source: gpfs_3.1_ker2.6.20/lpp/mmfs/src/gpl-linux/kx.c @ 214

/***************************************************************************
 *
 * Copyright (C) 2001 International Business Machines
 * All rights reserved.
 *
 * This file is part of the GPFS mmfslinux kernel module.
 *
 * Redistribution and use in source and binary forms, with or without 
 * modification, are permitted provided that the following conditions 
 * are met:
 *
 *  1. Redistributions of source code must retain the above copyright notice, 
 *     this list of conditions and the following disclaimer. 
 *  2. Redistributions in binary form must reproduce the above copyright 
 *     notice, this list of conditions and the following disclaimer in the
 *     documentation and/or other materials provided with the distribution. 
 *  3. The name of the author may not be used to endorse or promote products 
 *     derived from this software without specific prior written
 *     permission. 
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 
 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 
 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 
 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 
 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 *************************************************************************** */
/* @(#)39       1.56.1.2  src/avs/fs/mmfs/ts/kernext/gpl-linux/kx.c, mmfs, avs_rgpfs24, rgpfs24s007a 10/25/06 17:52:27 */

#include <Shark-gpl.h>

#ifdef MODULE
#include <linux/module.h>
#endif
#include <linux/string.h>
#include <linux/version.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/sched.h>
#include <linux/fs.h>
#include <linux/dcache.h>
#include <linux/file.h>
#include <linux/stat.h>
#include <linux/fcntl.h>


#ifndef GPFS_ARCH_X86_64
/* grab stuff for madvise */
#define __KERNEL_SYSCALLS__
#endif
#include <linux/unistd.h>
#include <linux/mman.h>

#include <verdep.h>
#include <cxiSystem.h>
#include <cxi2gpfs.h>
#include <cxiCred.h>

#include <linux2gpfs.h>
#include <Trace.h>
#include <cxiMode.h>
#include <cxiTSFattr.h>
#include <cxiVFSStats.h>

#ifdef API_32BIT

/* Equivalent structure for stat64 at 32-bit user level (for API remapping) */
/* gcc doesn't align double words with i386/ia32 like it does for ia64, so
 * we need to adjust for this. */
/* Ideally we would include a header file with this definition (ia32.h) instead
   of defining it ourselves, but this typically breaks things
   (such as with x86_64). */
typedef struct stat64_user32 {
  UInt64 st_dev;
  UInt32 __pad1;
  UInt32 __st_ino;
  UInt32 st_mode;
  UInt32 st_nlink;
  UInt32 st_uid;
  UInt32 st_gid;
  UInt64 st_rdev;
  UInt32 __pad2;
  UInt32 st_size_lower;     /* break up st_size to avoid extra padding */
  UInt32 st_size_upper;     /* w/ little Endian ordering */
  UInt32 st_blksize;
  UInt64 st_blocks;
  UInt32 st_atime;
  UInt32 __unused1;
  UInt32 st_mtime;
  UInt32 __unused2;
  UInt32 st_ctime;
  UInt32 __unused3;
  UInt64 st_ino;
} stat64_user32;
#endif /* API_32BIT */

/* check if inode belongs to GPFS */
#define GPFS_TYPE(IP) (!(iP) ? false : \
                        (!(iP->i_sb) ? false : \
                          (!(iP->i_sb->s_type) ? false : \
                            !strcmp(iP->i_sb->s_type->name, "gpfs"))))

int
kxPoll(void* fdListP, int nFDs, int timeoutMS)
{
  return -ENOSYS;
}

int cleanupFD = 0;

extern struct super_block *shutdownSuperP;

#ifdef API_32BIT
static void
vstat32(cxiVattr_t *vattrp, struct stat64_user32 *statbuf)
{
  ENTER(0);
  memset(statbuf, 0, sizeof(struct stat64_user32));

  statbuf->st_dev         = vattrp->va_dev;
  statbuf->st_ino         = vattrp->va_serialno;
  statbuf->st_mode        = vattrp->va_mode;
  statbuf->st_nlink       = vattrp->va_nlink;
  statbuf->st_uid         = vattrp->va_uid;
  statbuf->st_gid         = vattrp->va_gid;
  statbuf->st_rdev        = vattrp->va_rdev;
  statbuf->st_size_upper  = High32(vattrp->va_size);
  statbuf->st_size_lower  = Low32(vattrp->va_size);
  statbuf->st_atime       = vattrp->va_atime.tv_sec;
  statbuf->st_mtime       = vattrp->va_mtime.tv_sec;
  statbuf->st_ctime       = vattrp->va_ctime.tv_sec;
  statbuf->st_blksize     = vattrp->va_blocksize;
  statbuf->st_blocks      = vattrp->va_blocks;
  /* STAT64_HAS_BROKEN_ST_INO is set on i386 ... */
  statbuf->__st_ino       = vattrp->va_serialno;
  EXIT(0);
}
#endif /* API_32BIT */

void static
vstat(cxiVattr_t *vattrp, struct stat64 *statbuf)
{
  ENTER(0);
  memset(statbuf, 0, sizeof(struct stat64));

  statbuf->st_dev         = vattrp->va_dev;
  statbuf->st_ino         = vattrp->va_serialno;
  statbuf->st_mode        = vattrp->va_mode;
  statbuf->st_nlink       = vattrp->va_nlink;
  statbuf->st_uid         = vattrp->va_uid;
  statbuf->st_gid         = vattrp->va_gid;
  statbuf->st_rdev        = vattrp->va_rdev;
  statbuf->st_size        = vattrp->va_size;
  statbuf->st_atime       = vattrp->va_atime.tv_sec;
  statbuf->st_mtime       = vattrp->va_mtime.tv_sec;
  statbuf->st_ctime       = vattrp->va_ctime.tv_sec;
  statbuf->st_blksize     = vattrp->va_blocksize;
  statbuf->st_blocks      = vattrp->va_blocks;

#ifdef STAT64_HAS_BROKEN_ST_INO
  /* Linux has 2 struct stat64 definitions:
      1) /usr/include/asm/stat.h
      2) /usr/include/bits/stat.h
     Of course, they differ
      1) st_dev & st_rdev is 2 bytes in (1) and 8 bytes in (2),
         but the 2 definitions overlap.
      2) st_ino is 8 bytes in (1) and 4 bytes in (2)
         and they are in different places!
    Fortunately, (1) defines an ifdef telling us to assign st_ino
    to a second variable which just happens to exactly match the
    definition in (2). */
  statbuf->__st_ino       = vattrp->va_serialno;
#endif
  EXIT(0);
}

#ifdef GPFS_ARCH_PPC64
void static
vstat_ppc64(cxiVattr_t *vattrp, struct stat *statbuf)
{
  ENTER(0);
  memset(statbuf, 0, sizeof(struct stat));
                                                                                
  statbuf->st_dev         = vattrp->va_dev;
  statbuf->st_ino         = vattrp->va_serialno;
  statbuf->st_mode        = vattrp->va_mode;
  statbuf->st_nlink       = vattrp->va_nlink;
  statbuf->st_uid         = vattrp->va_uid;
  statbuf->st_gid         = vattrp->va_gid;
  statbuf->st_rdev        = vattrp->va_rdev;
  statbuf->st_size        = vattrp->va_size;
  statbuf->st_atime       = vattrp->va_atime.tv_sec;
  statbuf->st_mtime       = vattrp->va_mtime.tv_sec;
  statbuf->st_ctime       = vattrp->va_ctime.tv_sec;
  statbuf->st_blksize     = vattrp->va_blocksize;
  statbuf->st_blocks      = vattrp->va_blocks;
  EXIT(0);
}
#endif /* GPFS_ARCH_PPC64 */

/* a call to return exact file status information
 * using the file's pathname;
 */
int
tsstat(char *pathname, struct stat64 *statP)
{
  struct gpfsVfsData_t *privVfsP;
  cxiNode_t *cnP;
  int code = 0;
  int rc = 0;
  struct nameidata nd;
  struct inode *iP;
  Boolean relePath = false;
  cxiVattr_t vattr;
#ifdef GPFS_ARCH_PPC64
  struct stat statbuf;
#else
  struct stat64 statbuf;
#endif
#ifdef API_32BIT
  stat64_user32 statbuf32;
#endif

  ENTER(0);
  if (pathname == NULL || statP == NULL)
  {
    code = 1;
    rc = EINVAL;
    goto xerror;
  }

  rc = user_path_walk(pathname, &nd);
  if (rc)
  {
    rc = -rc;
    code = 2;
    goto xerror;
  }
  relePath = true;

  iP = nd.dentry->d_inode;
  DBGASSERT(iP != NULL);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_TSSTAT_ENTER,
         "tsstat enter: iP 0x%lX statP 0x%lX\n", iP, statP);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }
  cnP = VP_TO_CNP(iP);

  rc = gpfs_ops.gpfsGetattr(privVfsP, cnP, &vattr, true);
  vattr.va_rdev = cxiDev32ToDev(vattr.va_rdev);

  if (rc == 0)
  {
#ifdef API_32BIT
    /* Need to remap? */
    if (cxiIS64U((char *)statP))
    {
      vstat(&vattr, &statbuf);

      rc = cxiCopyOut((char *)&statbuf, (char *)statP, sizeof(struct stat64));
    }
    else
    {
      /* Need to remap for 32-bit API call */
      vstat32(&vattr, &statbuf32);

      rc = cxiCopyOut((char *)&statbuf32, (char *)statP,
                      sizeof(struct stat64_user32));
    }
#else

#ifdef GPFS_ARCH_PPC64
    /* On ppc64, struct stat64 layout is different for a 64-bit user app and
       kernel, and we can't use it for copying data out.  However, struct stat
       has the same layout in kernel and 64-bit userspace, and is identical
       to struct stat64 in 64-bit userspace, so we use it for copyout. */
    vstat_ppc64(&vattr, &statbuf);
#else
    vstat(&vattr, &statbuf);
#endif
    rc = cxiCopyOut((char *)&statbuf, (char *)statP, sizeof(statbuf));

#endif /* API_32BIT */
    if (rc != 0)
    {
      code = 5;
      goto xerror;
    }
  }

xerror:
  if (relePath)
    path_release(&nd);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_TSSTAT_EXIT,
         "tsstat exit: code %d rc %d\n", code, rc);
  EXIT(0);
  return -rc;
}

int
tsfstat(int fileDesc, struct stat64 *statP)
{
  struct gpfsVfsData_t *privVfsP;
  cxiNode_t *cnP;
  struct file *fP;
  struct inode *iP;
  int code = 0;
  int rc;
  Boolean releFile = false;
  cxiVattr_t vattr;
#ifdef GPFS_ARCH_PPC64
  struct stat statbuf;
#else
  struct stat64 statbuf;
#endif
#ifdef API_32BIT
  struct stat64_user32 statbuf32;
#endif /* API_32BIT */

  ENTER(0);
  if (statP == NULL)
  {
    code = 1;
    rc = EINVAL;
    goto xerror;
  }

  fP = fget(fileDesc);
  if (!fP)
  {
    code = 2;
    rc = EBADF;
    goto xerror;
  }
  releFile = true;

  iP = fP->f_dentry->d_inode;
  DBGASSERT(iP != NULL);

  TRACE4(TRACE_VNODE, 1, TRCID_LINUXOPS_TSFSTAT_ENTER,
         "tsfstat enter: fd %d fP 0x%lX iP 0x%lX statP 0x%lX\n",
         fileDesc, fP, iP, statP);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }
  cnP = VP_TO_CNP(iP);

  rc = gpfs_ops.gpfsGetattr(privVfsP, cnP, &vattr, true);
  vattr.va_rdev = cxiDev32ToDev(vattr.va_rdev);

  if (rc == 0)
  {
#ifdef API_32BIT
    /* Need to remap? */
    if (cxiIS64U((char *)statP))
    {
      vstat(&vattr, &statbuf);

      rc = cxiCopyOut((char *)&statbuf, (char *)statP, sizeof(struct stat64));
    }
    else
    {
      /* Need to remap for 32-bit API call */
      vstat32(&vattr, &statbuf32);

      rc = cxiCopyOut((char *)&statbuf32, (char *)statP,
                      sizeof(struct stat64_user32));
    }
#else

#ifdef GPFS_ARCH_PPC64
    /* On ppc64, struct stat64 layout is different for a 64-bit user app and
       kernel, and we can't use it for copying data out.  However, struct stat
       has the same layout in kernel and 64-bit userspace, and is identical
       to struct stat64 in 64-bit userspace, so we use it for copyout. */
    vstat_ppc64(&vattr, &statbuf);
#else
    vstat(&vattr, &statbuf);
#endif
    rc = cxiCopyOut((char *)&statbuf, (char *)statP, sizeof(statbuf));

#endif /* API_32BIT */
    if (rc != 0)
    {
      code = 5;
      goto xerror;
    }
  }

xerror:
  if (releFile)
    fput(fP);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_TSFSTAT_EXIT,
         "tsfstat exit: code %d rc %d\n", code, rc);
  EXIT(0);
  return -rc;
}

int
tsfattr(int fileDesc, int command, void *argP,
        tsfattrReasonCodeInfo *rCodeP)
{
  struct gpfsVfsData_t *privVfsP;
  cxiNode_t *cnP;
  struct MMFSVInfo *vinfoP = NULL;
  ext_cred_t eCred;
  struct file *fP;
  struct inode *iP;
  int reason = 0;
  int code = 0;
  int flags;
  int rc;
  Boolean releFile = false;

  VFS_STAT_START(tsfattrCall);
  ENTER(0);

#ifdef SMB_LOCKS_TEST
  /* hack to test setSMBOplock via tsfattr call */
  if (command == 999)
  {
    int args[2];
    int oplockGranted;
    struct file *fP;

    rc = cxiCopyIn((char *)argP, (char *)args, sizeof(args));
    if (rc)
      return -rc;

    /* Obtain the file pointer from the descriptor */
    fP = fget(fileDesc);
    if (!fP)
      return -EBADF;

    rc = setSMBOplock(fP, fP->private_data, args[0], args[1],
                      (void *)0xdeadbeef, &oplockGranted, false);

    fput(fP);

    if (rc)
      return rc;
    return oplockGranted;
  }
#endif

  /* if rCodeP was specified then immediately reset reason field */
  if (rCodeP != NULL)
  {
    rc = cxiCopyOut((char*)&reason, (char*)&(rCodeP->reason), sizeof(reason));
    if (rc != 0)
    {
      code = 1;
      goto xerror;
    }
  }

  fP = fget(fileDesc);
  if (!fP)
  {
    code = 2;
    rc = EBADF;
    goto xerror;
  }
  releFile = true;

  iP = fP->f_dentry->d_inode;
  DBGASSERT(iP != NULL);

  flags = cxiOpenFlagsXlate(fP->f_flags);

  TRACE8(TRACE_VNODE, 1, TRCID_LINUXOPS_TSFATTR_ENTER,
         "tsfattr enter: fd %d fP 0x%lX f_flags 0x%X flags 0x%X "
         "iP 0x%lX command %d argP 0x%lX rCodeP 0x%lX\n",
         fileDesc, fP, fP->f_flags, flags, iP, 
         command, argP, rCodeP);

  switch (iP->i_mode & S_IFMT)
  {
    case S_IFREG:
    case S_IFDIR:
      vinfoP = (struct MMFSVInfo *)fP->private_data;
      break;

    case S_IFLNK:
    case S_IFBLK:
    case S_IFCHR:
    case S_IFIFO:
    case S_IFSOCK:
      /* At the moment we do have have any file ops enabled for the
       * special files on Linux, and thus we don't have an OpenFile object 
       * or vinfoP stored in fP.  This means that a special file can't
       * have e.g. ACLs associated with it.  Fixing this properly will 
       * require a redesign of our LFS interface.  For now, just fix the 
       * kernel panic caused by invalid vinfoP. */
      vinfoP = NULL;
      if ((command != REMOTE_STAT) &&   // UIDREMAP
          (command != GET_XATTR) &&     // GPFS_LINUX
          (command != SET_XATTR))       // GPFS_LINUX
      {
        code = 5;
        rc = ENOSYS;
        goto xerror;
      }
      break;
  }

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }
  cnP = VP_TO_CNP(iP);

  setCred(&eCred);

  rc = gpfs_ops.gpfsFattr(privVfsP, cnP, vinfoP, flags, command, argP,
                          rCodeP, &eCred);

xerror:
  if (releFile)
    fput(fP);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_TSFATTR_EXIT,
         "tsfattr exit: code %d rc %d\n", code, rc);

  VFS_STAT_STOP;
  EXIT(0);
  return -rc;
}

int
tsfsattr(int command, void *argP)
{
  int rc;

  VFS_STAT_START(tsfattrCall);
  ENTER(0);

  TRACE1(TRACE_VNODE, 1, TRCID_LINUXOPS_TSFSATTR_ENTER,
        "tsfsattr enter: command %s\n",tsfsattrcmd_id2name(command));

  rc = gpfs_ops.gpfsFsAttr(command, argP);

  TRACE1(TRACE_VNODE, 1, TRCID_LINUXOPS_TSFSATTR_EXIT,
         "tsfsattr exit: rc %d\n", rc);

  VFS_STAT_STOP;
  EXIT(0);
  return -rc;
}

int
tsattr(char *pathname, int command, void *argP,
       tsfattrReasonCodeInfo *rCodeP)
{
  struct gpfsVfsData_t *privVfsP;
  cxiNode_t *cnP;
  int code = 0;
  int rc = 0;
  int reason = 0;       /* local reason code for resetting caller's */
  ext_cred_t eCred;
  struct nameidata nd;
  struct inode *iP;
  Boolean relePath = false;

  /* if rCodeP was specified then immediately reset reason field */
  ENTER(0);
  if (rCodeP != NULL)
  {
    rc = cxiCopyOut((char*)&reason, (char*)&(rCodeP->reason), sizeof(reason));
    if (rc != 0)
    {
      code = 1;
      goto xerror;
    }
  }

  rc = user_path_walk(pathname, &nd);
  if (rc)
  {
    rc = -rc;
    code = 2;
    goto xerror;
  }
  relePath = true;

  iP = nd.dentry->d_inode;
  DBGASSERT(iP != NULL);

  TRACE4(TRACE_VNODE, 1, TRCID_LINUXOPS_TSATTR_ENTER,
         "tsattr enter: iP 0x%lX command %d argP 0x%lX rCodeP 0x%lX\n",
         iP, command, argP, rCodeP);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }
  cnP = VP_TO_CNP(iP);

  setCred(&eCred);
  rc = gpfs_ops.gpfsFattr(privVfsP, cnP, NULL, 0, command, argP, rCodeP,
                          &eCred);

xerror:
  if (relePath)
    path_release(&nd);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_TSATTR_EXIT,
         "tsattr exit: code %d rc %d\n", code, rc);

  EXIT(0);
  return -rc;
}

int
kxGetACL(char *pathname, int flags, void *aclP)
{
  int rc = 0;
  int code = 0;
  cxiNode_t *cnP;
  struct gpfsVfsData_t *privVfsP;
  ext_cred_t eCred;
  struct nameidata nd;
  struct inode *iP;
  Boolean relePath = false;

  /* Verify input exists */
  ENTER(0);
  if (pathname == NULL || aclP == NULL)
  {
    code = 1;
    rc = EINVAL;
    goto xerror;
  }

  rc = user_path_walk(pathname, &nd);
  if (rc)
  {
    rc = -rc;
    code = 2;
    goto xerror;
  }
  relePath = true;

  iP = nd.dentry->d_inode;
  DBGASSERT(iP != NULL);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }

  cnP = VP_TO_CNP(iP);

  /* Retrieve the ACL data */
  setCred(&eCred);
  rc = gpfs_ops.gpfsGetAcl(privVfsP, cnP, flags, aclP, &eCred);

xerror:
  if (relePath)
    path_release(&nd);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_KXGETACL_EXIT,
         "kxGetACL exit: code %d rc %d\n", code, rc);
  EXIT(0);
  return -rc;
}

int
kxPutACL(char *pathname, int flags, void *aclP)
{
  int rc = 0;
  int code = 0;
  cxiNode_t *cnP;
  struct gpfsVfsData_t *privVfsP;
  ext_cred_t eCred;
  struct nameidata nd;
  struct inode *iP;
  Boolean relePath = false;

  /* Verify input exists */
  ENTER(0);
  if (pathname == NULL || aclP == NULL)
  {
    code = 1;
    rc = EINVAL;
    goto xerror;
  }

  rc = user_path_walk(pathname, &nd);
  if (rc)
  {
    rc = -rc;
    code = 2;
    goto xerror;
  }
  relePath = true;

  iP = nd.dentry->d_inode;
  DBGASSERT(iP != NULL);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }

  cnP = VP_TO_CNP(iP);

  /* Put the ACL data */
  setCred(&eCred);
  rc = gpfs_ops.gpfsPutAcl(privVfsP, cnP, flags, aclP, &eCred);

xerror:
  if (relePath)
    path_release(&nd);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_KXPUTACL_EXIT,
         "kxPutACL exit: code %d rc %d\n", code, rc);
  EXIT(0);
  return -rc;
}

#if !defined(GPFS_ARCH_X86_64) && !defined(GPFS_ARCH_IA64)
/* madvise is handled as a kernel extension / ioctl call until support
   is provided by glibc */
//static inline _syscall3(int,madvise,void*,addr,size_t,len,int,advice);
#endif

int
kxMadvise(void* addr, size_t len, int advice)
{
#if defined(GPFS_ARCH_X86_64) || defined(GPFS_ARCH_IA64)
  /* kxMadvise should never be called since glibc has madvice */
  return -ENOSYS;
#else
//  return madvise(addr, len, advice);
//    return stub_syscall2(__NR_madvise, len, advice); 
//    return sys_madvise(addr, len, advice);
  return -ENOSYS;
#endif
}

/* Get the kernel thread ID. */
cxiThreadId
kxGetThreadID()
{
  return (cxiThreadId)current->pid;
}

#ifdef GPFS_QUOTACTL
int kxQuotactl(char *pathname, int cmd, int qid, void *bufferP)
{
  int rc = 0;
  int code = 0;
  cxiNode_t *cnP;
  struct gpfsVfsData_t *privVfsP;
  ext_cred_t  eCred;
  ext_cred_t *credP = NULL;
  struct nameidata nd;
  struct inode *iP;
  Boolean relePath = false;
  ulong len;

  ENTER(0);

  TRACE4(TRACE_VNODE, 2, TRCID_LINUXOPS_TSQUOTACTL_K,
         "kxQuotactl: enter cmd 0x%x, qid %d, pathname 0x%x, euid %d\n",
          cmd, qid, pathname, current->euid);

  /* Verify input exists */
  if (pathname == NULL)
  {
    code = 1; // EBUF_ERROR
    rc = EINVAL;
    goto xerror;
  }

  TRACE0(TRACE_VNODE, 2, TRCID_LINUXOPS_TSQUOTACTL_K1,
       "kxQuotactl at [1]\n");

  rc = user_path_walk(pathname, &nd);

  TRACE1(TRACE_VNODE, 2, TRCID_LINUXOPS_TSQUOTACTL_K2,
        "kxQuotactl at [2] user_path_walk rc %d\n",
         rc);

  if (rc)
  {
    rc = -rc;
    code = 3;
    goto xerror;
  }

  relePath = true;

  iP = nd.dentry->d_inode;
  DBGASSERT(iP != NULL);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }

  cnP = VP_TO_CNP(iP);

  setCred(&eCred);
  credP = &eCred;
 
  TRACE4(TRACE_VNODE, 2, TRCID_LINUXOPS_TSQUOTACTL_K6,
         "kxQuotactl: gpfs_ops.gpfsQuotactl, cmd 0x%x, qid %d, credP 0x%x, bufferP 0x%x\n",
         cmd, qid, credP, bufferP);

  rc = gpfs_ops.gpfsQuotactl(privVfsP, cmd, qid, bufferP, credP);

  TRACE1(TRACE_VNODE, 2, TRCID_LINUXOPS_TSQUOTACTL_K7,
         "kxQuotactl: gpfs_ops.gpfsQuotactl rc %d\n", rc);
xerror:
  if (relePath)
    path_release(&nd);

  TRACE3(TRACE_VNODE, 1, TRCID_LINUXOPS_TSQUOTACTL_EXIT,
         "kxQuotactl exit: code %d rc %d, pathname %s\n", 
         code, rc, pathname ? pathname : "?");

  EXIT(0);

  return -rc;
}
#endif /* GPFS_QUOTACTL */

#ifdef SMB_LOCKS
/* Exchange control between Samba, GPFS */
int
setSMBOpenLockControl(void *vP, void *infoP, int command, int lockmode)
{
  struct inode *iP;
  struct gpfsVfsData_t *privVfsP;
  struct MMFSVInfo *vinfoP;
  struct file *fP = (struct file *)vP;
  int rc = EINVAL;

  ENTER(0);

  TRACE2(TRACE_VNODE, 9, TRCID_SMBOPENLOCKCONTROL_ENTERL,
         "setSMBOpenLockControl entry - command %d, mode 0x%x",
         command, lockmode);

  vinfoP = (struct MMFSVInfo *)fP->private_data;

  iP = fP->f_dentry->d_inode;
  DBGASSERT(iP != NULL);

  if (!GPFS_TYPE(iP))
  {
    rc = 0;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);

  rc = gpfs_ops.SMBOpenLockControl(command, lockmode, iP->i_ino, vinfoP,
                                   privVfsP);

xerror:

  TRACE1(TRACE_VNODE, 1, TRCID_SMBOPENLOCKCONTROL_EXITL,
         "setSMBOpLockControl exit: rc %d\n", rc);

  EXIT(0);
  return (rc);
}

#include <oplock.h>
#include <asm/uaccess.h>
int
setSMBOplock(void *vP, void *infoP, int accessWant, int oplockWant,
             void *breakArgP, int *oplockGrantedP, Boolean isAsync)
{
  struct file *filp = (struct file *)vP;
  struct gpfsVfsData_t *privVfsP;
  cxiNode_t *cnP;
  struct MMFSVInfo *vinfoP;
  struct file *fP;
  struct inode *iP;
  int reason = 0;
  int code = 0;
  int flags;
  int rc;
  Boolean releFile = false;

  /* get file pointer and inode */
  ENTER(0);
  fP = filp;
  if (!fP)
  {
    code = 2;
    rc = EBADF;
    goto xerror;
  }

  get_file(fP);

  releFile = true;

  vinfoP = (struct MMFSVInfo *)fP->private_data;

  iP = fP->f_dentry->d_inode;
  DBGASSERT(iP != NULL);

  TRACE5(TRACE_VNODE, 1, TRCID_LINUXOPS_SETSMBOPLOCK_E,
         "setSMBOplock enter: fP 0x%lX acc %d oplock %d argP 0x%lX async %d\n",
         fP, accessWant, oplockWant, breakArgP, isAsync);

  if (!GPFS_TYPE(iP))
  {
    code = 3;
    rc = EINVAL;
    goto xerror;
  }

  privVfsP = VP_TO_PVP(iP);
  if (privVfsP == NULL)
  {
    code = 4;
    rc = ENOENT;
    goto xerror;
  }
  cnP = VP_TO_CNP(iP);

  rc = gpfs_ops.gpfsSetSMBOplock(privVfsP, cnP, vinfoP, accessWant, oplockWant,
                                 breakArgP, oplockGrantedP, isAsync);

xerror:
  if (releFile)
    fput(fP);

  TRACE2(TRACE_VNODE, 1, TRCID_LINUXOPS_SETSMBOPLOCK_X,
         "setSMBOplock exit: granted %d rc %d", *oplockGrantedP, rc);

  EXIT(0);
  return -rc;
}

/* Share/Delegate interface for Samba */
int
kxGetShare(int fd, unsigned int share, unsigned int deny)
{
  int rc;
  struct file *fP = NULL;
  struct inode *iP;
  struct MMFSVInfo *vinfoP;
  struct gpfsVfsData_t *privVfsP;
  int shareWant;

  /* To be called from our Samba vfs_op library only. */
  ENTER(0);
  if (!cxiIsSambaThread())
  {
    rc = EPERM;
    goto xerror;
  }

  fP = fget(fd);
  if (!fP || !fP->f_dentry || !fP->f_dentry->d_inode)
  {
    rc = EINVAL;
    goto xerror;
  }

  vinfoP = (struct MMFSVInfo *)fP->private_data;

  iP = fP->f_dentry->d_inode;
  if (!iP || !(privVfsP = VP_TO_PVP(iP)))
  {
    rc = EBADF;
    goto xerror;
  }

  if (!GPFS_TYPE(iP))
  {
    rc = 0;
    goto xerror;
  }

  shareWant = ALLOW_SHARE_DELETE;

  TRACE6(TRACE_VNODE, 1, TRCID_LINUX_SHARE_ENTER,
         "kxGetShare: enter fd %d fP 0x%lX ino %d infoP 0x%lX access 0x%X "
         "deny 0x%X\n", fd, fP, iP->i_ino, vinfoP, share, deny);

  /* Translate (and validate) the Samba share/deny arguments to our shareWant */

  rc = EINVAL; /* setup for XLATE_SAMBA_ACCESS/DENY */

  XLATE_SAMBA_ACCESS(share, shareWant);
  XLATE_SAMBA_DENY(deny, shareWant);

  /* Call to make the reservation */
  if ((share != 0) || (deny != 0))
    rc = gpfs_ops.gpfsReserveShare(fP, vinfoP, privVfsP, 0, shareWant, NULL,NULL);
  else
    rc = 0;

  /* give back any pre-existing rights */
  if ((rc == 0) && (0 != (coMask & ~shareWant)))
    rc = gpfs_ops.gpfsReserveShare(fP, vinfoP, privVfsP, RESERVE_DOWNGRADE, 
                                   coMask & ~shareWant, NULL, NULL);
xerror:
  if (fP)
    fput(fP);

  TRACE1(TRACE_VNODE, 1, TRCID_LINUX_SHARE_EXIT, 
         "kxGetShare: exit rc %d\n", rc);

  EXIT(0);
  return -rc;
}

int
kxGetDelegation(int fd, unsigned int delegate_access,
                void *cb_token, void *cookie)
{
  int rc = 0;
  int code = 0;
  int flags = 0;
  struct file *fP = NULL;
  struct inode *iP;
  struct MMFSVInfo *vinfoP;
  struct gpfsVfsData_t *privVfsP;
  int oplockWant;

  /* To be called from our Samba vfs_op library only. */
  ENTER(0);
  if (!cxiIsSambaThread())
  {
    rc = EPERM;
    goto xerror;
  }

#if 0
  /* As long as Samba write leases are not as strict as NFSv4 write
     delegations (are not required to be revoked on stat) we can
     allow them.
  */
  if (delegate_access == CXI_DELEGATE_WRITE)
  {
    rc = EOPNOTSUPP;
    goto xerror;
  }
#endif

  fP = fget(fd);
  if (!fP || !fP->f_dentry || !fP->f_dentry->d_inode)
  {
    rc = EINVAL;
    goto xerror;
  }

  vinfoP = (struct MMFSVInfo *)fP->private_data;

  iP = fP->f_dentry->d_inode;
  if (!iP || !(privVfsP = VP_TO_PVP(iP)))
  {
    rc = EBADF;
    goto xerror;
  }

  if (!GPFS_TYPE(iP))
  {
    rc = 0;
    goto xerror;
  }


  TRACE5(TRACE_VNODE, 1, TRCID_DELEGATE_ENTER,
         "kxGetDelegation: enter fd %d fP 0x%lX ino %d vinfoP 0x%lX "
         "delegate 0x%x\n", fd, fP, iP->i_ino, vinfoP, delegate_access);

  /* Translate (and validate) the Samba delegate argument to our oplock flags */

  /* setup for XLATE_SAMBA_DELEGATE */
  rc = EINVAL;
  oplockWant = 0;

  if (delegate_access == CXI_DELEGATE_NONE)
    flags = RESERVE_DOWNGRADE;

  XLATE_SAMBA_DELEGATE(delegate_access, oplockWant);

  rc = gpfs_ops.gpfsReserveDelegation(fP, vinfoP, privVfsP, oplockWant, 
                                      flags, iP, NULL);

xerror:
  if (fP)
    fput(fP);

  TRACE1(TRACE_VNODE, 1, TRCID_DELEGATE_EXIT,
         "kxGetDelegation: exit rc %d\n", rc);

  EXIT(0);
  return -rc;
}
#endif

/* called from umount helper before calling umount system call  */

#define UMOUNT_BEGIN 0
#define UMOUNT_END 1
int kxUMount(int type, char *pathname, int force, int *ret_err)
{
    char kxnode[CXI_MAXPATHLEN+1];
    int rc = 0;
    int dmrc = 0;
    int code = 0;
    ulong len;
    struct cxiNode_t  *cnP = NULL;
    struct gpfsVfsData_t *privVfsP = NULL;
    void       *ndP = NULL;
    Boolean doDMEvents = false;

    ENTER(0);
    rc = cxiCopyInstr(pathname, kxnode, CXI_MAXPATHLEN, &len);
    LOGASSERT(rc == 0);
    TRACE4(TRACE_VNODE, 1, TRCID_KXUMOUNT_ENT,
         "kxUMount: enter type %d path %s force %d  ret_err %d\n",
          type,kxnode,force,*ret_err);
    
#ifdef DMAPI
    rc = cxiPathToVfsP((void **)&privVfsP,pathname, &ndP, (void **)&cnP, true);
    if (rc)
    {
      code = 1;
      goto exit;
    }         
    
    /* Generate PREUNMOUNT event for normal mounts. For force unmount
     it is generated from gpfs_s_umount_begin.
     umount helper calls with UMOUNT_BEGIN before calling umount
     system call.
    */

    if ( type == UMOUNT_BEGIN && !force)
    {
       dmrc = gpfs_ops.gpfsDmUnmountEvent(true,force,privVfsP,cnP,
                                        &doDMEvents, NULL, NULL, NULL, 0);
       if (doDMEvents )
         *ret_err = 1;
       else
         *ret_err = 0;
    }

  /* 
    umount helper calls with UMOUNT_END if umount system call fails.
    */

    if ( type == UMOUNT_END )
       dmrc = gpfs_ops.gpfsDmUnmountEvent(false,force,privVfsP,cnP,
                                     &doDMEvents, NULL, NULL, NULL, *ret_err);

#endif
exit:
    if (ndP)
      cxiPathRel(ndP);

    TRACE4(TRACE_VNODE, 1, TRCID_KXUMOUNT_EXT,
         "kxUMount: exit dmrc %d  ret_err %d rc %d code %d\n",
          dmrc,*ret_err,rc,code);
     EXIT(0);
     return rc;
}