Linux: Test for __vfs_write rather than __vfs_read

[pkg-k5-afs_openafs.git] / src / afs / LINUX / osi_vnodeops.c

/*
 * Copyright 2000, International Business Machines Corporation and others.
 * All Rights Reserved.
 *
 * This software has been released under the terms of the IBM Public
 * License.  For details, see the LICENSE file in the top-level source
 * directory or online at http://www.openafs.org/dl/license10.html
 */

/*
 * Linux specific vnodeops. Also includes the glue routines required to call
 * AFS vnodeops.
 *
 * So far the only truly scary part is that Linux relies on the inode cache
 * to be up to date. Don't you dare break a callback and expect an fstat
 * to give you meaningful information. This appears to be fixed in the 2.1
 * development kernels. As it is we can fix this now by intercepting the
 * stat calls.
 */

#include <afsconfig.h>
#include "afs/param.h"


#include "afs/sysincludes.h"
#include "afsincludes.h"
#include "afs/afs_stats.h"
#include <linux/mm.h>
#ifdef HAVE_MM_INLINE_H
#include <linux/mm_inline.h>
#endif
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/pagevec.h>
#include <linux/aio.h>
#include "afs/lock.h"
#include "afs/afs_bypasscache.h"

#include "osi_compat.h"
#include "osi_pagecopy.h"

#ifndef HAVE_LINUX_PAGEVEC_LRU_ADD_FILE
#define __pagevec_lru_add_file __pagevec_lru_add
#endif

#ifndef MAX_ERRNO
#define MAX_ERRNO 1000L
#endif

#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,34)
/* Enable our workaround for a race with d_splice_alias. The race was fixed in
 * 2.6.34, so don't do it after that point. */
# define D_SPLICE_ALIAS_RACE
#endif

int cachefs_noreadpage = 0;

extern struct backing_dev_info *afs_backing_dev_info;

extern struct vcache *afs_globalVp;

/* This function converts a positive error code from AFS into a negative
 * code suitable for passing into the Linux VFS layer. It checks that the
 * error code is within the permissable bounds for the ERR_PTR mechanism.
 *
 * _All_ error codes which come from the AFS layer should be passed through
 * this function before being returned to the kernel.
 */

static inline int
afs_convert_code(int code) {
    if ((code >= 0) && (code <= MAX_ERRNO))
        return -code;
    else
        return -EIO;
}

/* Linux doesn't require a credp for many functions, and crref is an expensive
 * operation. This helper function avoids obtaining it for VerifyVCache calls
 */

static inline int
afs_linux_VerifyVCache(struct vcache *avc, cred_t **retcred) {
    cred_t *credp = NULL;
    struct vrequest *treq = NULL;
    int code;

    if (avc->f.states & CStatd) {
        if (retcred)
            *retcred = NULL;
        return 0;
    }

    credp = crref();

    code = afs_CreateReq(&treq, credp);
    if (code == 0) {
        code = afs_VerifyVCache2(avc, treq);
        afs_DestroyReq(treq);
    }

    if (retcred != NULL)
        *retcred = credp;
    else
        crfree(credp);

    return afs_convert_code(code);
}

#if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER) || defined(HAVE_LINUX_GENERIC_FILE_AIO_READ)
# if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER)
static ssize_t
afs_linux_read_iter(struct kiocb *iocb, struct iov_iter *iter)
# elif defined(LINUX_HAS_NONVECTOR_AIO)
static ssize_t
afs_linux_aio_read(struct kiocb *iocb, char __user *buf, size_t bufsize,
                   loff_t pos)
# else
static ssize_t
afs_linux_aio_read(struct kiocb *iocb, const struct iovec *buf,
                   unsigned long bufsize, loff_t pos)
# endif
{
    struct file *fp = iocb->ki_filp;
    ssize_t code = 0;
    struct vcache *vcp = VTOAFS(fp->f_dentry->d_inode);
# if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER)
    loff_t pos = iocb->ki_pos;
    unsigned long bufsize = iter->nr_segs;
# endif


    AFS_GLOCK();
    afs_Trace4(afs_iclSetp, CM_TRACE_AIOREADOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, ICL_HANDLE_OFFSET(pos), ICL_TYPE_INT32,
               (afs_int32)bufsize, ICL_TYPE_INT32, 99999);
    code = afs_linux_VerifyVCache(vcp, NULL);

    if (code == 0) {
        /* Linux's FlushPages implementation doesn't ever use credp,
         * so we optimise by not using it */
        osi_FlushPages(vcp, NULL);      /* ensure stale pages are gone */
        AFS_GUNLOCK();
# if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER)
        code = generic_file_read_iter(iocb, iter);
# else
        code = generic_file_aio_read(iocb, buf, bufsize, pos);
# endif
        AFS_GLOCK();
    }

    afs_Trace4(afs_iclSetp, CM_TRACE_AIOREADOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, ICL_HANDLE_OFFSET(pos), ICL_TYPE_INT32,
               (afs_int32)bufsize, ICL_TYPE_INT32, code);
    AFS_GUNLOCK();
    return code;
}
#else
static ssize_t
afs_linux_read(struct file *fp, char *buf, size_t count, loff_t * offp)
{
    ssize_t code = 0;
    struct vcache *vcp = VTOAFS(fp->f_dentry->d_inode);

    AFS_GLOCK();
    afs_Trace4(afs_iclSetp, CM_TRACE_READOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, offp, ICL_TYPE_INT32, count, ICL_TYPE_INT32,
               99999);
    code = afs_linux_VerifyVCache(vcp, NULL);

    if (code == 0) {
        /* Linux's FlushPages implementation doesn't ever use credp,
         * so we optimise by not using it */
        osi_FlushPages(vcp, NULL);      /* ensure stale pages are gone */
        AFS_GUNLOCK();
        code = do_sync_read(fp, buf, count, offp);
        AFS_GLOCK();
    }

    afs_Trace4(afs_iclSetp, CM_TRACE_READOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, offp, ICL_TYPE_INT32, count, ICL_TYPE_INT32,
               code);
    AFS_GUNLOCK();
    return code;
}
#endif


/* Now we have integrated VM for writes as well as reads. the generic write operations
 * also take care of re-positioning the pointer if file is open in append
 * mode. Call fake open/close to ensure we do writes of core dumps.
 */
#if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER) || defined(HAVE_LINUX_GENERIC_FILE_AIO_READ)
# if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER)
static ssize_t
afs_linux_write_iter(struct kiocb *iocb, struct iov_iter *iter)
# elif defined(LINUX_HAS_NONVECTOR_AIO)
static ssize_t
afs_linux_aio_write(struct kiocb *iocb, const char __user *buf, size_t bufsize,
                    loff_t pos)
# else
static ssize_t
afs_linux_aio_write(struct kiocb *iocb, const struct iovec *buf,
                    unsigned long bufsize, loff_t pos)
# endif
{
    ssize_t code = 0;
    struct vcache *vcp = VTOAFS(iocb->ki_filp->f_dentry->d_inode);
    cred_t *credp;
# if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER)
    loff_t pos = iocb->ki_pos;
    unsigned long bufsize = iter->nr_segs;
# endif

    AFS_GLOCK();

    afs_Trace4(afs_iclSetp, CM_TRACE_AIOWRITEOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, ICL_HANDLE_OFFSET(pos), ICL_TYPE_INT32,
               (afs_int32)bufsize, ICL_TYPE_INT32,
               (iocb->ki_filp->f_flags & O_APPEND) ? 99998 : 99999);

    code = afs_linux_VerifyVCache(vcp, &credp);

    ObtainWriteLock(&vcp->lock, 529);
    afs_FakeOpen(vcp);
    ReleaseWriteLock(&vcp->lock);
    if (code == 0) {
            AFS_GUNLOCK();
# if defined(STRUCT_FILE_OPERATIONS_HAS_READ_ITER)
            code = generic_file_write_iter(iocb, iter);
# else
            code = generic_file_aio_write(iocb, buf, bufsize, pos);
# endif
            AFS_GLOCK();
    }

    ObtainWriteLock(&vcp->lock, 530);

    if (vcp->execsOrWriters == 1 && !credp)
      credp = crref();

    afs_FakeClose(vcp, credp);
    ReleaseWriteLock(&vcp->lock);

    afs_Trace4(afs_iclSetp, CM_TRACE_AIOWRITEOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, ICL_HANDLE_OFFSET(pos), ICL_TYPE_INT32,
               (afs_int32)bufsize, ICL_TYPE_INT32, code);

    if (credp)
      crfree(credp);
    AFS_GUNLOCK();
    return code;
}
#else
static ssize_t
afs_linux_write(struct file *fp, const char *buf, size_t count, loff_t * offp)
{
    ssize_t code = 0;
    struct vcache *vcp = VTOAFS(fp->f_dentry->d_inode);
    cred_t *credp;

    AFS_GLOCK();

    afs_Trace4(afs_iclSetp, CM_TRACE_WRITEOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, offp, ICL_TYPE_INT32, count, ICL_TYPE_INT32,
               (fp->f_flags & O_APPEND) ? 99998 : 99999);

    code = afs_linux_VerifyVCache(vcp, &credp);

    ObtainWriteLock(&vcp->lock, 529);
    afs_FakeOpen(vcp);
    ReleaseWriteLock(&vcp->lock);
    if (code == 0) {
            AFS_GUNLOCK();
            code = do_sync_write(fp, buf, count, offp);
            AFS_GLOCK();
    }

    ObtainWriteLock(&vcp->lock, 530);

    if (vcp->execsOrWriters == 1 && !credp)
      credp = crref();

    afs_FakeClose(vcp, credp);
    ReleaseWriteLock(&vcp->lock);

    afs_Trace4(afs_iclSetp, CM_TRACE_WRITEOP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_OFFSET, offp, ICL_TYPE_INT32, count, ICL_TYPE_INT32,
               code);

    if (credp)
      crfree(credp);
    AFS_GUNLOCK();
    return code;
}
#endif

extern int BlobScan(struct dcache * afile, afs_int32 ablob, afs_int32 *ablobOut);

/* This is a complete rewrite of afs_readdir, since we can make use of
 * filldir instead of afs_readdir_move. Note that changes to vcache/dcache
 * handling and use of bulkstats will need to be reflected here as well.
 */
static int
#if defined(STRUCT_FILE_OPERATIONS_HAS_ITERATE)
afs_linux_readdir(struct file *fp, struct dir_context *ctx)
#else
afs_linux_readdir(struct file *fp, void *dirbuf, filldir_t filldir)
#endif
{
    struct vcache *avc = VTOAFS(FILE_INODE(fp));
    struct vrequest *treq = NULL;
    struct dcache *tdc;
    int code;
    int offset;
    afs_int32 dirpos;
    struct DirEntry *de;
    struct DirBuffer entry;
    ino_t ino;
    int len;
    afs_size_t origOffset, tlen;
    cred_t *credp = crref();
    struct afs_fakestat_state fakestat;

    AFS_GLOCK();
    AFS_STATCNT(afs_readdir);

    code = afs_convert_code(afs_CreateReq(&treq, credp));
    crfree(credp);
    if (code)
        goto out1;

    afs_InitFakeStat(&fakestat);
    code = afs_convert_code(afs_EvalFakeStat(&avc, &fakestat, treq));
    if (code)
        goto out;

    /* update the cache entry */
  tagain:
    code = afs_convert_code(afs_VerifyVCache2(avc, treq));
    if (code)
        goto out;

    /* get a reference to the entire directory */
    tdc = afs_GetDCache(avc, (afs_size_t) 0, treq, &origOffset, &tlen, 1);
    len = tlen;
    if (!tdc) {
        code = -EIO;
        goto out;
    }
    ObtainWriteLock(&avc->lock, 811);
    ObtainReadLock(&tdc->lock);
    /*
     * Make sure that the data in the cache is current. There are two
     * cases we need to worry about:
     * 1. The cache data is being fetched by another process.
     * 2. The cache data is no longer valid
     */
    while ((avc->f.states & CStatd)
           && (tdc->dflags & DFFetching)
           && hsame(avc->f.m.DataVersion, tdc->f.versionNo)) {
        ReleaseReadLock(&tdc->lock);
        ReleaseWriteLock(&avc->lock);
        afs_osi_Sleep(&tdc->validPos);
        ObtainWriteLock(&avc->lock, 812);
        ObtainReadLock(&tdc->lock);
    }
    if (!(avc->f.states & CStatd)
        || !hsame(avc->f.m.DataVersion, tdc->f.versionNo)) {
        ReleaseReadLock(&tdc->lock);
        ReleaseWriteLock(&avc->lock);
        afs_PutDCache(tdc);
        goto tagain;
    }

    /* Set the readdir-in-progress flag, and downgrade the lock
     * to shared so others will be able to acquire a read lock.
     */
    avc->f.states |= CReadDir;
    avc->dcreaddir = tdc;
    avc->readdir_pid = MyPidxx2Pid(MyPidxx);
    ConvertWToSLock(&avc->lock);

    /* Fill in until we get an error or we're done. This implementation
     * takes an offset in units of blobs, rather than bytes.
     */
    code = 0;
#if defined(STRUCT_FILE_OPERATIONS_HAS_ITERATE)
    offset = ctx->pos;
#else
    offset = (int) fp->f_pos;
#endif
    while (1) {
        code = BlobScan(tdc, offset, &dirpos);
        if (code || !dirpos)
            break;

        code = afs_dir_GetVerifiedBlob(tdc, dirpos, &entry);
        if (code) {
            if (!(avc->f.states & CCorrupt)) {
                struct cell *tc = afs_GetCellStale(avc->f.fid.Cell, READ_LOCK);
                afs_warn("afs: Corrupt directory (%d.%d.%d.%d [%s] @%lx, pos %d)\n",
                         avc->f.fid.Cell, avc->f.fid.Fid.Volume,
                         avc->f.fid.Fid.Vnode, avc->f.fid.Fid.Unique,
                         tc ? tc->cellName : "",
                         (unsigned long)&tdc->f.inode, dirpos);
                if (tc)
                    afs_PutCell(tc, READ_LOCK);
                UpgradeSToWLock(&avc->lock, 814);
                avc->f.states |= CCorrupt;
            }
            code = -EIO;
            goto unlock_out;
        }

        de = (struct DirEntry *)entry.data;
        ino = afs_calc_inum (avc->f.fid.Cell, avc->f.fid.Fid.Volume,
                             ntohl(de->fid.vnode));
        len = strlen(de->name);

        /* filldir returns -EINVAL when the buffer is full. */
        {
            unsigned int type = DT_UNKNOWN;
            struct VenusFid afid;
            struct vcache *tvc;
            int vtype;
            afid.Cell = avc->f.fid.Cell;
            afid.Fid.Volume = avc->f.fid.Fid.Volume;
            afid.Fid.Vnode = ntohl(de->fid.vnode);
            afid.Fid.Unique = ntohl(de->fid.vunique);
            if ((avc->f.states & CForeign) == 0 && (ntohl(de->fid.vnode) & 1)) {
                type = DT_DIR;
            } else if ((tvc = afs_FindVCache(&afid, 0, 0))) {
                if (tvc->mvstat != AFS_MVSTAT_FILE) {
                    type = DT_DIR;
                } else if (((tvc->f.states) & (CStatd | CTruth))) {
                    /* CTruth will be set if the object has
                     *ever* been statd */
                    vtype = vType(tvc);
                    if (vtype == VDIR)
                        type = DT_DIR;
                    else if (vtype == VREG)
                        type = DT_REG;
                    /* Don't do this until we're sure it can't be a mtpt */
                    /* else if (vtype == VLNK)
                     * type=DT_LNK; */
                    /* what other types does AFS support? */
                }
                /* clean up from afs_FindVCache */
                afs_PutVCache(tvc);
            }
            /*
             * If this is NFS readdirplus, then the filler is going to
             * call getattr on this inode, which will deadlock if we're
             * holding the GLOCK.
             */
            AFS_GUNLOCK();
#if defined(STRUCT_FILE_OPERATIONS_HAS_ITERATE)
            /* dir_emit returns a bool - true when it succeeds.
             * Inverse the result to fit with how we check "code" */
            code = !dir_emit(ctx, de->name, len, ino, type);
#else
            code = (*filldir) (dirbuf, de->name, len, offset, ino, type);
#endif
            AFS_GLOCK();
        }
        DRelease(&entry, 0);
        if (code)
            break;
        offset = dirpos + 1 + ((len + 16) >> 5);
    }
    /* If filldir didn't fill in the last one this is still pointing to that
     * last attempt.
     */
    code = 0;

unlock_out:
#if defined(STRUCT_FILE_OPERATIONS_HAS_ITERATE)
    ctx->pos = (loff_t) offset;
#else
    fp->f_pos = (loff_t) offset;
#endif
    ReleaseReadLock(&tdc->lock);
    afs_PutDCache(tdc);
    UpgradeSToWLock(&avc->lock, 813);
    avc->f.states &= ~CReadDir;
    avc->dcreaddir = 0;
    avc->readdir_pid = 0;
    ReleaseSharedLock(&avc->lock);

out:
    afs_PutFakeStat(&fakestat);
    afs_DestroyReq(treq);
out1:
    AFS_GUNLOCK();
    return code;
}


/* in afs_pioctl.c */
extern int afs_xioctl(struct inode *ip, struct file *fp, unsigned int com,
                      unsigned long arg);

#if defined(HAVE_UNLOCKED_IOCTL) || defined(HAVE_COMPAT_IOCTL)
static long afs_unlocked_xioctl(struct file *fp, unsigned int com,
                               unsigned long arg) {
    return afs_xioctl(FILE_INODE(fp), fp, com, arg);

}
#endif


static int
afs_linux_mmap(struct file *fp, struct vm_area_struct *vmap)
{
    struct vcache *vcp = VTOAFS(FILE_INODE(fp));
    int code;

    AFS_GLOCK();
    afs_Trace3(afs_iclSetp, CM_TRACE_GMAP, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_POINTER, vmap->vm_start, ICL_TYPE_INT32,
               vmap->vm_end - vmap->vm_start);

    /* get a validated vcache entry */
    code = afs_linux_VerifyVCache(vcp, NULL);

    if (code == 0) {
        /* Linux's Flushpage implementation doesn't use credp, so optimise
         * our code to not need to crref() it */
        osi_FlushPages(vcp, NULL); /* ensure stale pages are gone */
        AFS_GUNLOCK();
        code = generic_file_mmap(fp, vmap);
        AFS_GLOCK();
        if (!code)
            vcp->f.states |= CMAPPED;
    }
    AFS_GUNLOCK();

    return code;
}

static int
afs_linux_open(struct inode *ip, struct file *fp)
{
    struct vcache *vcp = VTOAFS(ip);
    cred_t *credp = crref();
    int code;

    AFS_GLOCK();
    code = afs_open(&vcp, fp->f_flags, credp);
    AFS_GUNLOCK();

    crfree(credp);
    return afs_convert_code(code);
}

static int
afs_linux_release(struct inode *ip, struct file *fp)
{
    struct vcache *vcp = VTOAFS(ip);
    cred_t *credp = crref();
    int code = 0;

    AFS_GLOCK();
    code = afs_close(vcp, fp->f_flags, credp);
    ObtainWriteLock(&vcp->lock, 807);
    if (vcp->cred) {
        crfree(vcp->cred);
        vcp->cred = NULL;
    }
    ReleaseWriteLock(&vcp->lock);
    AFS_GUNLOCK();

    crfree(credp);
    return afs_convert_code(code);
}

static int
#if defined(FOP_FSYNC_TAKES_DENTRY)
afs_linux_fsync(struct file *fp, struct dentry *dp, int datasync)
#elif defined(FOP_FSYNC_TAKES_RANGE)
afs_linux_fsync(struct file *fp, loff_t start, loff_t end, int datasync)
#else
afs_linux_fsync(struct file *fp, int datasync)
#endif
{
    int code;
    struct inode *ip = FILE_INODE(fp);
    cred_t *credp = crref();

#if defined(FOP_FSYNC_TAKES_RANGE)
    afs_linux_lock_inode(ip);
#endif
    AFS_GLOCK();
    code = afs_fsync(VTOAFS(ip), credp);
    AFS_GUNLOCK();
#if defined(FOP_FSYNC_TAKES_RANGE)
    afs_linux_unlock_inode(ip);
#endif
    crfree(credp);
    return afs_convert_code(code);

}


static int
afs_linux_lock(struct file *fp, int cmd, struct file_lock *flp)
{
    int code = 0;
    struct vcache *vcp = VTOAFS(FILE_INODE(fp));
    cred_t *credp = crref();
    struct AFS_FLOCK flock;

    /* Convert to a lock format afs_lockctl understands. */
    memset(&flock, 0, sizeof(flock));
    flock.l_type = flp->fl_type;
    flock.l_pid = flp->fl_pid;
    flock.l_whence = 0;
    flock.l_start = flp->fl_start;
    if (flp->fl_end == OFFSET_MAX)
        flock.l_len = 0; /* Lock to end of file */
    else
        flock.l_len = flp->fl_end - flp->fl_start + 1;

    /* Safe because there are no large files, yet */
#if defined(F_GETLK64) && (F_GETLK != F_GETLK64)
    if (cmd == F_GETLK64)
        cmd = F_GETLK;
    else if (cmd == F_SETLK64)
        cmd = F_SETLK;
    else if (cmd == F_SETLKW64)
        cmd = F_SETLKW;
#endif /* F_GETLK64 && F_GETLK != F_GETLK64 */

    AFS_GLOCK();
    code = afs_convert_code(afs_lockctl(vcp, &flock, cmd, credp));
    AFS_GUNLOCK();

    if ((code == 0 || flp->fl_type == F_UNLCK) &&
        (cmd == F_SETLK || cmd == F_SETLKW)) {
        code = afs_posix_lock_file(fp, flp);
        if (code && flp->fl_type != F_UNLCK) {
            struct AFS_FLOCK flock2;
            flock2 = flock;
            flock2.l_type = F_UNLCK;
            AFS_GLOCK();
            afs_lockctl(vcp, &flock2, F_SETLK, credp);
            AFS_GUNLOCK();
        }
    }
    /* If lockctl says there are no conflicting locks, then also check with the
     * kernel, as lockctl knows nothing about byte range locks
     */
    if (code == 0 && cmd == F_GETLK && flock.l_type == F_UNLCK) {
        afs_posix_test_lock(fp, flp);
        /* If we found a lock in the kernel's structure, return it */
        if (flp->fl_type != F_UNLCK) {
            crfree(credp);
            return 0;
        }
    }

    /* Convert flock back to Linux's file_lock */
    flp->fl_type = flock.l_type;
    flp->fl_pid = flock.l_pid;
    flp->fl_start = flock.l_start;
    if (flock.l_len == 0)
        flp->fl_end = OFFSET_MAX; /* Lock to end of file */
    else
        flp->fl_end = flock.l_start + flock.l_len - 1;

    crfree(credp);
    return code;
}

#ifdef STRUCT_FILE_OPERATIONS_HAS_FLOCK
static int
afs_linux_flock(struct file *fp, int cmd, struct file_lock *flp) {
    int code = 0;
    struct vcache *vcp = VTOAFS(FILE_INODE(fp));
    cred_t *credp = crref();
    struct AFS_FLOCK flock;
    /* Convert to a lock format afs_lockctl understands. */
    memset(&flock, 0, sizeof(flock));
    flock.l_type = flp->fl_type;
    flock.l_pid = flp->fl_pid;
    flock.l_whence = 0;
    flock.l_start = 0;
    flock.l_len = 0;

    /* Safe because there are no large files, yet */
#if defined(F_GETLK64) && (F_GETLK != F_GETLK64)
    if (cmd == F_GETLK64)
        cmd = F_GETLK;
    else if (cmd == F_SETLK64)
        cmd = F_SETLK;
    else if (cmd == F_SETLKW64)
        cmd = F_SETLKW;
#endif /* F_GETLK64 && F_GETLK != F_GETLK64 */

    AFS_GLOCK();
    code = afs_convert_code(afs_lockctl(vcp, &flock, cmd, credp));
    AFS_GUNLOCK();

    if ((code == 0 || flp->fl_type == F_UNLCK) &&
        (cmd == F_SETLK || cmd == F_SETLKW)) {
        flp->fl_flags &=~ FL_SLEEP;
        code = flock_lock_file_wait(fp, flp);
        if (code && flp->fl_type != F_UNLCK) {
            struct AFS_FLOCK flock2;
            flock2 = flock;
            flock2.l_type = F_UNLCK;
            AFS_GLOCK();
            afs_lockctl(vcp, &flock2, F_SETLK, credp);
            AFS_GUNLOCK();
        }
    }
    /* Convert flock back to Linux's file_lock */
    flp->fl_type = flock.l_type;
    flp->fl_pid = flock.l_pid;

    crfree(credp);
    return code;
}
#endif

/* afs_linux_flush
 * essentially the same as afs_fsync() but we need to get the return
 * code for the sys_close() here, not afs_linux_release(), so call
 * afs_StoreAllSegments() with AFS_LASTSTORE
 */
static int
#if defined(FOP_FLUSH_TAKES_FL_OWNER_T)
afs_linux_flush(struct file *fp, fl_owner_t id)
#else
afs_linux_flush(struct file *fp)
#endif
{
    struct vrequest *treq = NULL;
    struct vcache *vcp;
    cred_t *credp;
    int code;
    int bypasscache = 0;

    AFS_GLOCK();

    if ((fp->f_flags & O_ACCMODE) == O_RDONLY) { /* readers dont flush */
        AFS_GUNLOCK();
        return 0;
    }

    AFS_DISCON_LOCK();

    credp = crref();
    vcp = VTOAFS(FILE_INODE(fp));

    code = afs_CreateReq(&treq, credp);
    if (code)
        goto out;
    /* If caching is bypassed for this file, or globally, just return 0 */
    if (cache_bypass_strategy == ALWAYS_BYPASS_CACHE)
        bypasscache = 1;
    else {
        ObtainReadLock(&vcp->lock);
        if (vcp->cachingStates & FCSBypass)
            bypasscache = 1;
        ReleaseReadLock(&vcp->lock);
    }
    if (bypasscache) {
        /* future proof: don't rely on 0 return from afs_InitReq */
        code = 0;
        goto out;
    }

    ObtainSharedLock(&vcp->lock, 535);
    if ((vcp->execsOrWriters > 0) && (file_count(fp) == 1)) {
        UpgradeSToWLock(&vcp->lock, 536);
        if (!AFS_IS_DISCONNECTED) {
                code = afs_StoreAllSegments(vcp,
                                treq,
                                AFS_SYNC | AFS_LASTSTORE);
        } else {
                afs_DisconAddDirty(vcp, VDisconWriteOsiFlush, 1);
        }
        ConvertWToSLock(&vcp->lock);
    }
    code = afs_CheckCode(code, treq, 54);
    ReleaseSharedLock(&vcp->lock);

out:
    afs_DestroyReq(treq);
    AFS_DISCON_UNLOCK();
    AFS_GUNLOCK();

    crfree(credp);
    return afs_convert_code(code);
}

struct file_operations afs_dir_fops = {
  .read =       generic_read_dir,
#if defined(STRUCT_FILE_OPERATIONS_HAS_ITERATE)
  .iterate =    afs_linux_readdir,
#else
  .readdir =    afs_linux_readdir,
#endif
#ifdef HAVE_UNLOCKED_IOCTL
  .unlocked_ioctl = afs_unlocked_xioctl,
#else
  .ioctl =      afs_xioctl,
#endif
#ifdef HAVE_COMPAT_IOCTL
  .compat_ioctl = afs_unlocked_xioctl,
#endif
  .open =       afs_linux_open,
  .release =    afs_linux_release,
  .llseek =     default_llseek,
#ifdef HAVE_LINUX_NOOP_FSYNC
  .fsync =      noop_fsync,
#else
  .fsync =      simple_sync_file,
#endif
};

struct file_operations afs_file_fops = {
#ifdef STRUCT_FILE_OPERATIONS_HAS_READ_ITER
  .read_iter =  afs_linux_read_iter,
  .write_iter = afs_linux_write_iter,
# if !defined(HAVE_LINUX___VFS_WRITE)
  .read =       new_sync_read,
  .write =      new_sync_write,
# endif
#elif defined(HAVE_LINUX_GENERIC_FILE_AIO_READ)
  .aio_read =   afs_linux_aio_read,
  .aio_write =  afs_linux_aio_write,
  .read =       do_sync_read,
  .write =      do_sync_write,
#else
  .read =       afs_linux_read,
  .write =      afs_linux_write,
#endif
#ifdef HAVE_UNLOCKED_IOCTL
  .unlocked_ioctl = afs_unlocked_xioctl,
#else
  .ioctl =      afs_xioctl,
#endif
#ifdef HAVE_COMPAT_IOCTL
  .compat_ioctl = afs_unlocked_xioctl,
#endif
  .mmap =       afs_linux_mmap,
  .open =       afs_linux_open,
  .flush =      afs_linux_flush,
#if defined(STRUCT_FILE_OPERATIONS_HAS_SENDFILE)
  .sendfile =   generic_file_sendfile,
#endif
#if defined(STRUCT_FILE_OPERATIONS_HAS_SPLICE) && !defined(HAVE_LINUX_DEFAULT_FILE_SPLICE_READ)
# if defined(HAVE_LINUX_ITER_FILE_SPLICE_WRITE)
  .splice_write = iter_file_splice_write,
# else
  .splice_write = generic_file_splice_write,
# endif
  .splice_read = generic_file_splice_read,
#endif
  .release =    afs_linux_release,
  .fsync =      afs_linux_fsync,
  .lock =       afs_linux_lock,
#ifdef STRUCT_FILE_OPERATIONS_HAS_FLOCK
  .flock =      afs_linux_flock,
#endif
  .llseek =     default_llseek,
};

static struct dentry *
canonical_dentry(struct inode *ip)
{
    struct vcache *vcp = VTOAFS(ip);
    struct dentry *first = NULL, *ret = NULL, *cur;
#if defined(D_ALIAS_IS_HLIST) && !defined(HLIST_ITERATOR_NO_NODE)
    struct hlist_node *p;
#endif

    /* general strategy:
     * if vcp->target_link is set, and can be found in ip->i_dentry, use that.
     * otherwise, use the first dentry in ip->i_dentry.
     * if ip->i_dentry is empty, use the 'dentry' argument we were given.
     */
    /* note that vcp->target_link specifies which dentry to use, but we have
     * no reference held on that dentry. so, we cannot use or dereference
     * vcp->target_link itself, since it may have been freed. instead, we only
     * use it to compare to pointers in the ip->i_dentry list. */

    d_prune_aliases(ip);

# ifdef HAVE_DCACHE_LOCK
    spin_lock(&dcache_lock);
# else
    spin_lock(&ip->i_lock);
# endif

#if defined(D_ALIAS_IS_HLIST)
# if defined(HLIST_ITERATOR_NO_NODE)
    hlist_for_each_entry(cur, &ip->i_dentry, d_alias) {
# else
    hlist_for_each_entry(cur, p, &ip->i_dentry, d_alias) {
# endif
#else
    list_for_each_entry_reverse(cur, &ip->i_dentry, d_alias) {
#endif

        if (!vcp->target_link || cur == vcp->target_link) {
            ret = cur;
            break;
        }

        if (!first) {
            first = cur;
        }
    }
    if (!ret && first) {
        ret = first;
    }

    vcp->target_link = ret;

# ifdef HAVE_DCACHE_LOCK
    if (ret) {
        dget_locked(ret);
    }
    spin_unlock(&dcache_lock);
# else
    if (ret) {
        dget(ret);
    }
    spin_unlock(&ip->i_lock);
# endif

    return ret;
}

/**********************************************************************
 * AFS Linux dentry operations
 **********************************************************************/

/* afs_linux_revalidate
 * Ensure vcache is stat'd before use. Return 0 if entry is valid.
 */
static int
afs_linux_revalidate(struct dentry *dp)
{
    struct vattr *vattr = NULL;
    struct vcache *vcp = VTOAFS(dp->d_inode);
    cred_t *credp;
    int code;

    if (afs_shuttingdown != AFS_RUNNING)
        return EIO;

    AFS_GLOCK();

    code = afs_CreateAttr(&vattr);
    if (code) {
        goto out;
    }

    /* This avoids the crref when we don't have to do it. Watch for
     * changes in afs_getattr that don't get replicated here!
     */
    if (vcp->f.states & CStatd &&
        (!afs_fakestat_enable || vcp->mvstat != AFS_MVSTAT_MTPT) &&
        !afs_nfsexporter &&
        (vType(vcp) == VDIR || vType(vcp) == VLNK)) {
        code = afs_CopyOutAttrs(vcp, vattr);
    } else {
        credp = crref();
        code = afs_getattr(vcp, vattr, credp);
        crfree(credp);
    }

    if (!code)
        afs_fill_inode(AFSTOV(vcp), vattr);

    afs_DestroyAttr(vattr);

out:
    AFS_GUNLOCK();

    return afs_convert_code(code);
}

/* vattr_setattr
 * Set iattr data into vattr. Assume vattr cleared before call.
 */
static void
iattr2vattr(struct vattr *vattrp, struct iattr *iattrp)
{
    vattrp->va_mask = iattrp->ia_valid;
    if (iattrp->ia_valid & ATTR_MODE)
        vattrp->va_mode = iattrp->ia_mode;
    if (iattrp->ia_valid & ATTR_UID)
        vattrp->va_uid = afs_from_kuid(iattrp->ia_uid);
    if (iattrp->ia_valid & ATTR_GID)
        vattrp->va_gid = afs_from_kgid(iattrp->ia_gid);
    if (iattrp->ia_valid & ATTR_SIZE)
        vattrp->va_size = iattrp->ia_size;
    if (iattrp->ia_valid & ATTR_ATIME) {
        vattrp->va_atime.tv_sec = iattrp->ia_atime.tv_sec;
        vattrp->va_atime.tv_usec = 0;
    }
    if (iattrp->ia_valid & ATTR_MTIME) {
        vattrp->va_mtime.tv_sec = iattrp->ia_mtime.tv_sec;
        vattrp->va_mtime.tv_usec = 0;
    }
    if (iattrp->ia_valid & ATTR_CTIME) {
        vattrp->va_ctime.tv_sec = iattrp->ia_ctime.tv_sec;
        vattrp->va_ctime.tv_usec = 0;
    }
}

/* vattr2inode
 * Rewrite the inode cache from the attr. Assumes all vattr fields are valid.
 */
void
vattr2inode(struct inode *ip, struct vattr *vp)
{
    ip->i_ino = vp->va_nodeid;
#ifdef HAVE_LINUX_SET_NLINK
    set_nlink(ip, vp->va_nlink);
#else
    ip->i_nlink = vp->va_nlink;
#endif
    ip->i_blocks = vp->va_blocks;
#ifdef STRUCT_INODE_HAS_I_BLKBITS
    ip->i_blkbits = AFS_BLKBITS;
#endif
#ifdef STRUCT_INODE_HAS_I_BLKSIZE
    ip->i_blksize = vp->va_blocksize;
#endif
    ip->i_rdev = vp->va_rdev;
    ip->i_mode = vp->va_mode;
    ip->i_uid = afs_make_kuid(vp->va_uid);
    ip->i_gid = afs_make_kgid(vp->va_gid);
    i_size_write(ip, vp->va_size);
    ip->i_atime.tv_sec = vp->va_atime.tv_sec;
    ip->i_atime.tv_nsec = 0;
    ip->i_mtime.tv_sec = vp->va_mtime.tv_sec;
    /* Set the mtime nanoseconds to the sysname generation number.
     * This convinces NFS clients that all directories have changed
     * any time the sysname list changes.
     */
    ip->i_mtime.tv_nsec = afs_sysnamegen;
    ip->i_ctime.tv_sec = vp->va_ctime.tv_sec;
    ip->i_ctime.tv_nsec = 0;
}

/* afs_notify_change
 * Linux version of setattr call. What to change is in the iattr struct.
 * We need to set bits in both the Linux inode as well as the vcache.
 */
static int
afs_notify_change(struct dentry *dp, struct iattr *iattrp)
{
    struct vattr *vattr = NULL;
    cred_t *credp = crref();
    struct inode *ip = dp->d_inode;
    int code;

    AFS_GLOCK();
    code = afs_CreateAttr(&vattr);
    if (code) {
        goto out;
    }

    iattr2vattr(vattr, iattrp); /* Convert for AFS vnodeops call. */

    code = afs_setattr(VTOAFS(ip), vattr, credp);
    if (!code) {
        afs_getattr(VTOAFS(ip), vattr, credp);
        vattr2inode(ip, vattr);
    }
    afs_DestroyAttr(vattr);

out:
    AFS_GUNLOCK();
    crfree(credp);
    return afs_convert_code(code);
}

#if defined(IOP_GETATTR_TAKES_PATH_STRUCT)
static int
afs_linux_getattr(const struct path *path, struct kstat *stat, u32 request_mask, unsigned int sync_mode)
{
        int err = afs_linux_revalidate(path->dentry);
        if (!err) {
                generic_fillattr(path->dentry->d_inode, stat);
        }
        return err;
}
#else
static int
afs_linux_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
{
        int err = afs_linux_revalidate(dentry);
        if (!err) {
                generic_fillattr(dentry->d_inode, stat);
        }
        return err;
}
#endif

static afs_uint32
parent_vcache_dv(struct inode *inode, cred_t *credp)
{
    int free_cred = 0;
    struct vcache *pvcp;

    /*
     * If parent is a mount point and we are using fakestat, we may need
     * to look at the fake vcache entry instead of what the vfs is giving
     * us.  The fake entry is the one with the useful DataVersion.
     */
    pvcp = VTOAFS(inode);
    if (pvcp->mvstat == AFS_MVSTAT_MTPT && afs_fakestat_enable) {
        struct vrequest treq;
        struct afs_fakestat_state fakestate;

        if (!credp) {
            credp = crref();
            free_cred = 1;
        }
        afs_InitReq(&treq, credp);
        afs_InitFakeStat(&fakestate);
        afs_TryEvalFakeStat(&pvcp, &fakestate, &treq);
        if (free_cred)
            crfree(credp);
        afs_PutFakeStat(&fakestate);
    }
    return hgetlo(pvcp->f.m.DataVersion);
}

#ifdef D_SPLICE_ALIAS_RACE
/* Leave some trace that this code is enabled; otherwise it's pretty hard to
 * tell. */
static __attribute__((used)) const char dentry_race_marker[] = "d_splice_alias race workaround enabled";

static int
check_dentry_race(struct dentry *dp)
{
    int raced = 0;
    if (!dp->d_inode) {
        struct dentry *parent = dget_parent(dp);

        /* In Linux, before commit 4919c5e45a91b5db5a41695fe0357fbdff0d5767,
         * d_splice_alias can momentarily hash a dentry before it's fully
         * populated. This only happens for a moment, since it's unhashed again
         * right after (in d_move), but this can make the dentry be found by
         * __d_lookup, and then given to us.
         *
         * So check if the dentry is unhashed; if it is, then the dentry is not
         * valid. We lock the parent inode to ensure that d_splice_alias is no
         * longer running (the inode mutex will be held during
         * afs_linux_lookup). Locking d_lock is required to check the dentry's
         * flags, so lock that, too.
         */
        afs_linux_lock_inode(parent->d_inode);
        spin_lock(&dp->d_lock);
        if (d_unhashed(dp)) {
            raced = 1;
        }
        spin_unlock(&dp->d_lock);
        afs_linux_unlock_inode(parent->d_inode);

        dput(parent);
    }
    return raced;
}
#endif /* D_SPLICE_ALIAS_RACE */

/* Validate a dentry. Return 1 if unchanged, 0 if VFS layer should re-evaluate.
 * In kernels 2.2.10 and above, we are passed an additional flags var which
 * may have either the LOOKUP_FOLLOW OR LOOKUP_DIRECTORY set in which case
 * we are advised to follow the entry if it is a link or to make sure that
 * it is a directory. But since the kernel itself checks these possibilities
 * later on, we shouldn't have to do it until later. Perhaps in the future..
 *
 * The code here assumes that on entry the global lock is not held
 */
static int
#if defined(DOP_REVALIDATE_TAKES_UNSIGNED)
afs_linux_dentry_revalidate(struct dentry *dp, unsigned int flags)
#elif defined(DOP_REVALIDATE_TAKES_NAMEIDATA)
afs_linux_dentry_revalidate(struct dentry *dp, struct nameidata *nd)
#else
afs_linux_dentry_revalidate(struct dentry *dp, int flags)
#endif
{
    cred_t *credp = NULL;
    struct vcache *vcp, *pvcp, *tvc = NULL;
    struct dentry *parent;
    int valid;
    struct afs_fakestat_state fakestate;
    int force_drop = 0;
    afs_uint32 parent_dv;

#ifdef LOOKUP_RCU
    /* We don't support RCU path walking */
# if defined(DOP_REVALIDATE_TAKES_UNSIGNED)
    if (flags & LOOKUP_RCU)
# else
    if (nd->flags & LOOKUP_RCU)
# endif
       return -ECHILD;
#endif

#ifdef D_SPLICE_ALIAS_RACE
    if (check_dentry_race(dp)) {
        valid = 0;
        return valid;
    }
#endif

    AFS_GLOCK();
    afs_InitFakeStat(&fakestate);

    if (dp->d_inode) {
        vcp = VTOAFS(dp->d_inode);

        if (vcp == afs_globalVp)
            goto good_dentry;

        if (vcp->mvstat == AFS_MVSTAT_MTPT) {
            if (vcp->mvid.target_root && (vcp->f.states & CMValid)) {
                int tryEvalOnly = 0;
                int code = 0;
                struct vrequest *treq = NULL;

                credp = crref();

                code = afs_CreateReq(&treq, credp);
                if (code) {
                    goto bad_dentry;
                }
                if ((strcmp(dp->d_name.name, ".directory") == 0)) {
                    tryEvalOnly = 1;
                }
                if (tryEvalOnly)
                    code = afs_TryEvalFakeStat(&vcp, &fakestate, treq);
                else
                    code = afs_EvalFakeStat(&vcp, &fakestate, treq);
                afs_DestroyReq(treq);
                if ((tryEvalOnly && vcp->mvstat == AFS_MVSTAT_MTPT) || code) {
                    /* a mount point, not yet replaced by its directory */
                    goto bad_dentry;
                }
            }
        } else if (vcp->mvstat == AFS_MVSTAT_ROOT && *dp->d_name.name != '/') {
            osi_Assert(vcp->mvid.parent != NULL);
        }

#ifdef notdef
        /* If the last looker changes, we should make sure the current
         * looker still has permission to examine this file.  This would
         * always require a crref() which would be "slow".
         */
        if (vcp->last_looker != treq.uid) {
            if (!afs_AccessOK(vcp, (vType(vcp) == VREG) ? PRSFS_READ : PRSFS_LOOKUP, &treq, CHECK_MODE_BITS)) {
                goto bad_dentry;
            }

            vcp->last_looker = treq.uid;
        }
#endif

        parent = dget_parent(dp);
        pvcp = VTOAFS(parent->d_inode);
        parent_dv = parent_vcache_dv(parent->d_inode, credp);

        /* If the parent's DataVersion has changed or the vnode
         * is longer valid, we need to do a full lookup.  VerifyVCache
         * isn't enough since the vnode may have been renamed.
         */

        if (parent_dv > dp->d_time || !(vcp->f.states & CStatd)) {
            struct vattr *vattr = NULL;
            int code;
            int lookup_good;

            if (credp == NULL) {
                credp = crref();
            }
            code = afs_lookup(pvcp, (char *)dp->d_name.name, &tvc, credp);

            if (code) {
                /* We couldn't perform the lookup, so we're not okay. */
                lookup_good = 0;

            } else if (tvc == vcp) {
                /* We got back the same vcache, so we're good. */
                lookup_good = 1;

            } else if (tvc == VTOAFS(dp->d_inode)) {
                /* We got back the same vcache, so we're good. This is
                 * different from the above case, because sometimes 'vcp' is
                 * not the same as the vcache for dp->d_inode, if 'vcp' was a
                 * mtpt and we evaluated it to a root dir. In rare cases,
                 * afs_lookup might not evalute the mtpt when we do, or vice
                 * versa, so the previous case will not succeed. But this is
                 * still 'correct', so make sure not to mark the dentry as
                 * invalid; it still points to the same thing! */
                lookup_good = 1;

            } else {
                /* We got back a different file, so we're definitely not
                 * okay. */
                lookup_good = 0;
            }

            if (!lookup_good) {
                dput(parent);
                /* Force unhash; the name doesn't point to this file
                 * anymore. */
                force_drop = 1;
                if (code && code != ENOENT) {
                    /* ...except if we couldn't perform the actual lookup,
                     * we don't know if the name points to this file or not. */
                    force_drop = 0;
                }
                goto bad_dentry;
            }

            code = afs_CreateAttr(&vattr);
            if (code) {
                dput(parent);
                goto bad_dentry;
            }

            if (afs_getattr(vcp, vattr, credp)) {
                dput(parent);
                afs_DestroyAttr(vattr);
                goto bad_dentry;
            }

            vattr2inode(AFSTOV(vcp), vattr);
            dp->d_time = parent_dv;

            afs_DestroyAttr(vattr);
        }

        /* should we always update the attributes at this point? */
        /* unlikely--the vcache entry hasn't changed */

        dput(parent);

    } else {

        /* 'dp' represents a cached negative lookup. */

        parent = dget_parent(dp);
        pvcp = VTOAFS(parent->d_inode);
        parent_dv = parent_vcache_dv(parent->d_inode, credp);

        if (parent_dv > dp->d_time || !(pvcp->f.states & CStatd)
            || afs_IsDynroot(pvcp)) {
            dput(parent);
            goto bad_dentry;
        }

        dput(parent);
    }

  good_dentry:
    valid = 1;
    goto done;

  bad_dentry:
    valid = 0;
#ifndef D_INVALIDATE_IS_VOID
    /* When (v3.18) d_invalidate was converted to void, it also started
     * being called automatically from revalidate, and automatically
     * handled:
     *  - shrink_dcache_parent
     *  - automatic detach of submounts
     *  - d_drop
     * Therefore, after that point, OpenAFS revalidate logic no longer needs
     * to do any of those things itself for invalid dentry structs.  We only need
     * to tell VFS it's invalid (by returning 0), and VFS will handle the rest.
     */
    if (have_submounts(dp))
        valid = 1;
#endif

  done:
    /* Clean up */
    if (tvc)
        afs_PutVCache(tvc);
    afs_PutFakeStat(&fakestate);
    AFS_GUNLOCK();
    if (credp)
        crfree(credp);

#ifndef D_INVALIDATE_IS_VOID
    if (!valid) {
        /*
         * If we had a negative lookup for the name we want to forcibly
         * unhash the dentry.
         * Otherwise use d_invalidate which will not unhash it if still in use.
         */
        if (force_drop) {
            shrink_dcache_parent(dp);
            d_drop(dp);
        } else
            d_invalidate(dp);
    }
#endif
    return valid;

}

static void
afs_dentry_iput(struct dentry *dp, struct inode *ip)
{
    struct vcache *vcp = VTOAFS(ip);

    AFS_GLOCK();
    if (!AFS_IS_DISCONNECTED || (vcp->f.states & CUnlinked)) {
        (void) afs_InactiveVCache(vcp, NULL);
    }
    AFS_GUNLOCK();
    afs_linux_clear_nfsfs_renamed(dp);

    iput(ip);
}

static int
#if defined(DOP_D_DELETE_TAKES_CONST)
afs_dentry_delete(const struct dentry *dp)
#else
afs_dentry_delete(struct dentry *dp)
#endif
{
    if (dp->d_inode && (VTOAFS(dp->d_inode)->f.states & CUnlinked))
        return 1;               /* bad inode? */

    return 0;
}

#ifdef STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT
static struct vfsmount *
afs_dentry_automount(afs_linux_path_t *path)
{
    struct dentry *target;

    /*
     * Avoid symlink resolution limits when resolving; we cannot contribute to
     * an infinite symlink loop.
     *
     * On newer kernels the field has moved to the private nameidata structure
     * so we can't adjust it here.  This may cause ELOOP when using a path with
     * 40 or more directories that are not already in the dentry cache.
     */
#if defined(STRUCT_TASK_STRUCT_HAS_TOTAL_LINK_COUNT)
    current->total_link_count--;
#endif

    target = canonical_dentry(path->dentry->d_inode);

    if (target == path->dentry) {
        dput(target);
        target = NULL;
    }

    if (target) {
        dput(path->dentry);
        path->dentry = target;

    } else {
        spin_lock(&path->dentry->d_lock);
        path->dentry->d_flags &= ~DCACHE_NEED_AUTOMOUNT;
        spin_unlock(&path->dentry->d_lock);
    }

    return NULL;
}
#endif /* STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT */

struct dentry_operations afs_dentry_operations = {
  .d_revalidate =       afs_linux_dentry_revalidate,
  .d_delete =           afs_dentry_delete,
  .d_iput =             afs_dentry_iput,
#ifdef STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT
  .d_automount =        afs_dentry_automount,
#endif /* STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT */
};

/**********************************************************************
 * AFS Linux inode operations
 **********************************************************************/

/* afs_linux_create
 *
 * Merely need to set enough of vattr to get us through the create. Note
 * that the higher level code (open_namei) will take care of any tuncation
 * explicitly. Exclusive open is also taken care of in open_namei.
 *
 * name is in kernel space at this point.
 */
static int
#if defined(IOP_CREATE_TAKES_BOOL)
afs_linux_create(struct inode *dip, struct dentry *dp, umode_t mode,
                 bool excl)
#elif defined(IOP_CREATE_TAKES_UMODE_T)
afs_linux_create(struct inode *dip, struct dentry *dp, umode_t mode,
                 struct nameidata *nd)
#elif defined(IOP_CREATE_TAKES_NAMEIDATA)
afs_linux_create(struct inode *dip, struct dentry *dp, int mode,
                 struct nameidata *nd)
#else
afs_linux_create(struct inode *dip, struct dentry *dp, int mode)
#endif
{
    struct vattr *vattr = NULL;
    cred_t *credp = crref();
    const char *name = dp->d_name.name;
    struct vcache *vcp;
    int code;

    AFS_GLOCK();

    code = afs_CreateAttr(&vattr);
    if (code) {
        goto out;
    }
    vattr->va_mode = mode;
    vattr->va_type = mode & S_IFMT;

    code = afs_create(VTOAFS(dip), (char *)name, vattr, NONEXCL, mode,
                      &vcp, credp);

    if (!code) {
        struct inode *ip = AFSTOV(vcp);

        afs_getattr(vcp, vattr, credp);
        afs_fill_inode(ip, vattr);
        insert_inode_hash(ip);
#if !defined(STRUCT_SUPER_BLOCK_HAS_S_D_OP)
        dp->d_op = &afs_dentry_operations;
#endif
        dp->d_time = parent_vcache_dv(dip, credp);
        d_instantiate(dp, ip);
    }

    afs_DestroyAttr(vattr);

out:
    AFS_GUNLOCK();

    crfree(credp);
    return afs_convert_code(code);
}

/* afs_linux_lookup */
static struct dentry *
#if defined(IOP_LOOKUP_TAKES_UNSIGNED)
afs_linux_lookup(struct inode *dip, struct dentry *dp,
                 unsigned flags)
#elif defined(IOP_LOOKUP_TAKES_NAMEIDATA)
afs_linux_lookup(struct inode *dip, struct dentry *dp,
                 struct nameidata *nd)
#else
afs_linux_lookup(struct inode *dip, struct dentry *dp)
#endif
{
    cred_t *credp = crref();
    struct vcache *vcp = NULL;
    const char *comp = dp->d_name.name;
    struct inode *ip = NULL;
    struct dentry *newdp = NULL;
    int code;

    AFS_GLOCK();

    code = afs_lookup(VTOAFS(dip), (char *)comp, &vcp, credp);
    if (code == ENOENT) {
        /* It's ok for the file to not be found. That's noted by the caller by
         * seeing that the dp->d_inode field is NULL (set by d_splice_alias or
         * d_add, below). */
        code = 0;
        osi_Assert(vcp == NULL);
    }
    if (code) {
        AFS_GUNLOCK();
        goto done;
    }

    if (vcp) {
        struct vattr *vattr = NULL;
        struct vcache *parent_vc = VTOAFS(dip);

        if (parent_vc == vcp) {
            /* This is possible if the parent dir is a mountpoint to a volume,
             * and the dir entry we looked up is a mountpoint to the same
             * volume. Linux cannot cope with this, so return an error instead
             * of risking a deadlock or panic. */
            afs_PutVCache(vcp);
            code = EDEADLK;
            AFS_GUNLOCK();
            goto done;
        }

        code = afs_CreateAttr(&vattr);
        if (code) {
            afs_PutVCache(vcp);
            AFS_GUNLOCK();
            goto done;
        }

        ip = AFSTOV(vcp);
        afs_getattr(vcp, vattr, credp);
        afs_fill_inode(ip, vattr);
        if (hlist_unhashed(&ip->i_hash))
            insert_inode_hash(ip);

        afs_DestroyAttr(vattr);
    }
#if !defined(STRUCT_SUPER_BLOCK_HAS_S_D_OP)
    dp->d_op = &afs_dentry_operations;
#endif
    dp->d_time = parent_vcache_dv(dip, credp);

    AFS_GUNLOCK();

    if (ip && S_ISDIR(ip->i_mode)) {
        d_prune_aliases(ip);

#ifdef STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT
        /* Only needed if this is a volume root */
        if (vcp->mvstat == 2)
            ip->i_flags |= S_AUTOMOUNT;
#endif
    }
    /*
     * Take an extra reference so the inode doesn't go away if
     * d_splice_alias drops our reference on error.
     */
    if (ip)
#ifdef HAVE_LINUX_IHOLD
        ihold(ip);
#else
        igrab(ip);
#endif

    newdp = d_splice_alias(ip, dp);

 done:
    crfree(credp);

    if (IS_ERR(newdp)) {
        /* d_splice_alias can return an error (EIO) if there is an existing
         * connected directory alias for this dentry. Add our dentry manually
         * ourselves if this happens. */
        d_add(dp, ip);

#if defined(D_SPLICE_ALIAS_LEAK_ON_ERROR)
        /* Depending on the kernel version, d_splice_alias may or may not drop
         * the inode reference on error. If it didn't, do it here. */
        iput(ip);
#endif
        return NULL;
    }

    if (code) {
        if (ip)
            iput(ip);
        return ERR_PTR(afs_convert_code(code));
    }

    iput(ip);
    return newdp;
}

static int
afs_linux_link(struct dentry *olddp, struct inode *dip, struct dentry *newdp)
{
    int code;
    cred_t *credp = crref();
    const char *name = newdp->d_name.name;
    struct inode *oldip = olddp->d_inode;

    /* If afs_link returned the vnode, we could instantiate the
     * dentry. Since it's not, we drop this one and do a new lookup.
     */
    d_drop(newdp);

    AFS_GLOCK();
    code = afs_link(VTOAFS(oldip), VTOAFS(dip), (char *)name, credp);

    AFS_GUNLOCK();
    crfree(credp);
    return afs_convert_code(code);
}

/* We have to have a Linux specific sillyrename function, because we
 * also have to keep the dcache up to date when we're doing a silly
 * rename - so we don't want the generic vnodeops doing this behind our
 * back.
 */

static int
afs_linux_sillyrename(struct inode *dir, struct dentry *dentry,
                      cred_t *credp)
{
    struct vcache *tvc = VTOAFS(dentry->d_inode);
    struct dentry *__dp = NULL;
    char *__name = NULL;
    int code;

    if (afs_linux_nfsfs_renamed(dentry))
        return EBUSY;

    do {
        dput(__dp);

        AFS_GLOCK();
        if (__name)
            osi_FreeSmallSpace(__name);
        __name = afs_newname();
        AFS_GUNLOCK();

        __dp = lookup_one_len(__name, dentry->d_parent, strlen(__name));

        if (IS_ERR(__dp)) {
            osi_FreeSmallSpace(__name);
            return EBUSY;
        }
    } while (__dp->d_inode != NULL);

    AFS_GLOCK();
    code = afs_rename(VTOAFS(dir), (char *)dentry->d_name.name,
                      VTOAFS(dir), (char *)__dp->d_name.name,
                      credp);
    if (!code) {
        tvc->mvid.silly_name = __name;
        crhold(credp);
        if (tvc->uncred) {
            crfree(tvc->uncred);
        }
        tvc->uncred = credp;
        tvc->f.states |= CUnlinked;
        afs_linux_set_nfsfs_renamed(dentry);

        __dp->d_time = 0;               /* force to revalidate */
        d_move(dentry, __dp);
    } else {
        osi_FreeSmallSpace(__name);
    }
    AFS_GUNLOCK();

    dput(__dp);

    return code;
}


static int
afs_linux_unlink(struct inode *dip, struct dentry *dp)
{
    int code = EBUSY;
    cred_t *credp = crref();
    const char *name = dp->d_name.name;
    struct vcache *tvc = VTOAFS(dp->d_inode);

    if (VREFCOUNT(tvc) > 1 && tvc->opens > 0
                                && !(tvc->f.states & CUnlinked)) {

        code = afs_linux_sillyrename(dip, dp, credp);
    } else {
        AFS_GLOCK();
        code = afs_remove(VTOAFS(dip), (char *)name, credp);
        AFS_GUNLOCK();
        if (!code)
            d_drop(dp);
    }

    crfree(credp);
    return afs_convert_code(code);
}


static int
afs_linux_symlink(struct inode *dip, struct dentry *dp, const char *target)
{
    int code;
    cred_t *credp = crref();
    struct vattr *vattr = NULL;
    const char *name = dp->d_name.name;

    /* If afs_symlink returned the vnode, we could instantiate the
     * dentry. Since it's not, we drop this one and do a new lookup.
     */
    d_drop(dp);

    AFS_GLOCK();
    code = afs_CreateAttr(&vattr);
    if (code) {
        goto out;
    }

    code = afs_symlink(VTOAFS(dip), (char *)name, vattr, (char *)target, NULL,
                        credp);
    afs_DestroyAttr(vattr);

out:
    AFS_GUNLOCK();
    crfree(credp);
    return afs_convert_code(code);
}

static int
#if defined(IOP_MKDIR_TAKES_UMODE_T)
afs_linux_mkdir(struct inode *dip, struct dentry *dp, umode_t mode)
#else
afs_linux_mkdir(struct inode *dip, struct dentry *dp, int mode)
#endif
{
    int code;
    cred_t *credp = crref();
    struct vcache *tvcp = NULL;
    struct vattr *vattr = NULL;
    const char *name = dp->d_name.name;

    AFS_GLOCK();
    code = afs_CreateAttr(&vattr);
    if (code) {
        goto out;
    }

    vattr->va_mask = ATTR_MODE;
    vattr->va_mode = mode;

    code = afs_mkdir(VTOAFS(dip), (char *)name, vattr, &tvcp, credp);

    if (tvcp) {
        struct inode *ip = AFSTOV(tvcp);

        afs_getattr(tvcp, vattr, credp);
        afs_fill_inode(ip, vattr);

#if !defined(STRUCT_SUPER_BLOCK_HAS_S_D_OP)
        dp->d_op = &afs_dentry_operations;
#endif
        dp->d_time = parent_vcache_dv(dip, credp);
        d_instantiate(dp, ip);
    }
    afs_DestroyAttr(vattr);

out:
    AFS_GUNLOCK();

    crfree(credp);
    return afs_convert_code(code);
}

static int
afs_linux_rmdir(struct inode *dip, struct dentry *dp)
{
    int code;
    cred_t *credp = crref();
    const char *name = dp->d_name.name;

    /* locking kernel conflicts with glock? */

    AFS_GLOCK();
    code = afs_rmdir(VTOAFS(dip), (char *)name, credp);
    AFS_GUNLOCK();

    /* Linux likes to see ENOTEMPTY returned from an rmdir() syscall
     * that failed because a directory is not empty. So, we map
     * EEXIST to ENOTEMPTY on linux.
     */
    if (code == EEXIST) {
        code = ENOTEMPTY;
    }

    if (!code) {
        d_drop(dp);
    }

    crfree(credp);
    return afs_convert_code(code);
}


static int
afs_linux_rename(struct inode *oldip, struct dentry *olddp,
                 struct inode *newip, struct dentry *newdp
#ifdef HAVE_LINUX_INODE_OPERATIONS_RENAME_TAKES_FLAGS
                 , unsigned int flags
#endif
                )
{
    int code;
    cred_t *credp = crref();
    const char *oldname = olddp->d_name.name;
    const char *newname = newdp->d_name.name;
    struct dentry *rehash = NULL;

#ifdef HAVE_LINUX_INODE_OPERATIONS_RENAME_TAKES_FLAGS
    if (flags)
        return -EINVAL;         /* no support for new flags yet */
#endif

    /* Prevent any new references during rename operation. */

    if (!d_unhashed(newdp)) {
        d_drop(newdp);
        rehash = newdp;
    }

    afs_maybe_shrink_dcache(olddp);

    AFS_GLOCK();
    code = afs_rename(VTOAFS(oldip), (char *)oldname, VTOAFS(newip), (char *)newname, credp);
    AFS_GUNLOCK();

    if (!code)
        olddp->d_time = 0;      /* force to revalidate */

    if (rehash)
        d_rehash(rehash);

    crfree(credp);
    return afs_convert_code(code);
}


/* afs_linux_ireadlink
 * Internal readlink which can return link contents to user or kernel space.
 * Note that the buffer is NOT supposed to be null-terminated.
 */
static int
afs_linux_ireadlink(struct inode *ip, char *target, int maxlen, uio_seg_t seg)
{
    int code;
    cred_t *credp = crref();
    struct uio tuio;
    struct iovec iov;

    memset(&tuio, 0, sizeof(tuio));
    memset(&iov, 0, sizeof(iov));

    setup_uio(&tuio, &iov, target, (afs_offs_t) 0, maxlen, UIO_READ, seg);
    code = afs_readlink(VTOAFS(ip), &tuio, credp);
    crfree(credp);

    if (!code)
        return maxlen - tuio.uio_resid;
    else
        return afs_convert_code(code);
}

#if !defined(USABLE_KERNEL_PAGE_SYMLINK_CACHE)
/* afs_linux_readlink
 * Fill target (which is in user space) with contents of symlink.
 */
static int
afs_linux_readlink(struct dentry *dp, char *target, int maxlen)
{
    int code;
    struct inode *ip = dp->d_inode;

    AFS_GLOCK();
    code = afs_linux_ireadlink(ip, target, maxlen, AFS_UIOUSER);
    AFS_GUNLOCK();
    return code;
}


/* afs_linux_follow_link
 * a file system dependent link following routine.
 */
#if defined(HAVE_LINUX_INODE_OPERATIONS_FOLLOW_LINK_NO_NAMEIDATA)
static const char *afs_linux_follow_link(struct dentry *dentry, void **link_data)
#else
static int afs_linux_follow_link(struct dentry *dentry, struct nameidata *nd)
#endif
{
    int code;
    char *name;

    name = kmalloc(PATH_MAX, GFP_NOFS);
    if (!name) {
#if defined(HAVE_LINUX_INODE_OPERATIONS_FOLLOW_LINK_NO_NAMEIDATA)
        return ERR_PTR(-EIO);
#else
        return -EIO;
#endif
    }

    AFS_GLOCK();
    code = afs_linux_ireadlink(dentry->d_inode, name, PATH_MAX - 1, AFS_UIOSYS);
    AFS_GUNLOCK();

    if (code < 0) {
#if defined(HAVE_LINUX_INODE_OPERATIONS_FOLLOW_LINK_NO_NAMEIDATA)
        return ERR_PTR(code);
#else
        return code;
#endif
    }

    name[code] = '\0';
#if defined(HAVE_LINUX_INODE_OPERATIONS_FOLLOW_LINK_NO_NAMEIDATA)
    return *link_data = name;
#else
    nd_set_link(nd, name);
    return 0;
#endif
}

#if defined(HAVE_LINUX_INODE_OPERATIONS_PUT_LINK_NO_NAMEIDATA)
static void
afs_linux_put_link(struct inode *inode, void *link_data)
{
    char *name = link_data;

    if (name && !IS_ERR(name))
        kfree(name);
}
#else
static void
afs_linux_put_link(struct dentry *dentry, struct nameidata *nd)
{
    char *name = nd_get_link(nd);

    if (name && !IS_ERR(name))
        kfree(name);
}
#endif /* HAVE_LINUX_INODE_OPERATIONS_PUT_LINK_NO_NAMEIDATA */

#endif /* USABLE_KERNEL_PAGE_SYMLINK_CACHE */

/* Populate a page by filling it from the cache file pointed at by cachefp
 * (which contains indicated chunk)
 * If task is NULL, the page copy occurs syncronously, and the routine
 * returns with page still locked. If task is non-NULL, then page copies
 * may occur in the background, and the page will be unlocked when it is
 * ready for use.
 */
static int
afs_linux_read_cache(struct file *cachefp, struct page *page,
                     int chunk, struct pagevec *lrupv,
                     struct afs_pagecopy_task *task) {
    loff_t offset = page_offset(page);
    struct inode *cacheinode = cachefp->f_dentry->d_inode;
    struct page *newpage, *cachepage;
    struct address_space *cachemapping;
    int pageindex;
    int code = 0;

    cachemapping = cacheinode->i_mapping;
    newpage = NULL;
    cachepage = NULL;

    /* If we're trying to read a page that's past the end of the disk
     * cache file, then just return a zeroed page */
    if (AFS_CHUNKOFFSET(offset) >= i_size_read(cacheinode)) {
        zero_user_segment(page, 0, PAGE_SIZE);
        SetPageUptodate(page);
        if (task)
            unlock_page(page);
        return 0;
    }

    /* From our offset, we now need to work out which page in the disk
     * file it corresponds to. This will be fun ... */
    pageindex = (offset - AFS_CHUNKTOBASE(chunk)) >> PAGE_SHIFT;

    while (cachepage == NULL) {
        cachepage = find_get_page(cachemapping, pageindex);
        if (!cachepage) {
            if (!newpage)
                newpage = page_cache_alloc_cold(cachemapping);
            if (!newpage) {
                code = -ENOMEM;
                goto out;
            }

            code = add_to_page_cache(newpage, cachemapping,
                                     pageindex, GFP_KERNEL);
            if (code == 0) {
                cachepage = newpage;
                newpage = NULL;

                get_page(cachepage);
                if (!pagevec_add(lrupv, cachepage))
                    __pagevec_lru_add_file(lrupv);

            } else {
                put_page(newpage);
                newpage = NULL;
                if (code != -EEXIST)
                    goto out;
            }
        } else {
            lock_page(cachepage);
        }
    }

    if (!PageUptodate(cachepage)) {
        ClearPageError(cachepage);
        code = cachemapping->a_ops->readpage(NULL, cachepage);
        if (!code && !task) {
            wait_on_page_locked(cachepage);
        }
    } else {
        unlock_page(cachepage);
    }

    if (!code) {
        if (PageUptodate(cachepage)) {
            copy_highpage(page, cachepage);
            flush_dcache_page(page);
            SetPageUptodate(page);

            if (task)
                unlock_page(page);
        } else if (task) {
            afs_pagecopy_queue_page(task, cachepage, page);
        } else {
            code = -EIO;
        }
    }

    if (code && task) {
        unlock_page(page);
    }

out:
    if (cachepage)
        put_page(cachepage);

    return code;
}

static int inline
afs_linux_readpage_fastpath(struct file *fp, struct page *pp, int *codep)
{
    loff_t offset = page_offset(pp);
    struct inode *ip = FILE_INODE(fp);
    struct vcache *avc = VTOAFS(ip);
    struct dcache *tdc;
    struct file *cacheFp = NULL;
    int code;
    int dcLocked = 0;
    struct pagevec lrupv;

    /* Not a UFS cache, don't do anything */
    if (cacheDiskType != AFS_FCACHE_TYPE_UFS)
        return 0;

    /* No readpage (ex: tmpfs) , skip */
    if (cachefs_noreadpage)
        return 0;

    /* Can't do anything if the vcache isn't statd , or if the read
     * crosses a chunk boundary.
     */
    if (!(avc->f.states & CStatd) ||
        AFS_CHUNK(offset) != AFS_CHUNK(offset + PAGE_SIZE)) {
        return 0;
    }

    ObtainWriteLock(&avc->lock, 911);

    /* XXX - See if hinting actually makes things faster !!! */

    /* See if we have a suitable entry already cached */
    tdc = avc->dchint;

    if (tdc) {
        /* We need to lock xdcache, then dcache, to handle situations where
         * the hint is on the free list. However, we can't safely do this
         * according to the locking hierarchy. So, use a non blocking lock.
         */
        ObtainReadLock(&afs_xdcache);
        dcLocked = ( 0 == NBObtainReadLock(&tdc->lock));

        if (dcLocked && (tdc->index != NULLIDX)
            && !FidCmp(&tdc->f.fid, &avc->f.fid)
            && tdc->f.chunk == AFS_CHUNK(offset)
            && !(afs_indexFlags[tdc->index] & (IFFree | IFDiscarded))) {
            /* Bonus - the hint was correct */
            afs_RefDCache(tdc);
        } else {
            /* Only destroy the hint if its actually invalid, not if there's
             * just been a locking failure */
            if (dcLocked) {
                ReleaseReadLock(&tdc->lock);
                avc->dchint = NULL;
            }

            tdc = NULL;
            dcLocked = 0;
        }
        ReleaseReadLock(&afs_xdcache);
    }

    /* No hint, or hint is no longer valid - see if we can get something
     * directly from the dcache
     */
    if (!tdc)
        tdc = afs_FindDCache(avc, offset);

    if (!tdc) {
        ReleaseWriteLock(&avc->lock);
        return 0;
    }

    if (!dcLocked)
        ObtainReadLock(&tdc->lock);

    /* Is the dcache we've been given currently up to date */
    if (!hsame(avc->f.m.DataVersion, tdc->f.versionNo) ||
        (tdc->dflags & DFFetching))
        goto out;

    /* Update our hint for future abuse */
    avc->dchint = tdc;

    /* Okay, so we've now got a cache file that is up to date */

    /* XXX - I suspect we should be locking the inodes before we use them! */
    AFS_GUNLOCK();
    cacheFp = afs_linux_raw_open(&tdc->f.inode);
    if (!cacheFp->f_dentry->d_inode->i_mapping->a_ops->readpage) {
        cachefs_noreadpage = 1;
        AFS_GLOCK();
        goto out;
    }
    pagevec_init(&lrupv, 0);

    code = afs_linux_read_cache(cacheFp, pp, tdc->f.chunk, &lrupv, NULL);

    if (pagevec_count(&lrupv))
       __pagevec_lru_add_file(&lrupv);

    filp_close(cacheFp, NULL);
    AFS_GLOCK();

    ReleaseReadLock(&tdc->lock);
    ReleaseWriteLock(&avc->lock);
    afs_PutDCache(tdc);

    *codep = code;
    return 1;

out:
    ReleaseWriteLock(&avc->lock);
    ReleaseReadLock(&tdc->lock);
    afs_PutDCache(tdc);
    return 0;
}

/* afs_linux_readpage
 *
 * This function is split into two, because prepare_write/begin_write
 * require a readpage call which doesn't unlock the resulting page upon
 * success.
 */
static int
afs_linux_fillpage(struct file *fp, struct page *pp)
{
    afs_int32 code;
    char *address;
    struct uio *auio;
    struct iovec *iovecp;
    struct inode *ip = FILE_INODE(fp);
    afs_int32 cnt = page_count(pp);
    struct vcache *avc = VTOAFS(ip);
    afs_offs_t offset = page_offset(pp);
    cred_t *credp;

    AFS_GLOCK();
    if (afs_linux_readpage_fastpath(fp, pp, &code)) {
        AFS_GUNLOCK();
        return code;
    }
    AFS_GUNLOCK();

    credp = crref();
    address = kmap(pp);
    ClearPageError(pp);

    auio = kmalloc(sizeof(struct uio), GFP_NOFS);
    iovecp = kmalloc(sizeof(struct iovec), GFP_NOFS);

    setup_uio(auio, iovecp, (char *)address, offset, PAGE_SIZE, UIO_READ,
              AFS_UIOSYS);

    AFS_GLOCK();
    AFS_DISCON_LOCK();
    afs_Trace4(afs_iclSetp, CM_TRACE_READPAGE, ICL_TYPE_POINTER, ip,
               ICL_TYPE_POINTER, pp, ICL_TYPE_INT32, cnt, ICL_TYPE_INT32,
               99999);  /* not a possible code value */

    code = afs_rdwr(avc, auio, UIO_READ, 0, credp);

    afs_Trace4(afs_iclSetp, CM_TRACE_READPAGE, ICL_TYPE_POINTER, ip,
               ICL_TYPE_POINTER, pp, ICL_TYPE_INT32, cnt, ICL_TYPE_INT32,
               code);
    AFS_DISCON_UNLOCK();
    AFS_GUNLOCK();
    if (!code) {
        /* XXX valid for no-cache also?  Check last bits of files... :)
         * Cognate code goes in afs_NoCacheFetchProc.  */
        if (auio->uio_resid)    /* zero remainder of page */
             memset((void *)(address + (PAGE_SIZE - auio->uio_resid)), 0,
                    auio->uio_resid);

        flush_dcache_page(pp);
        SetPageUptodate(pp);
    } /* !code */

    kunmap(pp);

    kfree(auio);
    kfree(iovecp);

    crfree(credp);
    return afs_convert_code(code);
}

static int
afs_linux_prefetch(struct file *fp, struct page *pp)
{
    int code = 0;
    struct vcache *avc = VTOAFS(FILE_INODE(fp));
    afs_offs_t offset = page_offset(pp);

    if (AFS_CHUNKOFFSET(offset) == 0) {
        struct dcache *tdc;
        struct vrequest *treq = NULL;
        cred_t *credp;

        credp = crref();
        AFS_GLOCK();
        code = afs_CreateReq(&treq, credp);
        if (!code && !NBObtainWriteLock(&avc->lock, 534)) {
            tdc = afs_FindDCache(avc, offset);
            if (tdc) {
                if (!(tdc->mflags & DFNextStarted))
                    afs_PrefetchChunk(avc, tdc, credp, treq);
                afs_PutDCache(tdc);
            }
            ReleaseWriteLock(&avc->lock);
        }
        afs_DestroyReq(treq);
        AFS_GUNLOCK();
        crfree(credp);
    }
    return afs_convert_code(code);

}

static int
afs_linux_bypass_readpages(struct file *fp, struct address_space *mapping,
                           struct list_head *page_list, unsigned num_pages)
{
    afs_int32 page_ix;
    struct uio *auio;
    afs_offs_t offset;
    struct iovec* iovecp;
    struct nocache_read_request *ancr;
    struct page *pp;
    struct pagevec lrupv;
    afs_int32 code = 0;

    cred_t *credp;
    struct inode *ip = FILE_INODE(fp);
    struct vcache *avc = VTOAFS(ip);
    afs_int32 base_index = 0;
    afs_int32 page_count = 0;
    afs_int32 isize;

    /* background thread must free: iovecp, auio, ancr */
    iovecp = osi_Alloc(num_pages * sizeof(struct iovec));

    auio = osi_Alloc(sizeof(struct uio));
    auio->uio_iov = iovecp;
    auio->uio_iovcnt = num_pages;
    auio->uio_flag = UIO_READ;
    auio->uio_seg = AFS_UIOSYS;
    auio->uio_resid = num_pages * PAGE_SIZE;

    ancr = osi_Alloc(sizeof(struct nocache_read_request));
    ancr->auio = auio;
    ancr->offset = auio->uio_offset;
    ancr->length = auio->uio_resid;

    pagevec_init(&lrupv, 0);

    for(page_ix = 0; page_ix < num_pages; ++page_ix) {

        if(list_empty(page_list))
            break;

        pp = list_entry(page_list->prev, struct page, lru);
        /* If we allocate a page and don't remove it from page_list,
         * the page cache gets upset. */
        list_del(&pp->lru);
        isize = (i_size_read(fp->f_mapping->host) - 1) >> PAGE_SHIFT;
        if(pp->index > isize) {
            if(PageLocked(pp))
                unlock_page(pp);
            continue;
        }

        if(page_ix == 0) {
            offset = page_offset(pp);
            ancr->offset = auio->uio_offset = offset;
            base_index = pp->index;
        }
        iovecp[page_ix].iov_len = PAGE_SIZE;
        code = add_to_page_cache(pp, mapping, pp->index, GFP_KERNEL);
        if(base_index != pp->index) {
            if(PageLocked(pp))
                 unlock_page(pp);
            put_page(pp);
            iovecp[page_ix].iov_base = (void *) 0;
            base_index++;
            ancr->length -= PAGE_SIZE;
            continue;
        }
        base_index++;
        if(code) {
            if(PageLocked(pp))
                unlock_page(pp);
            put_page(pp);
            iovecp[page_ix].iov_base = (void *) 0;
        } else {
            page_count++;
            if(!PageLocked(pp)) {
                lock_page(pp);
            }

            /* increment page refcount--our original design assumed
             * that locking it would effectively pin it;  protect
             * ourselves from the possiblity that this assumption is
             * is faulty, at low cost (provided we do not fail to
             * do the corresponding decref on the other side) */
            get_page(pp);

            /* save the page for background map */
            iovecp[page_ix].iov_base = (void*) pp;

            /* and put it on the LRU cache */
            if (!pagevec_add(&lrupv, pp))
                __pagevec_lru_add_file(&lrupv);
        }
    }

    /* If there were useful pages in the page list, make sure all pages
     * are in the LRU cache, then schedule the read */
    if(page_count) {
        if (pagevec_count(&lrupv))
            __pagevec_lru_add_file(&lrupv);
        credp = crref();
        code = afs_ReadNoCache(avc, ancr, credp);
        crfree(credp);
    } else {
        /* If there is nothing for the background thread to handle,
         * it won't be freeing the things that we never gave it */
        osi_Free(iovecp, num_pages * sizeof(struct iovec));
        osi_Free(auio, sizeof(struct uio));
        osi_Free(ancr, sizeof(struct nocache_read_request));
    }
    /* we do not flush, release, or unmap pages--that will be
     * done for us by the background thread as each page comes in
     * from the fileserver */
    return afs_convert_code(code);
}


static int
afs_linux_bypass_readpage(struct file *fp, struct page *pp)
{
    cred_t *credp = NULL;
    struct uio *auio;
    struct iovec *iovecp;
    struct nocache_read_request *ancr;
    int code;

    /*
     * Special case: if page is at or past end of file, just zero it and set
     * it as up to date.
     */
    if (page_offset(pp) >=  i_size_read(fp->f_mapping->host)) {
        zero_user_segment(pp, 0, PAGE_SIZE);
        SetPageUptodate(pp);
        unlock_page(pp);
        return 0;
    }

    ClearPageError(pp);

    /* receiver frees */
    auio = osi_Alloc(sizeof(struct uio));
    iovecp = osi_Alloc(sizeof(struct iovec));

    /* address can be NULL, because we overwrite it with 'pp', below */
    setup_uio(auio, iovecp, NULL, page_offset(pp),
              PAGE_SIZE, UIO_READ, AFS_UIOSYS);

    /* save the page for background map */
    get_page(pp); /* see above */
    auio->uio_iov->iov_base = (void*) pp;
    /* the background thread will free this */
    ancr = osi_Alloc(sizeof(struct nocache_read_request));
    ancr->auio = auio;
    ancr->offset = page_offset(pp);
    ancr->length = PAGE_SIZE;

    credp = crref();
    code = afs_ReadNoCache(VTOAFS(FILE_INODE(fp)), ancr, credp);
    crfree(credp);

    return afs_convert_code(code);
}

static inline int
afs_linux_can_bypass(struct inode *ip) {

    switch(cache_bypass_strategy) {
        case NEVER_BYPASS_CACHE:
            return 0;
        case ALWAYS_BYPASS_CACHE:
            return 1;
        case LARGE_FILES_BYPASS_CACHE:
            if (i_size_read(ip) > cache_bypass_threshold)
                return 1;
        default:
            return 0;
     }
}

/* Check if a file is permitted to bypass the cache by policy, and modify
 * the cache bypass state recorded for that file */

static inline int
afs_linux_bypass_check(struct inode *ip) {
    cred_t* credp;

    int bypass = afs_linux_can_bypass(ip);

    credp = crref();
    trydo_cache_transition(VTOAFS(ip), credp, bypass);
    crfree(credp);

    return bypass;
}


static int
afs_linux_readpage(struct file *fp, struct page *pp)
{
    int code;

    if (afs_linux_bypass_check(FILE_INODE(fp))) {
        code = afs_linux_bypass_readpage(fp, pp);
    } else {
        code = afs_linux_fillpage(fp, pp);
        if (!code)
            code = afs_linux_prefetch(fp, pp);
        unlock_page(pp);
    }

    return code;
}

/* Readpages reads a number of pages for a particular file. We use
 * this to optimise the reading, by limiting the number of times upon which
 * we have to lookup, lock and open vcaches and dcaches
 */

static int
afs_linux_readpages(struct file *fp, struct address_space *mapping,
                    struct list_head *page_list, unsigned int num_pages)
{
    struct inode *inode = mapping->host;
    struct vcache *avc = VTOAFS(inode);
    struct dcache *tdc;
    struct file *cacheFp = NULL;
    int code;
    unsigned int page_idx;
    loff_t offset;
    struct pagevec lrupv;
    struct afs_pagecopy_task *task;

    if (afs_linux_bypass_check(inode))
        return afs_linux_bypass_readpages(fp, mapping, page_list, num_pages);

    if (cacheDiskType == AFS_FCACHE_TYPE_MEM)
        return 0;

    /* No readpage (ex: tmpfs) , skip */
    if (cachefs_noreadpage)
        return 0;

    AFS_GLOCK();
    if ((code = afs_linux_VerifyVCache(avc, NULL))) {
        AFS_GUNLOCK();
        return code;
    }

    ObtainWriteLock(&avc->lock, 912);
    AFS_GUNLOCK();

    task = afs_pagecopy_init_task();

    tdc = NULL;
    pagevec_init(&lrupv, 0);
    for (page_idx = 0; page_idx < num_pages; page_idx++) {
        struct page *page = list_entry(page_list->prev, struct page, lru);
        list_del(&page->lru);
        offset = page_offset(page);

        if (tdc && tdc->f.chunk != AFS_CHUNK(offset)) {
            AFS_GLOCK();
            ReleaseReadLock(&tdc->lock);
            afs_PutDCache(tdc);
            AFS_GUNLOCK();
            tdc = NULL;
            if (cacheFp)
                filp_close(cacheFp, NULL);
        }

        if (!tdc) {
            AFS_GLOCK();
            if ((tdc = afs_FindDCache(avc, offset))) {
                ObtainReadLock(&tdc->lock);
                if (!hsame(avc->f.m.DataVersion, tdc->f.versionNo) ||
                    (tdc->dflags & DFFetching)) {
                    ReleaseReadLock(&tdc->lock);
                    afs_PutDCache(tdc);
                    tdc = NULL;
                }
            }
            AFS_GUNLOCK();
            if (tdc) {
                cacheFp = afs_linux_raw_open(&tdc->f.inode);
                if (!cacheFp->f_dentry->d_inode->i_mapping->a_ops->readpage) {
                    cachefs_noreadpage = 1;
                    goto out;
                }
            }
        }

        if (tdc && !add_to_page_cache(page, mapping, page->index,
                                      GFP_KERNEL)) {
            get_page(page);
            if (!pagevec_add(&lrupv, page))
                __pagevec_lru_add_file(&lrupv);

            afs_linux_read_cache(cacheFp, page, tdc->f.chunk, &lrupv, task);
        }
        put_page(page);
    }
    if (pagevec_count(&lrupv))
       __pagevec_lru_add_file(&lrupv);

out:
    if (tdc)
        filp_close(cacheFp, NULL);

    afs_pagecopy_put_task(task);

    AFS_GLOCK();
    if (tdc) {
        ReleaseReadLock(&tdc->lock);
        afs_PutDCache(tdc);
    }

    ReleaseWriteLock(&avc->lock);
    AFS_GUNLOCK();
    return 0;
}

/* Prepare an AFS vcache for writeback. Should be called with the vcache
 * locked */
static inline int
afs_linux_prepare_writeback(struct vcache *avc) {
    pid_t pid;
    struct pagewriter *pw;

    pid = MyPidxx2Pid(MyPidxx);
    /* Prevent recursion into the writeback code */
    spin_lock(&avc->pagewriter_lock);
    list_for_each_entry(pw, &avc->pagewriters, link) {
        if (pw->writer == pid) {
            spin_unlock(&avc->pagewriter_lock);
            return AOP_WRITEPAGE_ACTIVATE;
        }
    }
    spin_unlock(&avc->pagewriter_lock);

    /* Add ourselves to writer list */
    pw = osi_Alloc(sizeof(struct pagewriter));
    pw->writer = pid;
    spin_lock(&avc->pagewriter_lock);
    list_add_tail(&pw->link, &avc->pagewriters);
    spin_unlock(&avc->pagewriter_lock);

    return 0;
}

static inline int
afs_linux_dopartialwrite(struct vcache *avc, cred_t *credp) {
    struct vrequest *treq = NULL;
    int code = 0;

    if (!afs_CreateReq(&treq, credp)) {
        code = afs_DoPartialWrite(avc, treq);
        afs_DestroyReq(treq);
    }

    return afs_convert_code(code);
}

static inline void
afs_linux_complete_writeback(struct vcache *avc) {
    struct pagewriter *pw, *store;
    pid_t pid;
    struct list_head tofree;

    INIT_LIST_HEAD(&tofree);
    pid = MyPidxx2Pid(MyPidxx);
    /* Remove ourselves from writer list */
    spin_lock(&avc->pagewriter_lock);
    list_for_each_entry_safe(pw, store, &avc->pagewriters, link) {
        if (pw->writer == pid) {
            list_del(&pw->link);
            /* osi_Free may sleep so we need to defer it */
            list_add_tail(&pw->link, &tofree);
        }
    }
    spin_unlock(&avc->pagewriter_lock);
    list_for_each_entry_safe(pw, store, &tofree, link) {
        list_del(&pw->link);
        osi_Free(pw, sizeof(struct pagewriter));
    }
}

/* Writeback a given page syncronously. Called with no AFS locks held */
static int
afs_linux_page_writeback(struct inode *ip, struct page *pp,
                         unsigned long offset, unsigned int count,
                         cred_t *credp)
{
    struct vcache *vcp = VTOAFS(ip);
    char *buffer;
    afs_offs_t base;
    int code = 0;
    struct uio tuio;
    struct iovec iovec;
    int f_flags = 0;

    memset(&tuio, 0, sizeof(tuio));
    memset(&iovec, 0, sizeof(iovec));

    buffer = kmap(pp) + offset;
    base = page_offset(pp) + offset;

    AFS_GLOCK();
    afs_Trace4(afs_iclSetp, CM_TRACE_UPDATEPAGE, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_POINTER, pp, ICL_TYPE_INT32, page_count(pp),
               ICL_TYPE_INT32, 99999);

    setup_uio(&tuio, &iovec, buffer, base, count, UIO_WRITE, AFS_UIOSYS);

    code = afs_write(vcp, &tuio, f_flags, credp, 0);

    i_size_write(ip, vcp->f.m.Length);
    ip->i_blocks = ((vcp->f.m.Length + 1023) >> 10) << 1;

    code = code ? afs_convert_code(code) : count - tuio.uio_resid;

    afs_Trace4(afs_iclSetp, CM_TRACE_UPDATEPAGE, ICL_TYPE_POINTER, vcp,
               ICL_TYPE_POINTER, pp, ICL_TYPE_INT32, page_count(pp),
               ICL_TYPE_INT32, code);

    AFS_GUNLOCK();
    kunmap(pp);

    return code;
}

static int
afs_linux_writepage_sync(struct inode *ip, struct page *pp,
                         unsigned long offset, unsigned int count)
{
    int code;
    int code1 = 0;
    struct vcache *vcp = VTOAFS(ip);
    cred_t *credp;

    /* Catch recursive writeback. This occurs if the kernel decides
     * writeback is required whilst we are writing to the cache, or
     * flushing to the server. When we're running syncronously (as
     * opposed to from writepage) we can't actually do anything about
     * this case - as we can't return AOP_WRITEPAGE_ACTIVATE to write()
     */
    AFS_GLOCK();
    ObtainWriteLock(&vcp->lock, 532);
    afs_linux_prepare_writeback(vcp);
    ReleaseWriteLock(&vcp->lock);
    AFS_GUNLOCK();

    credp = crref();
    code = afs_linux_page_writeback(ip, pp, offset, count, credp);

    AFS_GLOCK();
    ObtainWriteLock(&vcp->lock, 533);
    if (code > 0)
        code1 = afs_linux_dopartialwrite(vcp, credp);
    afs_linux_complete_writeback(vcp);
    ReleaseWriteLock(&vcp->lock);
    AFS_GUNLOCK();
    crfree(credp);

    if (code1)
        return code1;

    return code;
}

static int
#ifdef AOP_WRITEPAGE_TAKES_WRITEBACK_CONTROL
afs_linux_writepage(struct page *pp, struct writeback_control *wbc)
#else
afs_linux_writepage(struct page *pp)
#endif
{
    struct address_space *mapping = pp->mapping;
    struct inode *inode;
    struct vcache *vcp;
    cred_t *credp;
    unsigned int to = PAGE_SIZE;
    loff_t isize;
    int code = 0;
    int code1 = 0;

    get_page(pp);

    inode = mapping->host;
    vcp = VTOAFS(inode);
    isize = i_size_read(inode);

    /* Don't defeat an earlier truncate */
    if (page_offset(pp) > isize) {
        set_page_writeback(pp);
        unlock_page(pp);
        goto done;
    }

    AFS_GLOCK();
    ObtainWriteLock(&vcp->lock, 537);
    code = afs_linux_prepare_writeback(vcp);
    if (code == AOP_WRITEPAGE_ACTIVATE) {
        /* WRITEPAGE_ACTIVATE is the only return value that permits us
         * to return with the page still locked */
        ReleaseWriteLock(&vcp->lock);
        AFS_GUNLOCK();
        return code;
    }

    /* Grab the creds structure currently held in the vnode, and
     * get a reference to it, in case it goes away ... */
    credp = vcp->cred;
    if (credp)
        crhold(credp);
    else
        credp = crref();
    ReleaseWriteLock(&vcp->lock);
    AFS_GUNLOCK();

    set_page_writeback(pp);

    SetPageUptodate(pp);

    /* We can unlock the page here, because it's protected by the
     * page_writeback flag. This should make us less vulnerable to
     * deadlocking in afs_write and afs_DoPartialWrite
     */
    unlock_page(pp);

    /* If this is the final page, then just write the number of bytes that
     * are actually in it */
    if ((isize - page_offset(pp)) < to )
        to = isize - page_offset(pp);

    code = afs_linux_page_writeback(inode, pp, 0, to, credp);

    AFS_GLOCK();
    ObtainWriteLock(&vcp->lock, 538);

    /* As much as we might like to ignore a file server error here,
     * and just try again when we close(), unfortunately StoreAllSegments
     * will invalidate our chunks if the server returns a permanent error,
     * so we need to at least try and get that error back to the user
     */
    if (code == to)
        code1 = afs_linux_dopartialwrite(vcp, credp);

    afs_linux_complete_writeback(vcp);
    ReleaseWriteLock(&vcp->lock);
    crfree(credp);
    AFS_GUNLOCK();

done:
    end_page_writeback(pp);
    put_page(pp);

    if (code1)
        return code1;

    if (code == to)
        return 0;

    return code;
}

/* afs_linux_permission
 * Check access rights - returns error if can't check or permission denied.
 */
static int
#if defined(IOP_PERMISSION_TAKES_FLAGS)
afs_linux_permission(struct inode *ip, int mode, unsigned int flags)
#elif defined(IOP_PERMISSION_TAKES_NAMEIDATA)
afs_linux_permission(struct inode *ip, int mode, struct nameidata *nd)
#else
afs_linux_permission(struct inode *ip, int mode)
#endif
{
    int code;
    cred_t *credp;
    int tmp = 0;

    /* Check for RCU path walking */
#if defined(IOP_PERMISSION_TAKES_FLAGS)
    if (flags & IPERM_FLAG_RCU)
       return -ECHILD;
#elif defined(MAY_NOT_BLOCK)
    if (mode & MAY_NOT_BLOCK)
       return -ECHILD;
#endif

    credp = crref();
    AFS_GLOCK();
    if (mode & MAY_EXEC)
        tmp |= VEXEC;
    if (mode & MAY_READ)
        tmp |= VREAD;
    if (mode & MAY_WRITE)
        tmp |= VWRITE;
    code = afs_access(VTOAFS(ip), tmp, credp);

    AFS_GUNLOCK();
    crfree(credp);
    return afs_convert_code(code);
}

static int
afs_linux_commit_write(struct file *file, struct page *page, unsigned offset,
                       unsigned to)
{
    int code;
    struct inode *inode = FILE_INODE(file);
    loff_t pagebase = page_offset(page);

    if (i_size_read(inode) < (pagebase + offset))
        i_size_write(inode, pagebase + offset);

    if (PageChecked(page)) {
        SetPageUptodate(page);
        ClearPageChecked(page);
    }

    code = afs_linux_writepage_sync(inode, page, offset, to - offset);

    return code;
}

static int
afs_linux_prepare_write(struct file *file, struct page *page, unsigned from,
                        unsigned to)
{

    /* http://kerneltrap.org/node/4941 details the expected behaviour of
     * prepare_write. Essentially, if the page exists within the file,
     * and is not being fully written, then we should populate it.
     */

    if (!PageUptodate(page)) {
        loff_t pagebase = page_offset(page);
        loff_t isize = i_size_read(page->mapping->host);

        /* Is the location we are writing to beyond the end of the file? */
        if (pagebase >= isize ||
            ((from == 0) && (pagebase + to) >= isize)) {
            zero_user_segments(page, 0, from, to, PAGE_SIZE);
            SetPageChecked(page);
        /* Are we we writing a full page */
        } else if (from == 0 && to == PAGE_SIZE) {
            SetPageChecked(page);
        /* Is the page readable, if it's wronly, we don't care, because we're
         * not actually going to read from it ... */
        } else if ((file->f_flags && O_ACCMODE) != O_WRONLY) {
            /* We don't care if fillpage fails, because if it does the page
             * won't be marked as up to date
             */
            afs_linux_fillpage(file, page);
        }
    }
    return 0;
}

#if defined(STRUCT_ADDRESS_SPACE_OPERATIONS_HAS_WRITE_BEGIN)
static int
afs_linux_write_end(struct file *file, struct address_space *mapping,
                                loff_t pos, unsigned len, unsigned copied,
                                struct page *page, void *fsdata)
{
    int code;
    unsigned int from = pos & (PAGE_SIZE - 1);

    code = afs_linux_commit_write(file, page, from, from + copied);

    unlock_page(page);
    put_page(page);
    return code;
}

static int
afs_linux_write_begin(struct file *file, struct address_space *mapping,
                                loff_t pos, unsigned len, unsigned flags,
                                struct page **pagep, void **fsdata)
{
    struct page *page;
    pgoff_t index = pos >> PAGE_SHIFT;
    unsigned int from = pos & (PAGE_SIZE - 1);
    int code;

    page = grab_cache_page_write_begin(mapping, index, flags);
    *pagep = page;

    code = afs_linux_prepare_write(file, page, from, from + len);
    if (code) {
        unlock_page(page);
        put_page(page);
    }

    return code;
}
#endif

#ifndef STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT
static void *
afs_linux_dir_follow_link(struct dentry *dentry, struct nameidata *nd)
{
    struct dentry **dpp;
    struct dentry *target;

    if (current->total_link_count > 0) {
        /* avoid symlink resolution limits when resolving; we cannot contribute to
         * an infinite symlink loop */
        /* only do this for follow_link when total_link_count is positive to be
         * on the safe side; there is at least one code path in the Linux
         * kernel where it seems like it may be possible to get here without
         * total_link_count getting incremented. it is not clear on how that
         * path is actually reached, but guard against it just to be safe */
        current->total_link_count--;
    }

    target = canonical_dentry(dentry->d_inode);

# ifdef STRUCT_NAMEIDATA_HAS_PATH
    dpp = &nd->path.dentry;
# else
    dpp = &nd->dentry;
# endif

    dput(*dpp);

    if (target) {
        *dpp = target;
    } else {
        *dpp = dget(dentry);
    }

    nd->last_type = LAST_BIND;

    return NULL;
}
#endif /* !STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT */


static struct inode_operations afs_file_iops = {
  .permission =         afs_linux_permission,
  .getattr =            afs_linux_getattr,
  .setattr =            afs_notify_change,
};

static struct address_space_operations afs_file_aops = {
  .readpage =           afs_linux_readpage,
  .readpages =          afs_linux_readpages,
  .writepage =          afs_linux_writepage,
#if defined (STRUCT_ADDRESS_SPACE_OPERATIONS_HAS_WRITE_BEGIN)
  .write_begin =        afs_linux_write_begin,
  .write_end =          afs_linux_write_end,
#else
  .commit_write =       afs_linux_commit_write,
  .prepare_write =      afs_linux_prepare_write,
#endif
};


/* Separate ops vector for directories. Linux 2.2 tests type of inode
 * by what sort of operation is allowed.....
 */

static struct inode_operations afs_dir_iops = {
  .setattr =            afs_notify_change,
  .create =             afs_linux_create,
  .lookup =             afs_linux_lookup,
  .link =               afs_linux_link,
  .unlink =             afs_linux_unlink,
  .symlink =            afs_linux_symlink,
  .mkdir =              afs_linux_mkdir,
  .rmdir =              afs_linux_rmdir,
  .rename =             afs_linux_rename,
  .getattr =            afs_linux_getattr,
  .permission =         afs_linux_permission,
#ifndef STRUCT_DENTRY_OPERATIONS_HAS_D_AUTOMOUNT
  .follow_link =        afs_linux_dir_follow_link,
#endif
};

/* We really need a separate symlink set of ops, since do_follow_link()
 * determines if it _is_ a link by checking if the follow_link op is set.
 */
#if defined(USABLE_KERNEL_PAGE_SYMLINK_CACHE)
static int
afs_symlink_filler(struct file *file, struct page *page)
{
    struct inode *ip = (struct inode *)page->mapping->host;
    char *p = (char *)kmap(page);
    int code;

    AFS_GLOCK();
    code = afs_linux_ireadlink(ip, p, PAGE_SIZE, AFS_UIOSYS);
    AFS_GUNLOCK();

    if (code < 0)
        goto fail;
    p[code] = '\0';             /* null terminate? */

    SetPageUptodate(page);
    kunmap(page);
    unlock_page(page);
    return 0;

  fail:
    SetPageError(page);
    kunmap(page);
    unlock_page(page);
    return code;
}

static struct address_space_operations afs_symlink_aops = {
  .readpage =   afs_symlink_filler
};
#endif  /* USABLE_KERNEL_PAGE_SYMLINK_CACHE */

static struct inode_operations afs_symlink_iops = {
#if defined(USABLE_KERNEL_PAGE_SYMLINK_CACHE)
  .readlink =           page_readlink,
# if defined(HAVE_LINUX_PAGE_GET_LINK)
  .get_link =           page_get_link,
# elif defined(HAVE_LINUX_PAGE_FOLLOW_LINK)
  .follow_link =        page_follow_link,
# else
  .follow_link =        page_follow_link_light,
  .put_link =           page_put_link,
# endif
#else /* !defined(USABLE_KERNEL_PAGE_SYMLINK_CACHE) */
  .readlink =           afs_linux_readlink,
  .follow_link =        afs_linux_follow_link,
  .put_link =           afs_linux_put_link,
#endif /* USABLE_KERNEL_PAGE_SYMLINK_CACHE */
  .setattr =            afs_notify_change,
};

void
afs_fill_inode(struct inode *ip, struct vattr *vattr)
{
    if (vattr)
        vattr2inode(ip, vattr);

#ifdef STRUCT_ADDRESS_SPACE_HAS_BACKING_DEV_INFO
    ip->i_mapping->backing_dev_info = afs_backing_dev_info;
#endif
/* Reset ops if symlink or directory. */
    if (S_ISREG(ip->i_mode)) {
        ip->i_op = &afs_file_iops;
        ip->i_fop = &afs_file_fops;
        ip->i_data.a_ops = &afs_file_aops;

    } else if (S_ISDIR(ip->i_mode)) {
        ip->i_op = &afs_dir_iops;
        ip->i_fop = &afs_dir_fops;

    } else if (S_ISLNK(ip->i_mode)) {
        ip->i_op = &afs_symlink_iops;
#if defined(HAVE_LINUX_INODE_NOHIGHMEM)
        inode_nohighmem(ip);
#endif
#if defined(USABLE_KERNEL_PAGE_SYMLINK_CACHE)
        ip->i_data.a_ops = &afs_symlink_aops;
        ip->i_mapping = &ip->i_data;
#endif
    }

}