem to need them, and we can treat them as local only */
+        if ((oflags & (FMODE_READ | FMODE_WRITE)) ==
+                (FMODE_READ | FMODE_WRITE))
+                posix_flags = SMB_O_RDWR;
+        else if (oflags & FMODE_READ)
+                posix_flags = SMB_O_RDONLY;
+        else if (oflags & FMODE_WRITE)
+                posix_flags = SMB_O_WRONLY;
+        if (oflags & O_CREAT)
+                posix_flags |= SMB_O_CREAT;
+        if (oflags & O_EXCL)
+                posix_flags |= SMB_O_EXCL;
+        if (oflags & O_TRUNC)
+                posix_flags |= SMB_O_TRUNC;
+        if (oflags & O_APPEND)
+                posix_flags |= SMB_O_APPEND;
+        if (oflags & O_SYNC)
+                posix_flags |= SMB_O_SYNC;
+        if (oflags & O_DIRECTORY)
+                posix_flags |= SMB_O_DIRECTORY;
+        if (oflags & O_NOFOLLOW)
+                posix_flags |= SMB_O_NOFOLLOW;
+        if (oflags & O_DIRECT)
+                posix_flags |= SMB_O_DIRECT;
+
+
+        rc = CIFSPOSIXCreate(xid, cifs_sb->tcon, posix_flags, mode,
+                        pnetfid, presp_data, &oplock, full_path,
+                        cifs_sb->local_nls, cifs_sb->mnt_cifs_flags &
+                                        CIFS_MOUNT_MAP_SPECIAL_CHR);
+        if (rc)
+                goto posix_open_ret;
+
+        if (presp_data->Type == cpu_to_le32(-1))
+                goto posix_open_ret; /* open ok, caller does qpathinfo */
+
+        /* get new inode and set it up */
+        if (!pinode)
+                goto posix_open_ret; /* caller does not need info */
+
+        *pinode = cifs_new_inode(sb, &presp_data->UniqueId);
+
+        /* We do not need to close the file if new_inode fails since
+           the caller will retry qpathinfo as long as inode is null */
+        if (*pinode == NULL)
+                goto posix_open_ret;
+
+        posix_fill_in_inode(*pinode, presp_data, 1);
+
+posix_open_ret:
+        kfree(presp_data);
+        return rc;
+}
+
 static void setup_cifs_dentry(struct cifsTconInfo *tcon,
                               struct dentry *direntry,
                               struct inode *newinode)
@@ -150,7 +222,14 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode,
         int xid;
         int create_options = CREATE_NOT_DIR;
         int oplock = 0;
-        /* BB below access is too much for the mknod to request */
+        int oflags;
+        /*
+         * 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.
+         */
         int desiredAccess = GENERIC_READ | GENERIC_WRITE;
         __u16 fileHandle;
         struct cifs_sb_info *cifs_sb;
@@ -174,13 +253,43 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode,
         }
 
         mode &= ~current->fs->umask;
+        if (oplockEnabled)
+                oplock = REQ_OPLOCK;
 
-        if (nd && (nd->flags & LOOKUP_OPEN)) {
-                int oflags = nd->intent.open.flags;
+        if (nd && (nd->flags & LOOKUP_OPEN))
+                oflags = nd->intent.open.flags;
+        else
+                oflags = FMODE_READ;
+
+        if (tcon->unix_ext && (tcon->ses->capabilities & CAP_UNIX) &&
+            (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, &fileHandle, xid);
+                /* EIO could indicate that (posix open) operation is not
+                   supported, despite what server claimed in capability
+                   negotation.  EREMOTE indicates DFS junction, which is not
+                   handled in posix open */
+
+                if ((rc == 0) && (newinode == NULL))
+                        goto cifs_create_get_file_info; /* query inode info */
+                else if (rc == 0) /* success, no need to query */
+                        goto cifs_create_set_dentry;
+                else if ((rc != -EIO) && (rc != -EREMOTE) &&
+                         (rc != -EOPNOTSUPP)) /* path not found or net err */
+                        goto cifs_create_out;
+                /* else 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) */
+        }
 
+        if (nd && (nd->flags & LOOKUP_OPEN)) {
+                /* if the file is going to stay open, then we
+                   need to set the desired access properly */
                 desiredAccess = 0;
                 if (oflags & FMODE_READ)
-                        desiredAccess |= GENERIC_READ;
+                        desiredAccess |= GENERIC_READ; /* is this too little? */
                 if (oflags & FMODE_WRITE) {
                         desiredAccess |= GENERIC_WRITE;
                         if (!(oflags & FMODE_READ))
@@ -199,8 +308,6 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode,
 
         /* BB add processing to set equivalent of mode - e.g. via CreateX with
            ACLs */
-        if (oplockEnabled)
-                oplock = REQ_OPLOCK;
 
         buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL);
         if (buf == NULL) {
@@ -233,116 +340,112 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode,
         }
         if (rc) {
                 cFYI(1, ("cifs_create returned 0x%x", rc));
-        } else {
-                /* 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 = {
+                goto cifs_create_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 = (__u64) current_fsuid();
-                                if (inode->i_mode & S_ISGID)
-                                        args.gid = (__u64) inode->i_gid;
-                                else
-                                        args.gid = (__u64) current_fsgid();
-                        } else {
-                                args.uid = NO_CHANGE_64;
-                                args.gid = NO_CHANGE_64;
-                        }
-                        CIFSSMBUnixSetInfo(xid, tcon, full_path, &args,
-                                cifs_sb->local_nls,
-                                cifs_sb->mnt_cifs_flags &
-                                        CIFS_MOUNT_MAP_SPECIAL_CHR);
+                if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) {
+                        args.uid = (__u64) current_fsuid();
+                        if (inode->i_mode & S_ISGID)
+                                args.gid = (__u64) inode->i_gid;
+                        else
+                                args.gid = (__u64) current_fsgid();
                 } 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 */
+                        args.uid = NO_CHANGE_64;
+                        args.gid = NO_CHANGE_64;
                 }
+                CIFSSMBUnixSetInfo(xid, tcon, full_path, &args,
+                        cifs_sb->local_nls,
+                        cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR);
+        } else {
+                /* BB implement mode setting via Windows security
+                   descriptors e.g. */
+                /* CIFSSMBWinSetPerms(xid,tcon,path,mode,-1,-1,nls);*/
 
-                /* 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,
-                                                 &fileHandle);
-                        if (newinode) {
-                                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();
-                                }
+                /* 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, &fileHandle);
+                if (newinode) {
+                        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();
                         }
                 }
+        }
 
-                if (rc != 0) {
-                        cFYI(1, ("Create worked, get_inode_info failed rc = %d",
-                                 rc));
-                } else
-                        setup_cifs_dentry(tcon, direntry, newinode);
-
-                if ((nd == NULL /* nfsd case - nfs srv does not set nd */) ||
-                        (!(nd->flags & LOOKUP_OPEN))) {
-                        /* mknod case - do not leave file open */
-                        CIFSSMBClose(xid, tcon, fileHandle);
-                } else if (newinode) {
-                        struct cifsFileInfo *pCifsFile =
-                           kzalloc(sizeof(struct cifsFileInfo), GFP_KERNEL);
-
-                        if (pCifsFile == NULL)
-                                goto cifs_create_out;
-                        pCifsFile->netfid = fileHandle;
-                        pCifsFile->pid = current->tgid;
-                        pCifsFile->pInode = newinode;
-                        pCifsFile->invalidHandle = false;
-                        pCifsFile->closePend     = false;
-                        init_MUTEX(&pCifsFile->fh_sem);
-                        mutex_init(&pCifsFile->lock_mutex);
-                        INIT_LIST_HEAD(&pCifsFile->llist);
-                        atomic_set(&pCifsFile->wrtPending, 0);
-
-                        /* set the following in open now
+cifs_create_set_dentry:
+        if (rc == 0)
+                setup_cifs_dentry(tcon, direntry, newinode);
+        else
+                cFYI(1, ("Create worked, get_inode_info failed rc = %d", rc));
+
+        /* nfsd case - nfs srv does not set nd */
+        if ((nd == NULL) || (!(nd->flags & LOOKUP_OPEN))) {
+                /* mknod case - do not leave file open */
+                CIFSSMBClose(xid, tcon, fileHandle);
+        } else if (newinode) {
+                struct cifsFileInfo *pCifsFile =
+                        kzalloc(sizeof(struct cifsFileInfo), GFP_KERNEL);
+
+                if (pCifsFile == NULL)
+                        goto cifs_create_out;
+                pCifsFile->netfid = fileHandle;
+                pCifsFile->pid = current->tgid;
+                pCifsFile->pInode = newinode;
+                pCifsFile->invalidHandle = false;
+                pCifsFile->closePend     = false;
+                init_MUTEX(&pCifsFile->fh_sem);
+                mutex_init(&pCifsFile->lock_mutex);
+                INIT_LIST_HEAD(&pCifsFile->llist);
+                atomic_set(&pCifsFile->wrtPending, 0);
+
+                /* set the following in open now
                                 pCifsFile->pfile = file; */
-                        write_lock(&GlobalSMBSeslock);
-                        list_add(&pCifsFile->tlist, &tcon->openFileList);
-                        pCifsInode = CIFS_I(newinode);
-                        if (pCifsInode) {
-                                /* if readable file instance put first in list*/
-                                if (write_only) {
-                                        list_add_tail(&pCifsFile->flist,
-                                                &pCifsInode->openFileList);
-                                } else {
-                                        list_add(&pCifsFile->flist,
-                                                &pCifsInode->openFileList);
-                                }
-                                if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) {
-                                        pCifsInode->clientCanCacheAll = true;
-                                        pCifsInode->clientCanCacheRead = true;
-                                        cFYI(1, ("Exclusive Oplock inode %p",
-                                                newinode));
-                                } else if ((oplock & 0xF) == OPLOCK_READ)
-                                        pCifsInode->clientCanCacheRead = true;
+                write_lock(&GlobalSMBSeslock);
+                list_add(&pCifsFile->tlist, &tcon->openFileList);
+                pCifsInode = CIFS_I(newinode);
+                if (pCifsInode) {
+                        /* if readable file instance put first in list*/
+                        if (write_only) {
+                                list_add_tail(&pCifsFile->flist,
+                                              &pCifsInode->openFileList);
+                        } else {
+                                list_add(&pCifsFile->flist,
+                                         &pCifsInode->openFileList);
                         }
-                        write_unlock(&GlobalSMBSeslock);
+                        if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) {
+                                pCifsInode->clientCanCacheAll = true;
+                                pCifsInode->clientCanCacheRead = true;
+                                cFYI(1, ("Exclusive Oplock inode %p",
+                                        newinode));
+                        } else if ((oplock & 0xF) == OPLOCK_READ)
+                                pCifsInode->clientCanCacheRead = true;
                 }
+                write_unlock(&GlobalSMBSeslock);
         }
 cifs_create_out:
         kfree(buf);
diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index bcf7b5184664..4690a360c855 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -199,6 +199,49 @@ static void fill_fake_finddataunix(FILE_UNIX_BASIC_INFO *pfnd_dat,
         pfnd_dat->Gid = cpu_to_le64(pinode->i_gid);
 }
 
+/**
+ * cifs_new inode - create new inode, initialize, and hash it
+ * @sb - pointer to superblock
+ * @inum - if valid pointer and serverino is enabled, replace i_ino with val
+ *
+ * Create a new inode, initialize it for CIFS and hash it. Returns the new
+ * inode or NULL if one couldn't be allocated.
+ *
+ * If the share isn't mounted with "serverino" or inum is a NULL pointer then
+ * we'll just use the inode number assigned by new_inode(). Note that this can
+ * mean i_ino collisions since the i_ino assigned by new_inode is not
+ * guaranteed to be unique.
+ */
+struct inode *
+cifs_new_inode(struct super_block *sb, __u64 *inum)
+{
+        struct inode *inode;
+
+        inode = new_inode(sb);
+        if (inode == NULL)
+                return NULL;
+
+        /*
+         * BB: Is i_ino == 0 legal? Here, we assume that it is. If it isn't we
+         *     stop passing inum as ptr. Are there sanity checks we can use to
+         *     ensure that the server is really filling in that field? Also,
+         *     if serverino is disabled, perhaps we should be using iunique()?
+         */
+        if (inum && (CIFS_SB(sb)->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM))
+                inode->i_ino = (unsigned long) *inum;
+
+        /*
+         * must set this here instead of cifs_alloc_inode since VFS will
+         * clobber i_flags
+         */
+        if (sb->s_flags & MS_NOATIME)
+                inode->i_flags |= S_NOATIME | S_NOCMTIME;
+
+        insert_inode_hash(inode);
+
+        return inode;
+}
+
 int cifs_get_inode_info_unix(struct inode **pinode,
         const unsigned char *full_path, struct super_block *sb, int xid)
 {
@@ -233,22 +276,11 @@ int cifs_get_inode_info_unix(struct inode **pinode,
 
         /* get new inode */
         if (*pinode == NULL) {
-                *pinode = new_inode(sb);
+                *pinode = cifs_new_inode(sb, &find_data.UniqueId);
                 if (*pinode == NULL) {
                         rc = -ENOMEM;
                         goto cgiiu_exit;
                 }
-                /* Is an i_ino of zero legal? */
-                /* note ino incremented to unique num in new_inode */
-                /* Are there sanity checks we can use to ensure that
-                   the server is really filling in that field? */
-                if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM)
-                        (*pinode)->i_ino = (unsigned long)find_data.UniqueId;
-
-                if (sb->s_flags & MS_NOATIME)
-                        (*pinode)->i_flags |= S_NOATIME | S_NOCMTIME;
-
-                insert_inode_hash(*pinode);
         }
 
         inode = *pinode;
@@ -465,11 +497,9 @@ int cifs_get_inode_info(struct inode **pinode,
 
         /* get new inode */
         if (*pinode == NULL) {
-                *pinode = new_inode(sb);
-                if (*pinode == NULL) {
-                        rc = -ENOMEM;
-                        goto cgii_exit;
-                }
+                __u64 inode_num;
+                __u64 *pinum = &inode_num;
+
                 /* Is an i_ino of zero legal? Can we use that to check
                    if the server supports returning inode numbers?  Are
                    there other sanity checks we can use to ensure that
@@ -486,22 +516,26 @@ int cifs_get_inode_info(struct inode **pinode,
 
                 if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) {
                         int rc1 = 0;
-                        __u64 inode_num;
 
                         rc1 = CIFSGetSrvInodeNumber(xid, pTcon,
-                                        full_path, &inode_num,
+                                        full_path, pinum,
                                         cifs_sb->local_nls,
                                         cifs_sb->mnt_cifs_flags &
                                                 CIFS_MOUNT_MAP_SPECIAL_CHR);
                         if (rc1) {
                                 cFYI(1, ("GetSrvInodeNum rc %d", rc1));
+                                pinum = NULL;
                                 /* BB EOPNOSUPP disable SERVER_INUM? */
-                        } else /* do we need cast or hash to ino? */
-                                (*pinode)->i_ino = inode_num;
-                } /* else ino incremented to unique num in new_inode*/
-                if (sb->s_flags & MS_NOATIME)
-                        (*pinode)->i_flags |= S_NOATIME | S_NOCMTIME;
-                insert_inode_hash(*pinode);
+                        }
+                } else {
+                        pinum = NULL;
+                }
+
+                *pinode = cifs_new_inode(sb, pinum);
+                if (*pinode == NULL) {
+                        rc = -ENOMEM;
+                        goto cgii_exit;
+                }
         }
         inode = *pinode;
         cifsInfo = CIFS_I(inode);
@@ -621,7 +655,7 @@ static const struct inode_operations cifs_ipc_inode_ops = {
         .lookup = cifs_lookup,
 };
 
-static char *build_path_to_root(struct cifs_sb_info *cifs_sb)
+char *cifs_build_path_to_root(struct cifs_sb_info *cifs_sb)
 {
         int pplen = cifs_sb->prepathlen;
         int dfsplen;
@@ -678,7 +712,7 @@ struct inode *cifs_iget(struct super_block *sb, unsigned long ino)
                 return inode;
 
         cifs_sb = CIFS_SB(inode->i_sb);
-        full_path = build_path_to_root(cifs_sb);
+        full_path = cifs_build_path_to_root(cifs_sb);
         if (full_path == NULL)
                 return ERR_PTR(-ENOMEM);
 
@@ -1017,7 +1051,7 @@ out_reval:
         return rc;
 }
 
-static void posix_fill_in_inode(struct inode *tmp_inode,
+void posix_fill_in_inode(struct inode *tmp_inode,
         FILE_UNIX_BASIC_INFO *pData, int isNewInode)
 {
         struct cifsInodeInfo *cifsInfo = CIFS_I(tmp_inode);
@@ -1114,24 +1148,14 @@ int cifs_mkdir(struct inode *inode, struct dentry *direntry, int mode)
                         else
                                 direntry->d_op = &cifs_dentry_ops;
 
-                        newinode = new_inode(inode->i_sb);
+                        newinode = cifs_new_inode(inode->i_sb,
+                                                  &pInfo->UniqueId);
                         if (newinode == NULL) {
                                 kfree(pInfo);
                                 goto mkdir_get_info;
                         }
 
-                        /* Is an i_ino of zero legal? */
-                        /* Are there sanity checks we can use to ensure that
-                           the server is really filling in that field? */
-                        if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) {
-                                newinode->i_ino =
-                                        (unsigned long)pInfo->UniqueId;
-                        } /* note ino incremented to unique num in new_inode */
-                        if (inode->i_sb->s_flags & MS_NOATIME)
-                                newinode->i_flags |= S_NOATIME | S_NOCMTIME;
                         newinode->i_nlink = 2;
-
-                        insert_inode_hash(newinode);
                         d_instantiate(direntry, newinode);
 
                         /* we already checked in POSIXCreate whether
diff --git a/fs/cifs/readdir.c b/fs/cifs/readdir.c
index 9f51f9bf0292..c2c01ff4c32c 100644
--- a/fs/cifs/readdir.c
+++ b/fs/cifs/readdir.c
@@ -56,35 +56,34 @@ static inline void dump_cifs_file_struct(struct file *file, char *label)
 }
 #endif /* DEBUG2 */
 
-/* Returns one if new inode created (which therefore needs to be hashed) */
+/* Returns 1 if new inode created, 2 if both dentry and inode were */
 /* Might check in the future if inode number changed so we can rehash inode */
-static int construct_dentry(struct qstr *qstring, struct file *file,
-        struct inode **ptmp_inode, struct dentry **pnew_dentry)
+static int
+construct_dentry(struct qstr *qstring, struct file *file,
+                 struct inode **ptmp_inode, struct dentry **pnew_dentry,
+                 __u64 *inum)
 {
-        struct dentry *tmp_dentry;
-        struct cifs_sb_info *cifs_sb;
-        struct cifsTconInfo *pTcon;
+        struct dentry *tmp_dentry = NULL;
+        struct super_block *sb = file->f_path.dentry->d_sb;
         int rc = 0;
 
         cFYI(1, ("For %s", qstring->name));
-        cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
-        pTcon = cifs_sb->tcon;
 
         qstring->hash = full_name_hash(qstring->name, qstring->len);
         tmp_dentry = d_lookup(file->f_path.dentry, qstring);
         if (tmp_dentry) {
+                /* BB: overwrite old name? i.e. tmp_dentry->d_name and
+                 * tmp_dentry->d_name.len??
+                 */
                 cFYI(0, ("existing dentry with inode 0x%p",
                          tmp_dentry->d_inode));
                 *ptmp_inode = tmp_dentry->d_inode;
-/* BB overwrite old name? i.e. tmp_dentry->d_name and tmp_dentry->d_name.len??*/
                 if (*ptmp_inode == NULL) {
-                        *ptmp_inode = new_inode(file->f_path.dentry->d_sb);
+                        *ptmp_inode = cifs_new_inode(sb, inum);
                         if (*ptmp_inode == NULL)
                                 return rc;
                         rc = 1;
                 }
-                if (file->f_path.dentry->d_sb->s_flags & MS_NOATIME)
-                        (*ptmp_inode)->i_flags |= S_NOATIME | S_NOCMTIME;
         } else {
                 tmp_dentry = d_alloc(file->f_path.dentry, qstring);
                 if (tmp_dentry == NULL) {
@@ -93,15 +92,14 @@ static int construct_dentry(struct qstr *qstring, struct file *file,
                         return rc;
                 }
 
-                *ptmp_inode = new_inode(file->f_path.dentry->d_sb);
-                if (pTcon->nocase)
+                if (CIFS_SB(sb)->tcon->nocase)
                         tmp_dentry->d_op = &cifs_ci_dentry_ops;
                 else
                         tmp_dentry->d_op = &cifs_dentry_ops;
+
+                *ptmp_inode = cifs_new_inode(sb, inum);
                 if (*ptmp_inode == NULL)
                         return rc;
-                if (file->f_path.dentry->d_sb->s_flags & MS_NOATIME)
-                        (*ptmp_inode)->i_flags |= S_NOATIME | S_NOCMTIME;
                 rc = 2;
         }
 
@@ -822,7 +820,7 @@ static int find_cifs_entry(const int xid, struct cifsTconInfo *pTcon,
 /* inode num, inode type and filename returned */
 static int cifs_get_name_from_search_buf(struct qstr *pqst,
         char *current_entry, __u16 level, unsigned int unicode,
-        struct cifs_sb_info *cifs_sb, int max_len, ino_t *pinum)
+        struct cifs_sb_info *cifs_sb, int max_len, __u64 *pinum)
 {
         int rc = 0;
         unsigned int len = 0;
@@ -842,9 +840,7 @@ static int cifs_get_name_from_search_buf(struct qstr *pqst,
                         len = strnlen(filename, PATH_MAX);
                 }
 
-                /* BB fixme - hash low and high 32 bits if not 64 bit arch BB */
-                if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM)
-                        *pinum = pFindData->UniqueId;
+                *pinum = pFindData->UniqueId;
         } else if (level == SMB_FIND_FILE_DIRECTORY_INFO) {
                 FILE_DIRECTORY_INFO *pFindData =
                         (FILE_DIRECTORY_INFO *)current_entry;
@@ -907,7 +903,7 @@ static int cifs_filldir(char *pfindEntry, struct file *file,
         struct qstr qstring;
         struct cifsFileInfo *pCifsF;
         unsigned int obj_type;
-        ino_t  inum;
+        __u64  inum;
         struct cifs_sb_info *cifs_sb;
         struct inode *tmp_inode;
         struct dentry *tmp_dentry;
@@ -940,20 +936,18 @@ static int cifs_filldir(char *pfindEntry, struct file *file,
         if (rc)
                 return rc;
 
-        rc = construct_dentry(&qstring, file, &tmp_inode, &tmp_dentry);
+        /* only these two infolevels return valid inode numbers */
+        if (pCifsF->srch_inf.info_level == SMB_FIND_FILE_UNIX ||
+            pCifsF->srch_inf.info_level == SMB_FIND_FILE_ID_FULL_DIR_INFO)
+                rc = construct_dentry(&qstring, file, &tmp_inode, &tmp_dentry,
+                                        &inum);
+        else
+                rc = construct_dentry(&qstring, file, &tmp_inode, &tmp_dentry,
+                                        NULL);
+
         if ((tmp_inode == NULL) || (tmp_dentry == NULL))
                 return -ENOMEM;
 
-        if (rc) {
-                /* inode created, we need to hash it with right inode number */
-                if (inum != 0) {
-                        /* BB fixme - hash the 2 32 quantities bits together if
-                         *  necessary BB */
-                        tmp_inode->i_ino = inum;
-                }
-                insert_inode_hash(tmp_inode);
-        }
-
         /* we pass in rc below, indicating whether it is a new inode,
            so we can figure out whether to invalidate the inode cached
            data if the file has changed */
diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c
index 5f22de7b79a9..5c68b4282be9 100644
--- a/fs/cifs/sess.c
+++ b/fs/cifs/sess.c
@@ -34,15 +34,99 @@
 extern void SMBNTencrypt(unsigned char *passwd, unsigned char *c8,
                          unsigned char *p24);
 
+/* Checks if this is the first smb session to be reconnected after
+   the socket has been reestablished (so we know whether to use vc 0).
+   Called while holding the cifs_tcp_ses_lock, so do not block */
+static bool is_first_ses_reconnect(struct cifsSesInfo *ses)
+{
+        struct list_head *tmp;
+        struct cifsSesInfo *tmp_ses;
+
+        list_for_each(tmp, &ses->server->smb_ses_list) {
+                tmp_ses = list_entry(tmp, struct cifsSesInfo,
+                                     smb_ses_list);
+                if (tmp_ses->need_reconnect == false)
+                        return false;
+        }
+        /* could not find a session that was already connected,
+           this must be the first one we are reconnecting */
+        return true;
+}
+
+/*
+ *      vc number 0 is treated specially by some servers, and should be the
+ *      first one we request.  After that we can use vcnumbers up to maxvcs,
+ *      one for each smb session (some Windows versions set maxvcs incorrectly
+ *      so maxvc=1 can be ignored).  If we have too many vcs, we can reuse
+ *      any vc but zero (some servers reset the connection on vcnum zero)
+ *
+ */
+static __le16 get_next_vcnum(struct cifsSesInfo *ses)
+{
+        __u16 vcnum = 0;
+        struct list_head *tmp;
+        struct cifsSesInfo *tmp_ses;
+        __u16 max_vcs = ses->server->max_vcs;
+        __u16 i;
+        int free_vc_found = 0;
+
+        /* Quoting the MS-SMB specification: "Windows-based SMB servers set this
+        field to one but do not enforce this limit, which allows an SMB client
+        to establish more virtual circuits than allowed by this value ... but
+        other server implementations can enforce this limit." */
+        if (max_vcs < 2)
+                max_vcs = 0xFFFF;
+
+        write_lock(&cifs_tcp_ses_lock);
+        if ((ses->need_reconnect) && is_first_ses_reconnect(ses))
+                        goto get_vc_num_exit;  /* vcnum will be zero */
+        for (i = ses->server->srv_count - 1; i < max_vcs; i++) {
+                if (i == 0) /* this is the only connection, use vc 0 */
+                        break;
+
+                free_vc_found = 1;
+
+                list_for_each(tmp, &ses->server->smb_ses_list) {
+                        tmp_ses = list_entry(tmp, struct cifsSesInfo,
+                                             smb_ses_list);
+                        if (tmp_ses->vcnum == i) {
+                                free_vc_found = 0;
+                                break; /* found duplicate, try next vcnum */
+                        }
+                }
+                if (free_vc_found)
+                        break; /* we found a vcnumber that will work - use it */
+        }
+
+        if (i == 0)
+                vcnum = 0; /* for most common case, ie if one smb session, use
+                              vc zero.  Also for case when no free vcnum, zero
+                              is safest to send (some clients only send zero) */
+        else if (free_vc_found == 0)
+                vcnum = 1;  /* we can not reuse vc=0 safely, since some servers
+                                reset all uids on that, but 1 is ok. */
+        else
+                vcnum = i;
+        ses->vcnum = vcnum;
+get_vc_num_exit:
+        write_unlock(&cifs_tcp_ses_lock);
+
+        return le16_to_cpu(vcnum);
+}
+
 static __u32 cifs_ssetup_hdr(struct cifsSesInfo *ses, SESSION_SETUP_ANDX *pSMB)
 {
         __u32 capabilities = 0;
 
         /* init fields common to all four types of SessSetup */
-        /* note that header is initialized to zero in header_assemble */
+        /* Note that offsets for first seven fields in req struct are same  */
+        /*      in CIFS Specs so does not matter which of 3 forms of struct */
+        /*      that we use in next few lines                               */
+        /* Note that header is initialized to zero in header_assemble */
         pSMB->req.AndXCommand = 0xFF;
         pSMB->req.MaxBufferSize = cpu_to_le16(ses->server->maxBuf);
         pSMB->req.MaxMpxCount = cpu_to_le16(ses->server->maxReq);
+        pSMB->req.VcNumber = get_next_vcnum(ses);
 
         /* Now no need to set SMBFLG_CASELESS or obsolete CANONICAL PATH */
 
@@ -71,7 +155,6 @@ static __u32 cifs_ssetup_hdr(struct cifsSesInfo *ses, SESSION_SETUP_ANDX *pSMB)
         if (ses->capabilities & CAP_UNIX)
                 capabilities |= CAP_UNIX;
 
-        /* BB check whether to init vcnum BB */
         return capabilities;
 }
 
@@ -228,7 +311,7 @@ static int decode_unicode_ssetup(char **pbcc_area, int bleft,
 
         kfree(ses->serverOS);
         /* UTF-8 string will not grow more than four times as big as UCS-16 */
-        ses->serverOS = kzalloc(4 * len, GFP_KERNEL);
+        ses->serverOS = kzalloc((4 * len) + 2 /* trailing null */, GFP_KERNEL);
         if (ses->serverOS != NULL)
                 cifs_strfromUCS_le(ses->serverOS, (__le16 *)data, len, nls_cp);
         data += 2 * (len + 1);
@@ -241,7 +324,7 @@ static int decode_unicode_ssetup(char **pbcc_area, int bleft,
                 return rc;
 
         kfree(ses->serverNOS);
-        ses->serverNOS = kzalloc(4 * len, GFP_KERNEL); /* BB this is wrong length FIXME BB */
+        ses->serverNOS = kzalloc((4 * len) + 2 /* trailing null */, GFP_KERNEL);
         if (ses->serverNOS != NULL) {
                 cifs_strfromUCS_le(ses->serverNOS, (__le16 *)data, len,
                                    nls_cp);
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index d71dc44e21ed..cb329edc925b 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -166,6 +166,75 @@ test_page_region(
 }
 
 /*
+ *      Mapping of multi-page buffers into contiguous virtual space
+ */
+
+typedef struct a_list {
+        void            *vm_addr;
+        struct a_list   *next;
+} a_list_t;
+
+static a_list_t         *as_free_head;
+static int              as_list_len;
+static DEFINE_SPINLOCK(as_lock);
+
+/*
+ *      Try to batch vunmaps because they are costly.
+ */
+STATIC void
+free_address(
+        void            *addr)
+{
+        a_list_t        *aentry;
+
+#ifdef CONFIG_XEN
+        /*
+         * Xen needs to be able to make sure it can get an exclusive
+         * RO mapping of pages it wants to turn into a pagetable.  If
+         * a newly allocated page is also still being vmap()ed by xfs,
+         * it will cause pagetable construction to fail.  This is a
+         * quick workaround to always eagerly unmap pages so that Xen
+         * is happy.
+         */
+        vunmap(addr);
+        return;
+#endif
+
+        aentry = kmalloc(sizeof(a_list_t), GFP_NOWAIT);
+        if (likely(aentry)) {
+                spin_lock(&as_lock);
+                aentry->next = as_free_head;
+                aentry->vm_addr = addr;
+                as_free_head = aentry;
+                as_list_len++;
+                spin_unlock(&as_lock);
+        } else {
+                vunmap(addr);
+        }
+}
+
+STATIC void
+purge_addresses(void)
+{
+        a_list_t        *aentry, *old;
+
+        if (as_free_head == NULL)
+                return;
+
+        spin_lock(&as_lock);
+        aentry = as_free_head;
+        as_free_head = NULL;
+        as_list_len = 0;
+        spin_unlock(&as_lock);
+
+        while ((old = aentry) != NULL) {
+                vunmap(aentry->vm_addr);
+                aentry = aentry->next;
+                kfree(old);
+        }
+}
+
+/*
  *      Internal xfs_buf_t object manipulation
  */
 
@@ -264,7 +333,7 @@ xfs_buf_free(
                 uint            i;
 
                 if ((bp->b_flags & XBF_MAPPED) && (bp->b_page_count > 1))
-                       vm_unmap_ram(bp->b_addr - bp->b_offset, bp->b_page_count);
+                        free_address(bp->b_addr - bp->b_offset);
 
                 for (i = 0; i < bp->b_page_count; i++) {
                         struct page     *page = bp->b_pages[i];
@@ -386,8 +455,10 @@ _xfs_buf_map_pages(
                 bp->b_addr = page_address(bp->b_pages[0]) + bp->b_offset;
                 bp->b_flags |= XBF_MAPPED;
         } else if (flags & XBF_MAPPED) {
-               bp->b_addr = vm_map_ram(bp->b_pages, bp->b_page_count,
-                                       -1, PAGE_KERNEL);
+                if (as_list_len > 64)
+                        purge_addresses();
+                bp->b_addr = vmap(bp->b_pages, bp->b_page_count,
+                                        VM_MAP, PAGE_KERNEL);
                 if (unlikely(bp->b_addr == NULL))
                         return -ENOMEM;
                 bp->b_addr += bp->b_offset;
@@ -1672,6 +1743,8 @@ xfsbufd(
                         count++;
                 }
 
+                if (as_list_len > 0)
+                        purge_addresses();
                 if (count)
                         blk_run_address_space(target->bt_mapping);