Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / xfs / xfs_ioctl.c
diff --git a/kernel/fs/xfs/xfs_ioctl.c b/kernel/fs/xfs/xfs_ioctl.c
new file mode 100644 (file)
index 0000000..87f67c6
--- /dev/null
@@ -0,0 +1,1806 @@
+/*
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.
+ * All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_inode.h"
+#include "xfs_ioctl.h"
+#include "xfs_alloc.h"
+#include "xfs_rtalloc.h"
+#include "xfs_itable.h"
+#include "xfs_error.h"
+#include "xfs_attr.h"
+#include "xfs_bmap.h"
+#include "xfs_bmap_util.h"
+#include "xfs_fsops.h"
+#include "xfs_discard.h"
+#include "xfs_quota.h"
+#include "xfs_export.h"
+#include "xfs_trace.h"
+#include "xfs_icache.h"
+#include "xfs_symlink.h"
+#include "xfs_trans.h"
+#include "xfs_pnfs.h"
+
+#include <linux/capability.h>
+#include <linux/dcache.h>
+#include <linux/mount.h>
+#include <linux/namei.h>
+#include <linux/pagemap.h>
+#include <linux/slab.h>
+#include <linux/exportfs.h>
+
+/*
+ * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
+ * a file or fs handle.
+ *
+ * XFS_IOC_PATH_TO_FSHANDLE
+ *    returns fs handle for a mount point or path within that mount point
+ * XFS_IOC_FD_TO_HANDLE
+ *    returns full handle for a FD opened in user space
+ * XFS_IOC_PATH_TO_HANDLE
+ *    returns full handle for a path
+ */
+int
+xfs_find_handle(
+       unsigned int            cmd,
+       xfs_fsop_handlereq_t    *hreq)
+{
+       int                     hsize;
+       xfs_handle_t            handle;
+       struct inode            *inode;
+       struct fd               f = {NULL};
+       struct path             path;
+       int                     error;
+       struct xfs_inode        *ip;
+
+       if (cmd == XFS_IOC_FD_TO_HANDLE) {
+               f = fdget(hreq->fd);
+               if (!f.file)
+                       return -EBADF;
+               inode = file_inode(f.file);
+       } else {
+               error = user_lpath((const char __user *)hreq->path, &path);
+               if (error)
+                       return error;
+               inode = d_inode(path.dentry);
+       }
+       ip = XFS_I(inode);
+
+       /*
+        * We can only generate handles for inodes residing on a XFS filesystem,
+        * and only for regular files, directories or symbolic links.
+        */
+       error = -EINVAL;
+       if (inode->i_sb->s_magic != XFS_SB_MAGIC)
+               goto out_put;
+
+       error = -EBADF;
+       if (!S_ISREG(inode->i_mode) &&
+           !S_ISDIR(inode->i_mode) &&
+           !S_ISLNK(inode->i_mode))
+               goto out_put;
+
+
+       memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
+
+       if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
+               /*
+                * This handle only contains an fsid, zero the rest.
+                */
+               memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
+               hsize = sizeof(xfs_fsid_t);
+       } else {
+               handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
+                                       sizeof(handle.ha_fid.fid_len);
+               handle.ha_fid.fid_pad = 0;
+               handle.ha_fid.fid_gen = ip->i_d.di_gen;
+               handle.ha_fid.fid_ino = ip->i_ino;
+
+               hsize = XFS_HSIZE(handle);
+       }
+
+       error = -EFAULT;
+       if (copy_to_user(hreq->ohandle, &handle, hsize) ||
+           copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
+               goto out_put;
+
+       error = 0;
+
+ out_put:
+       if (cmd == XFS_IOC_FD_TO_HANDLE)
+               fdput(f);
+       else
+               path_put(&path);
+       return error;
+}
+
+/*
+ * No need to do permission checks on the various pathname components
+ * as the handle operations are privileged.
+ */
+STATIC int
+xfs_handle_acceptable(
+       void                    *context,
+       struct dentry           *dentry)
+{
+       return 1;
+}
+
+/*
+ * Convert userspace handle data into a dentry.
+ */
+struct dentry *
+xfs_handle_to_dentry(
+       struct file             *parfilp,
+       void __user             *uhandle,
+       u32                     hlen)
+{
+       xfs_handle_t            handle;
+       struct xfs_fid64        fid;
+
+       /*
+        * Only allow handle opens under a directory.
+        */
+       if (!S_ISDIR(file_inode(parfilp)->i_mode))
+               return ERR_PTR(-ENOTDIR);
+
+       if (hlen != sizeof(xfs_handle_t))
+               return ERR_PTR(-EINVAL);
+       if (copy_from_user(&handle, uhandle, hlen))
+               return ERR_PTR(-EFAULT);
+       if (handle.ha_fid.fid_len !=
+           sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
+               return ERR_PTR(-EINVAL);
+
+       memset(&fid, 0, sizeof(struct fid));
+       fid.ino = handle.ha_fid.fid_ino;
+       fid.gen = handle.ha_fid.fid_gen;
+
+       return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
+                       FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
+                       xfs_handle_acceptable, NULL);
+}
+
+STATIC struct dentry *
+xfs_handlereq_to_dentry(
+       struct file             *parfilp,
+       xfs_fsop_handlereq_t    *hreq)
+{
+       return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
+}
+
+int
+xfs_open_by_handle(
+       struct file             *parfilp,
+       xfs_fsop_handlereq_t    *hreq)
+{
+       const struct cred       *cred = current_cred();
+       int                     error;
+       int                     fd;
+       int                     permflag;
+       struct file             *filp;
+       struct inode            *inode;
+       struct dentry           *dentry;
+       fmode_t                 fmode;
+       struct path             path;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       dentry = xfs_handlereq_to_dentry(parfilp, hreq);
+       if (IS_ERR(dentry))
+               return PTR_ERR(dentry);
+       inode = d_inode(dentry);
+
+       /* Restrict xfs_open_by_handle to directories & regular files. */
+       if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
+               error = -EPERM;
+               goto out_dput;
+       }
+
+#if BITS_PER_LONG != 32
+       hreq->oflags |= O_LARGEFILE;
+#endif
+
+       permflag = hreq->oflags;
+       fmode = OPEN_FMODE(permflag);
+       if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
+           (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
+               error = -EPERM;
+               goto out_dput;
+       }
+
+       if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
+               error = -EACCES;
+               goto out_dput;
+       }
+
+       /* Can't write directories. */
+       if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
+               error = -EISDIR;
+               goto out_dput;
+       }
+
+       fd = get_unused_fd_flags(0);
+       if (fd < 0) {
+               error = fd;
+               goto out_dput;
+       }
+
+       path.mnt = parfilp->f_path.mnt;
+       path.dentry = dentry;
+       filp = dentry_open(&path, hreq->oflags, cred);
+       dput(dentry);
+       if (IS_ERR(filp)) {
+               put_unused_fd(fd);
+               return PTR_ERR(filp);
+       }
+
+       if (S_ISREG(inode->i_mode)) {
+               filp->f_flags |= O_NOATIME;
+               filp->f_mode |= FMODE_NOCMTIME;
+       }
+
+       fd_install(fd, filp);
+       return fd;
+
+ out_dput:
+       dput(dentry);
+       return error;
+}
+
+int
+xfs_readlink_by_handle(
+       struct file             *parfilp,
+       xfs_fsop_handlereq_t    *hreq)
+{
+       struct dentry           *dentry;
+       __u32                   olen;
+       void                    *link;
+       int                     error;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       dentry = xfs_handlereq_to_dentry(parfilp, hreq);
+       if (IS_ERR(dentry))
+               return PTR_ERR(dentry);
+
+       /* Restrict this handle operation to symlinks only. */
+       if (!d_is_symlink(dentry)) {
+               error = -EINVAL;
+               goto out_dput;
+       }
+
+       if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
+               error = -EFAULT;
+               goto out_dput;
+       }
+
+       link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
+       if (!link) {
+               error = -ENOMEM;
+               goto out_dput;
+       }
+
+       error = xfs_readlink(XFS_I(d_inode(dentry)), link);
+       if (error)
+               goto out_kfree;
+       error = readlink_copy(hreq->ohandle, olen, link);
+       if (error)
+               goto out_kfree;
+
+ out_kfree:
+       kfree(link);
+ out_dput:
+       dput(dentry);
+       return error;
+}
+
+int
+xfs_set_dmattrs(
+       xfs_inode_t     *ip,
+       u_int           evmask,
+       u_int16_t       state)
+{
+       xfs_mount_t     *mp = ip->i_mount;
+       xfs_trans_t     *tp;
+       int             error;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       if (XFS_FORCED_SHUTDOWN(mp))
+               return -EIO;
+
+       tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
+       error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
+       if (error) {
+               xfs_trans_cancel(tp, 0);
+               return error;
+       }
+       xfs_ilock(ip, XFS_ILOCK_EXCL);
+       xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
+
+       ip->i_d.di_dmevmask = evmask;
+       ip->i_d.di_dmstate  = state;
+
+       xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
+       error = xfs_trans_commit(tp, 0);
+
+       return error;
+}
+
+STATIC int
+xfs_fssetdm_by_handle(
+       struct file             *parfilp,
+       void                    __user *arg)
+{
+       int                     error;
+       struct fsdmidata        fsd;
+       xfs_fsop_setdm_handlereq_t dmhreq;
+       struct dentry           *dentry;
+
+       if (!capable(CAP_MKNOD))
+               return -EPERM;
+       if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
+               return -EFAULT;
+
+       error = mnt_want_write_file(parfilp);
+       if (error)
+               return error;
+
+       dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
+       if (IS_ERR(dentry)) {
+               mnt_drop_write_file(parfilp);
+               return PTR_ERR(dentry);
+       }
+
+       if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
+               error = -EPERM;
+               goto out;
+       }
+
+       if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
+               error = -EFAULT;
+               goto out;
+       }
+
+       error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
+                                fsd.fsd_dmstate);
+
+ out:
+       mnt_drop_write_file(parfilp);
+       dput(dentry);
+       return error;
+}
+
+STATIC int
+xfs_attrlist_by_handle(
+       struct file             *parfilp,
+       void                    __user *arg)
+{
+       int                     error = -ENOMEM;
+       attrlist_cursor_kern_t  *cursor;
+       xfs_fsop_attrlist_handlereq_t al_hreq;
+       struct dentry           *dentry;
+       char                    *kbuf;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+       if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
+               return -EFAULT;
+       if (al_hreq.buflen < sizeof(struct attrlist) ||
+           al_hreq.buflen > XATTR_LIST_MAX)
+               return -EINVAL;
+
+       /*
+        * Reject flags, only allow namespaces.
+        */
+       if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
+               return -EINVAL;
+
+       dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
+       if (IS_ERR(dentry))
+               return PTR_ERR(dentry);
+
+       kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
+       if (!kbuf)
+               goto out_dput;
+
+       cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
+       error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
+                                       al_hreq.flags, cursor);
+       if (error)
+               goto out_kfree;
+
+       if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
+               error = -EFAULT;
+
+out_kfree:
+       kmem_free(kbuf);
+out_dput:
+       dput(dentry);
+       return error;
+}
+
+int
+xfs_attrmulti_attr_get(
+       struct inode            *inode,
+       unsigned char           *name,
+       unsigned char           __user *ubuf,
+       __uint32_t              *len,
+       __uint32_t              flags)
+{
+       unsigned char           *kbuf;
+       int                     error = -EFAULT;
+
+       if (*len > XATTR_SIZE_MAX)
+               return -EINVAL;
+       kbuf = kmem_zalloc_large(*len, KM_SLEEP);
+       if (!kbuf)
+               return -ENOMEM;
+
+       error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
+       if (error)
+               goto out_kfree;
+
+       if (copy_to_user(ubuf, kbuf, *len))
+               error = -EFAULT;
+
+out_kfree:
+       kmem_free(kbuf);
+       return error;
+}
+
+int
+xfs_attrmulti_attr_set(
+       struct inode            *inode,
+       unsigned char           *name,
+       const unsigned char     __user *ubuf,
+       __uint32_t              len,
+       __uint32_t              flags)
+{
+       unsigned char           *kbuf;
+
+       if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
+               return -EPERM;
+       if (len > XATTR_SIZE_MAX)
+               return -EINVAL;
+
+       kbuf = memdup_user(ubuf, len);
+       if (IS_ERR(kbuf))
+               return PTR_ERR(kbuf);
+
+       return xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
+}
+
+int
+xfs_attrmulti_attr_remove(
+       struct inode            *inode,
+       unsigned char           *name,
+       __uint32_t              flags)
+{
+       if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
+               return -EPERM;
+       return xfs_attr_remove(XFS_I(inode), name, flags);
+}
+
+STATIC int
+xfs_attrmulti_by_handle(
+       struct file             *parfilp,
+       void                    __user *arg)
+{
+       int                     error;
+       xfs_attr_multiop_t      *ops;
+       xfs_fsop_attrmulti_handlereq_t am_hreq;
+       struct dentry           *dentry;
+       unsigned int            i, size;
+       unsigned char           *attr_name;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+       if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
+               return -EFAULT;
+
+       /* overflow check */
+       if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
+               return -E2BIG;
+
+       dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
+       if (IS_ERR(dentry))
+               return PTR_ERR(dentry);
+
+       error = -E2BIG;
+       size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
+       if (!size || size > 16 * PAGE_SIZE)
+               goto out_dput;
+
+       ops = memdup_user(am_hreq.ops, size);
+       if (IS_ERR(ops)) {
+               error = PTR_ERR(ops);
+               goto out_dput;
+       }
+
+       error = -ENOMEM;
+       attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
+       if (!attr_name)
+               goto out_kfree_ops;
+
+       error = 0;
+       for (i = 0; i < am_hreq.opcount; i++) {
+               ops[i].am_error = strncpy_from_user((char *)attr_name,
+                               ops[i].am_attrname, MAXNAMELEN);
+               if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
+                       error = -ERANGE;
+               if (ops[i].am_error < 0)
+                       break;
+
+               switch (ops[i].am_opcode) {
+               case ATTR_OP_GET:
+                       ops[i].am_error = xfs_attrmulti_attr_get(
+                                       d_inode(dentry), attr_name,
+                                       ops[i].am_attrvalue, &ops[i].am_length,
+                                       ops[i].am_flags);
+                       break;
+               case ATTR_OP_SET:
+                       ops[i].am_error = mnt_want_write_file(parfilp);
+                       if (ops[i].am_error)
+                               break;
+                       ops[i].am_error = xfs_attrmulti_attr_set(
+                                       d_inode(dentry), attr_name,
+                                       ops[i].am_attrvalue, ops[i].am_length,
+                                       ops[i].am_flags);
+                       mnt_drop_write_file(parfilp);
+                       break;
+               case ATTR_OP_REMOVE:
+                       ops[i].am_error = mnt_want_write_file(parfilp);
+                       if (ops[i].am_error)
+                               break;
+                       ops[i].am_error = xfs_attrmulti_attr_remove(
+                                       d_inode(dentry), attr_name,
+                                       ops[i].am_flags);
+                       mnt_drop_write_file(parfilp);
+                       break;
+               default:
+                       ops[i].am_error = -EINVAL;
+               }
+       }
+
+       if (copy_to_user(am_hreq.ops, ops, size))
+               error = -EFAULT;
+
+       kfree(attr_name);
+ out_kfree_ops:
+       kfree(ops);
+ out_dput:
+       dput(dentry);
+       return error;
+}
+
+int
+xfs_ioc_space(
+       struct xfs_inode        *ip,
+       struct inode            *inode,
+       struct file             *filp,
+       int                     ioflags,
+       unsigned int            cmd,
+       xfs_flock64_t           *bf)
+{
+       struct iattr            iattr;
+       enum xfs_prealloc_flags flags = 0;
+       uint                    iolock = XFS_IOLOCK_EXCL;
+       int                     error;
+
+       /*
+        * Only allow the sys admin to reserve space unless
+        * unwritten extents are enabled.
+        */
+       if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
+           !capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
+               return -EPERM;
+
+       if (!(filp->f_mode & FMODE_WRITE))
+               return -EBADF;
+
+       if (!S_ISREG(inode->i_mode))
+               return -EINVAL;
+
+       if (filp->f_flags & O_DSYNC)
+               flags |= XFS_PREALLOC_SYNC;
+       if (ioflags & XFS_IO_INVIS)
+               flags |= XFS_PREALLOC_INVISIBLE;
+
+       error = mnt_want_write_file(filp);
+       if (error)
+               return error;
+
+       xfs_ilock(ip, iolock);
+       error = xfs_break_layouts(inode, &iolock, false);
+       if (error)
+               goto out_unlock;
+
+       xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
+       iolock |= XFS_MMAPLOCK_EXCL;
+
+       switch (bf->l_whence) {
+       case 0: /*SEEK_SET*/
+               break;
+       case 1: /*SEEK_CUR*/
+               bf->l_start += filp->f_pos;
+               break;
+       case 2: /*SEEK_END*/
+               bf->l_start += XFS_ISIZE(ip);
+               break;
+       default:
+               error = -EINVAL;
+               goto out_unlock;
+       }
+
+       /*
+        * length of <= 0 for resv/unresv/zero is invalid.  length for
+        * alloc/free is ignored completely and we have no idea what userspace
+        * might have set it to, so set it to zero to allow range
+        * checks to pass.
+        */
+       switch (cmd) {
+       case XFS_IOC_ZERO_RANGE:
+       case XFS_IOC_RESVSP:
+       case XFS_IOC_RESVSP64:
+       case XFS_IOC_UNRESVSP:
+       case XFS_IOC_UNRESVSP64:
+               if (bf->l_len <= 0) {
+                       error = -EINVAL;
+                       goto out_unlock;
+               }
+               break;
+       default:
+               bf->l_len = 0;
+               break;
+       }
+
+       if (bf->l_start < 0 ||
+           bf->l_start > inode->i_sb->s_maxbytes ||
+           bf->l_start + bf->l_len < 0 ||
+           bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
+               error = -EINVAL;
+               goto out_unlock;
+       }
+
+       switch (cmd) {
+       case XFS_IOC_ZERO_RANGE:
+               flags |= XFS_PREALLOC_SET;
+               error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
+               break;
+       case XFS_IOC_RESVSP:
+       case XFS_IOC_RESVSP64:
+               flags |= XFS_PREALLOC_SET;
+               error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
+                                               XFS_BMAPI_PREALLOC);
+               break;
+       case XFS_IOC_UNRESVSP:
+       case XFS_IOC_UNRESVSP64:
+               error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
+               break;
+       case XFS_IOC_ALLOCSP:
+       case XFS_IOC_ALLOCSP64:
+       case XFS_IOC_FREESP:
+       case XFS_IOC_FREESP64:
+               flags |= XFS_PREALLOC_CLEAR;
+               if (bf->l_start > XFS_ISIZE(ip)) {
+                       error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
+                                       bf->l_start - XFS_ISIZE(ip), 0);
+                       if (error)
+                               goto out_unlock;
+               }
+
+               iattr.ia_valid = ATTR_SIZE;
+               iattr.ia_size = bf->l_start;
+
+               error = xfs_setattr_size(ip, &iattr);
+               break;
+       default:
+               ASSERT(0);
+               error = -EINVAL;
+       }
+
+       if (error)
+               goto out_unlock;
+
+       error = xfs_update_prealloc_flags(ip, flags);
+
+out_unlock:
+       xfs_iunlock(ip, iolock);
+       mnt_drop_write_file(filp);
+       return error;
+}
+
+STATIC int
+xfs_ioc_bulkstat(
+       xfs_mount_t             *mp,
+       unsigned int            cmd,
+       void                    __user *arg)
+{
+       xfs_fsop_bulkreq_t      bulkreq;
+       int                     count;  /* # of records returned */
+       xfs_ino_t               inlast; /* last inode number */
+       int                     done;
+       int                     error;
+
+       /* done = 1 if there are more stats to get and if bulkstat */
+       /* should be called again (unused here, but used in dmapi) */
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       if (XFS_FORCED_SHUTDOWN(mp))
+               return -EIO;
+
+       if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
+               return -EFAULT;
+
+       if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
+               return -EFAULT;
+
+       if ((count = bulkreq.icount) <= 0)
+               return -EINVAL;
+
+       if (bulkreq.ubuffer == NULL)
+               return -EINVAL;
+
+       if (cmd == XFS_IOC_FSINUMBERS)
+               error = xfs_inumbers(mp, &inlast, &count,
+                                       bulkreq.ubuffer, xfs_inumbers_fmt);
+       else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
+               error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
+                                       sizeof(xfs_bstat_t), NULL, &done);
+       else    /* XFS_IOC_FSBULKSTAT */
+               error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
+                                    sizeof(xfs_bstat_t), bulkreq.ubuffer,
+                                    &done);
+
+       if (error)
+               return error;
+
+       if (bulkreq.ocount != NULL) {
+               if (copy_to_user(bulkreq.lastip, &inlast,
+                                               sizeof(xfs_ino_t)))
+                       return -EFAULT;
+
+               if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
+                       return -EFAULT;
+       }
+
+       return 0;
+}
+
+STATIC int
+xfs_ioc_fsgeometry_v1(
+       xfs_mount_t             *mp,
+       void                    __user *arg)
+{
+       xfs_fsop_geom_t         fsgeo;
+       int                     error;
+
+       error = xfs_fs_geometry(mp, &fsgeo, 3);
+       if (error)
+               return error;
+
+       /*
+        * Caller should have passed an argument of type
+        * xfs_fsop_geom_v1_t.  This is a proper subset of the
+        * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
+        */
+       if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
+               return -EFAULT;
+       return 0;
+}
+
+STATIC int
+xfs_ioc_fsgeometry(
+       xfs_mount_t             *mp,
+       void                    __user *arg)
+{
+       xfs_fsop_geom_t         fsgeo;
+       int                     error;
+
+       error = xfs_fs_geometry(mp, &fsgeo, 4);
+       if (error)
+               return error;
+
+       if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
+               return -EFAULT;
+       return 0;
+}
+
+/*
+ * Linux extended inode flags interface.
+ */
+
+STATIC unsigned int
+xfs_merge_ioc_xflags(
+       unsigned int    flags,
+       unsigned int    start)
+{
+       unsigned int    xflags = start;
+
+       if (flags & FS_IMMUTABLE_FL)
+               xflags |= XFS_XFLAG_IMMUTABLE;
+       else
+               xflags &= ~XFS_XFLAG_IMMUTABLE;
+       if (flags & FS_APPEND_FL)
+               xflags |= XFS_XFLAG_APPEND;
+       else
+               xflags &= ~XFS_XFLAG_APPEND;
+       if (flags & FS_SYNC_FL)
+               xflags |= XFS_XFLAG_SYNC;
+       else
+               xflags &= ~XFS_XFLAG_SYNC;
+       if (flags & FS_NOATIME_FL)
+               xflags |= XFS_XFLAG_NOATIME;
+       else
+               xflags &= ~XFS_XFLAG_NOATIME;
+       if (flags & FS_NODUMP_FL)
+               xflags |= XFS_XFLAG_NODUMP;
+       else
+               xflags &= ~XFS_XFLAG_NODUMP;
+
+       return xflags;
+}
+
+STATIC unsigned int
+xfs_di2lxflags(
+       __uint16_t      di_flags)
+{
+       unsigned int    flags = 0;
+
+       if (di_flags & XFS_DIFLAG_IMMUTABLE)
+               flags |= FS_IMMUTABLE_FL;
+       if (di_flags & XFS_DIFLAG_APPEND)
+               flags |= FS_APPEND_FL;
+       if (di_flags & XFS_DIFLAG_SYNC)
+               flags |= FS_SYNC_FL;
+       if (di_flags & XFS_DIFLAG_NOATIME)
+               flags |= FS_NOATIME_FL;
+       if (di_flags & XFS_DIFLAG_NODUMP)
+               flags |= FS_NODUMP_FL;
+       return flags;
+}
+
+STATIC int
+xfs_ioc_fsgetxattr(
+       xfs_inode_t             *ip,
+       int                     attr,
+       void                    __user *arg)
+{
+       struct fsxattr          fa;
+
+       memset(&fa, 0, sizeof(struct fsxattr));
+
+       xfs_ilock(ip, XFS_ILOCK_SHARED);
+       fa.fsx_xflags = xfs_ip2xflags(ip);
+       fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
+       fa.fsx_projid = xfs_get_projid(ip);
+
+       if (attr) {
+               if (ip->i_afp) {
+                       if (ip->i_afp->if_flags & XFS_IFEXTENTS)
+                               fa.fsx_nextents = ip->i_afp->if_bytes /
+                                                       sizeof(xfs_bmbt_rec_t);
+                       else
+                               fa.fsx_nextents = ip->i_d.di_anextents;
+               } else
+                       fa.fsx_nextents = 0;
+       } else {
+               if (ip->i_df.if_flags & XFS_IFEXTENTS)
+                       fa.fsx_nextents = ip->i_df.if_bytes /
+                                               sizeof(xfs_bmbt_rec_t);
+               else
+                       fa.fsx_nextents = ip->i_d.di_nextents;
+       }
+       xfs_iunlock(ip, XFS_ILOCK_SHARED);
+
+       if (copy_to_user(arg, &fa, sizeof(fa)))
+               return -EFAULT;
+       return 0;
+}
+
+STATIC void
+xfs_set_diflags(
+       struct xfs_inode        *ip,
+       unsigned int            xflags)
+{
+       unsigned int            di_flags;
+
+       /* can't set PREALLOC this way, just preserve it */
+       di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
+       if (xflags & XFS_XFLAG_IMMUTABLE)
+               di_flags |= XFS_DIFLAG_IMMUTABLE;
+       if (xflags & XFS_XFLAG_APPEND)
+               di_flags |= XFS_DIFLAG_APPEND;
+       if (xflags & XFS_XFLAG_SYNC)
+               di_flags |= XFS_DIFLAG_SYNC;
+       if (xflags & XFS_XFLAG_NOATIME)
+               di_flags |= XFS_DIFLAG_NOATIME;
+       if (xflags & XFS_XFLAG_NODUMP)
+               di_flags |= XFS_DIFLAG_NODUMP;
+       if (xflags & XFS_XFLAG_NODEFRAG)
+               di_flags |= XFS_DIFLAG_NODEFRAG;
+       if (xflags & XFS_XFLAG_FILESTREAM)
+               di_flags |= XFS_DIFLAG_FILESTREAM;
+       if (S_ISDIR(ip->i_d.di_mode)) {
+               if (xflags & XFS_XFLAG_RTINHERIT)
+                       di_flags |= XFS_DIFLAG_RTINHERIT;
+               if (xflags & XFS_XFLAG_NOSYMLINKS)
+                       di_flags |= XFS_DIFLAG_NOSYMLINKS;
+               if (xflags & XFS_XFLAG_EXTSZINHERIT)
+                       di_flags |= XFS_DIFLAG_EXTSZINHERIT;
+               if (xflags & XFS_XFLAG_PROJINHERIT)
+                       di_flags |= XFS_DIFLAG_PROJINHERIT;
+       } else if (S_ISREG(ip->i_d.di_mode)) {
+               if (xflags & XFS_XFLAG_REALTIME)
+                       di_flags |= XFS_DIFLAG_REALTIME;
+               if (xflags & XFS_XFLAG_EXTSIZE)
+                       di_flags |= XFS_DIFLAG_EXTSIZE;
+       }
+
+       ip->i_d.di_flags = di_flags;
+}
+
+STATIC void
+xfs_diflags_to_linux(
+       struct xfs_inode        *ip)
+{
+       struct inode            *inode = VFS_I(ip);
+       unsigned int            xflags = xfs_ip2xflags(ip);
+
+       if (xflags & XFS_XFLAG_IMMUTABLE)
+               inode->i_flags |= S_IMMUTABLE;
+       else
+               inode->i_flags &= ~S_IMMUTABLE;
+       if (xflags & XFS_XFLAG_APPEND)
+               inode->i_flags |= S_APPEND;
+       else
+               inode->i_flags &= ~S_APPEND;
+       if (xflags & XFS_XFLAG_SYNC)
+               inode->i_flags |= S_SYNC;
+       else
+               inode->i_flags &= ~S_SYNC;
+       if (xflags & XFS_XFLAG_NOATIME)
+               inode->i_flags |= S_NOATIME;
+       else
+               inode->i_flags &= ~S_NOATIME;
+}
+
+static int
+xfs_ioctl_setattr_xflags(
+       struct xfs_trans        *tp,
+       struct xfs_inode        *ip,
+       struct fsxattr          *fa)
+{
+       struct xfs_mount        *mp = ip->i_mount;
+
+       /* Can't change realtime flag if any extents are allocated. */
+       if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
+           XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & XFS_XFLAG_REALTIME))
+               return -EINVAL;
+
+       /* If realtime flag is set then must have realtime device */
+       if (fa->fsx_xflags & XFS_XFLAG_REALTIME) {
+               if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
+                   (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
+                       return -EINVAL;
+       }
+
+       /*
+        * Can't modify an immutable/append-only file unless
+        * we have appropriate permission.
+        */
+       if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
+            (fa->fsx_xflags & (XFS_XFLAG_IMMUTABLE | XFS_XFLAG_APPEND))) &&
+           !capable(CAP_LINUX_IMMUTABLE))
+               return -EPERM;
+
+       xfs_set_diflags(ip, fa->fsx_xflags);
+       xfs_diflags_to_linux(ip);
+       xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
+       xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
+       XFS_STATS_INC(xs_ig_attrchg);
+       return 0;
+}
+
+/*
+ * Set up the transaction structure for the setattr operation, checking that we
+ * have permission to do so. On success, return a clean transaction and the
+ * inode locked exclusively ready for further operation specific checks. On
+ * failure, return an error without modifying or locking the inode.
+ */
+static struct xfs_trans *
+xfs_ioctl_setattr_get_trans(
+       struct xfs_inode        *ip)
+{
+       struct xfs_mount        *mp = ip->i_mount;
+       struct xfs_trans        *tp;
+       int                     error;
+
+       if (mp->m_flags & XFS_MOUNT_RDONLY)
+               return ERR_PTR(-EROFS);
+       if (XFS_FORCED_SHUTDOWN(mp))
+               return ERR_PTR(-EIO);
+
+       tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
+       error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
+       if (error)
+               goto out_cancel;
+
+       xfs_ilock(ip, XFS_ILOCK_EXCL);
+       xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
+
+       /*
+        * CAP_FOWNER overrides the following restrictions:
+        *
+        * The user ID of the calling process must be equal to the file owner
+        * ID, except in cases where the CAP_FSETID capability is applicable.
+        */
+       if (!inode_owner_or_capable(VFS_I(ip))) {
+               error = -EPERM;
+               goto out_cancel;
+       }
+
+       if (mp->m_flags & XFS_MOUNT_WSYNC)
+               xfs_trans_set_sync(tp);
+
+       return tp;
+
+out_cancel:
+       xfs_trans_cancel(tp, 0);
+       return ERR_PTR(error);
+}
+
+/*
+ * extent size hint validation is somewhat cumbersome. Rules are:
+ *
+ * 1. extent size hint is only valid for directories and regular files
+ * 2. XFS_XFLAG_EXTSIZE is only valid for regular files
+ * 3. XFS_XFLAG_EXTSZINHERIT is only valid for directories.
+ * 4. can only be changed on regular files if no extents are allocated
+ * 5. can be changed on directories at any time
+ * 6. extsize hint of 0 turns off hints, clears inode flags.
+ * 7. Extent size must be a multiple of the appropriate block size.
+ * 8. for non-realtime files, the extent size hint must be limited
+ *    to half the AG size to avoid alignment extending the extent beyond the
+ *    limits of the AG.
+ */
+static int
+xfs_ioctl_setattr_check_extsize(
+       struct xfs_inode        *ip,
+       struct fsxattr          *fa)
+{
+       struct xfs_mount        *mp = ip->i_mount;
+
+       if ((fa->fsx_xflags & XFS_XFLAG_EXTSIZE) && !S_ISREG(ip->i_d.di_mode))
+               return -EINVAL;
+
+       if ((fa->fsx_xflags & XFS_XFLAG_EXTSZINHERIT) &&
+           !S_ISDIR(ip->i_d.di_mode))
+               return -EINVAL;
+
+       if (S_ISREG(ip->i_d.di_mode) && ip->i_d.di_nextents &&
+           ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
+               return -EINVAL;
+
+       if (fa->fsx_extsize != 0) {
+               xfs_extlen_t    size;
+               xfs_fsblock_t   extsize_fsb;
+
+               extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
+               if (extsize_fsb > MAXEXTLEN)
+                       return -EINVAL;
+
+               if (XFS_IS_REALTIME_INODE(ip) ||
+                   (fa->fsx_xflags & XFS_XFLAG_REALTIME)) {
+                       size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
+               } else {
+                       size = mp->m_sb.sb_blocksize;
+                       if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
+                               return -EINVAL;
+               }
+
+               if (fa->fsx_extsize % size)
+                       return -EINVAL;
+       } else
+               fa->fsx_xflags &= ~(XFS_XFLAG_EXTSIZE | XFS_XFLAG_EXTSZINHERIT);
+
+       return 0;
+}
+
+static int
+xfs_ioctl_setattr_check_projid(
+       struct xfs_inode        *ip,
+       struct fsxattr          *fa)
+{
+       /* Disallow 32bit project ids if projid32bit feature is not enabled. */
+       if (fa->fsx_projid > (__uint16_t)-1 &&
+           !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
+               return -EINVAL;
+
+       /*
+        * Project Quota ID state is only allowed to change from within the init
+        * namespace. Enforce that restriction only if we are trying to change
+        * the quota ID state. Everything else is allowed in user namespaces.
+        */
+       if (current_user_ns() == &init_user_ns)
+               return 0;
+
+       if (xfs_get_projid(ip) != fa->fsx_projid)
+               return -EINVAL;
+       if ((fa->fsx_xflags & XFS_XFLAG_PROJINHERIT) !=
+           (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
+               return -EINVAL;
+
+       return 0;
+}
+
+STATIC int
+xfs_ioctl_setattr(
+       xfs_inode_t             *ip,
+       struct fsxattr          *fa)
+{
+       struct xfs_mount        *mp = ip->i_mount;
+       struct xfs_trans        *tp;
+       struct xfs_dquot        *udqp = NULL;
+       struct xfs_dquot        *pdqp = NULL;
+       struct xfs_dquot        *olddquot = NULL;
+       int                     code;
+
+       trace_xfs_ioctl_setattr(ip);
+
+       code = xfs_ioctl_setattr_check_projid(ip, fa);
+       if (code)
+               return code;
+
+       /*
+        * If disk quotas is on, we make sure that the dquots do exist on disk,
+        * before we start any other transactions. Trying to do this later
+        * is messy. We don't care to take a readlock to look at the ids
+        * in inode here, because we can't hold it across the trans_reserve.
+        * If the IDs do change before we take the ilock, we're covered
+        * because the i_*dquot fields will get updated anyway.
+        */
+       if (XFS_IS_QUOTA_ON(mp)) {
+               code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
+                                        ip->i_d.di_gid, fa->fsx_projid,
+                                        XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
+               if (code)
+                       return code;
+       }
+
+       tp = xfs_ioctl_setattr_get_trans(ip);
+       if (IS_ERR(tp)) {
+               code = PTR_ERR(tp);
+               goto error_free_dquots;
+       }
+
+
+       if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
+           xfs_get_projid(ip) != fa->fsx_projid) {
+               code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
+                               capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
+               if (code)       /* out of quota */
+                       goto error_trans_cancel;
+       }
+
+       code = xfs_ioctl_setattr_check_extsize(ip, fa);
+       if (code)
+               goto error_trans_cancel;
+
+       code = xfs_ioctl_setattr_xflags(tp, ip, fa);
+       if (code)
+               goto error_trans_cancel;
+
+       /*
+        * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
+        * overrides the following restrictions:
+        *
+        * The set-user-ID and set-group-ID bits of a file will be cleared upon
+        * successful return from chown()
+        */
+
+       if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) &&
+           !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
+               ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
+
+       /* Change the ownerships and register project quota modifications */
+       if (xfs_get_projid(ip) != fa->fsx_projid) {
+               if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
+                       olddquot = xfs_qm_vop_chown(tp, ip,
+                                               &ip->i_pdquot, pdqp);
+               }
+               ASSERT(ip->i_d.di_version > 1);
+               xfs_set_projid(ip, fa->fsx_projid);
+       }
+
+       /*
+        * Only set the extent size hint if we've already determined that the
+        * extent size hint should be set on the inode. If no extent size flags
+        * are set on the inode then unconditionally clear the extent size hint.
+        */
+       if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
+               ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
+       else
+               ip->i_d.di_extsize = 0;
+
+       code = xfs_trans_commit(tp, 0);
+
+       /*
+        * Release any dquot(s) the inode had kept before chown.
+        */
+       xfs_qm_dqrele(olddquot);
+       xfs_qm_dqrele(udqp);
+       xfs_qm_dqrele(pdqp);
+
+       return code;
+
+error_trans_cancel:
+       xfs_trans_cancel(tp, 0);
+error_free_dquots:
+       xfs_qm_dqrele(udqp);
+       xfs_qm_dqrele(pdqp);
+       return code;
+}
+
+STATIC int
+xfs_ioc_fssetxattr(
+       xfs_inode_t             *ip,
+       struct file             *filp,
+       void                    __user *arg)
+{
+       struct fsxattr          fa;
+       int error;
+
+       if (copy_from_user(&fa, arg, sizeof(fa)))
+               return -EFAULT;
+
+       error = mnt_want_write_file(filp);
+       if (error)
+               return error;
+       error = xfs_ioctl_setattr(ip, &fa);
+       mnt_drop_write_file(filp);
+       return error;
+}
+
+STATIC int
+xfs_ioc_getxflags(
+       xfs_inode_t             *ip,
+       void                    __user *arg)
+{
+       unsigned int            flags;
+
+       flags = xfs_di2lxflags(ip->i_d.di_flags);
+       if (copy_to_user(arg, &flags, sizeof(flags)))
+               return -EFAULT;
+       return 0;
+}
+
+STATIC int
+xfs_ioc_setxflags(
+       struct xfs_inode        *ip,
+       struct file             *filp,
+       void                    __user *arg)
+{
+       struct xfs_trans        *tp;
+       struct fsxattr          fa;
+       unsigned int            flags;
+       int                     error;
+
+       if (copy_from_user(&flags, arg, sizeof(flags)))
+               return -EFAULT;
+
+       if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
+                     FS_NOATIME_FL | FS_NODUMP_FL | \
+                     FS_SYNC_FL))
+               return -EOPNOTSUPP;
+
+       fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
+
+       error = mnt_want_write_file(filp);
+       if (error)
+               return error;
+
+       tp = xfs_ioctl_setattr_get_trans(ip);
+       if (IS_ERR(tp)) {
+               error = PTR_ERR(tp);
+               goto out_drop_write;
+       }
+
+       error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
+       if (error) {
+               xfs_trans_cancel(tp, 0);
+               goto out_drop_write;
+       }
+
+       error = xfs_trans_commit(tp, 0);
+out_drop_write:
+       mnt_drop_write_file(filp);
+       return error;
+}
+
+STATIC int
+xfs_getbmap_format(void **ap, struct getbmapx *bmv, int *full)
+{
+       struct getbmap __user   *base = (struct getbmap __user *)*ap;
+
+       /* copy only getbmap portion (not getbmapx) */
+       if (copy_to_user(base, bmv, sizeof(struct getbmap)))
+               return -EFAULT;
+
+       *ap += sizeof(struct getbmap);
+       return 0;
+}
+
+STATIC int
+xfs_ioc_getbmap(
+       struct xfs_inode        *ip,
+       int                     ioflags,
+       unsigned int            cmd,
+       void                    __user *arg)
+{
+       struct getbmapx         bmx;
+       int                     error;
+
+       if (copy_from_user(&bmx, arg, sizeof(struct getbmapx)))
+               return -EFAULT;
+
+       if (bmx.bmv_count < 2)
+               return -EINVAL;
+
+       bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
+       if (ioflags & XFS_IO_INVIS)
+               bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
+
+       error = xfs_getbmap(ip, &bmx, xfs_getbmap_format,
+                           (__force struct getbmap *)arg+1);
+       if (error)
+               return error;
+
+       /* copy back header - only size of getbmap */
+       if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
+               return -EFAULT;
+       return 0;
+}
+
+STATIC int
+xfs_getbmapx_format(void **ap, struct getbmapx *bmv, int *full)
+{
+       struct getbmapx __user  *base = (struct getbmapx __user *)*ap;
+
+       if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
+               return -EFAULT;
+
+       *ap += sizeof(struct getbmapx);
+       return 0;
+}
+
+STATIC int
+xfs_ioc_getbmapx(
+       struct xfs_inode        *ip,
+       void                    __user *arg)
+{
+       struct getbmapx         bmx;
+       int                     error;
+
+       if (copy_from_user(&bmx, arg, sizeof(bmx)))
+               return -EFAULT;
+
+       if (bmx.bmv_count < 2)
+               return -EINVAL;
+
+       if (bmx.bmv_iflags & (~BMV_IF_VALID))
+               return -EINVAL;
+
+       error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
+                           (__force struct getbmapx *)arg+1);
+       if (error)
+               return error;
+
+       /* copy back header */
+       if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
+               return -EFAULT;
+
+       return 0;
+}
+
+int
+xfs_ioc_swapext(
+       xfs_swapext_t   *sxp)
+{
+       xfs_inode_t     *ip, *tip;
+       struct fd       f, tmp;
+       int             error = 0;
+
+       /* Pull information for the target fd */
+       f = fdget((int)sxp->sx_fdtarget);
+       if (!f.file) {
+               error = -EINVAL;
+               goto out;
+       }
+
+       if (!(f.file->f_mode & FMODE_WRITE) ||
+           !(f.file->f_mode & FMODE_READ) ||
+           (f.file->f_flags & O_APPEND)) {
+               error = -EBADF;
+               goto out_put_file;
+       }
+
+       tmp = fdget((int)sxp->sx_fdtmp);
+       if (!tmp.file) {
+               error = -EINVAL;
+               goto out_put_file;
+       }
+
+       if (!(tmp.file->f_mode & FMODE_WRITE) ||
+           !(tmp.file->f_mode & FMODE_READ) ||
+           (tmp.file->f_flags & O_APPEND)) {
+               error = -EBADF;
+               goto out_put_tmp_file;
+       }
+
+       if (IS_SWAPFILE(file_inode(f.file)) ||
+           IS_SWAPFILE(file_inode(tmp.file))) {
+               error = -EINVAL;
+               goto out_put_tmp_file;
+       }
+
+       ip = XFS_I(file_inode(f.file));
+       tip = XFS_I(file_inode(tmp.file));
+
+       if (ip->i_mount != tip->i_mount) {
+               error = -EINVAL;
+               goto out_put_tmp_file;
+       }
+
+       if (ip->i_ino == tip->i_ino) {
+               error = -EINVAL;
+               goto out_put_tmp_file;
+       }
+
+       if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
+               error = -EIO;
+               goto out_put_tmp_file;
+       }
+
+       error = xfs_swap_extents(ip, tip, sxp);
+
+ out_put_tmp_file:
+       fdput(tmp);
+ out_put_file:
+       fdput(f);
+ out:
+       return error;
+}
+
+/*
+ * Note: some of the ioctl's return positive numbers as a
+ * byte count indicating success, such as readlink_by_handle.
+ * So we don't "sign flip" like most other routines.  This means
+ * true errors need to be returned as a negative value.
+ */
+long
+xfs_file_ioctl(
+       struct file             *filp,
+       unsigned int            cmd,
+       unsigned long           p)
+{
+       struct inode            *inode = file_inode(filp);
+       struct xfs_inode        *ip = XFS_I(inode);
+       struct xfs_mount        *mp = ip->i_mount;
+       void                    __user *arg = (void __user *)p;
+       int                     ioflags = 0;
+       int                     error;
+
+       if (filp->f_mode & FMODE_NOCMTIME)
+               ioflags |= XFS_IO_INVIS;
+
+       trace_xfs_file_ioctl(ip);
+
+       switch (cmd) {
+       case FITRIM:
+               return xfs_ioc_trim(mp, arg);
+       case XFS_IOC_ALLOCSP:
+       case XFS_IOC_FREESP:
+       case XFS_IOC_RESVSP:
+       case XFS_IOC_UNRESVSP:
+       case XFS_IOC_ALLOCSP64:
+       case XFS_IOC_FREESP64:
+       case XFS_IOC_RESVSP64:
+       case XFS_IOC_UNRESVSP64:
+       case XFS_IOC_ZERO_RANGE: {
+               xfs_flock64_t           bf;
+
+               if (copy_from_user(&bf, arg, sizeof(bf)))
+                       return -EFAULT;
+               return xfs_ioc_space(ip, inode, filp, ioflags, cmd, &bf);
+       }
+       case XFS_IOC_DIOINFO: {
+               struct dioattr  da;
+               xfs_buftarg_t   *target =
+                       XFS_IS_REALTIME_INODE(ip) ?
+                       mp->m_rtdev_targp : mp->m_ddev_targp;
+
+               da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
+               da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
+
+               if (copy_to_user(arg, &da, sizeof(da)))
+                       return -EFAULT;
+               return 0;
+       }
+
+       case XFS_IOC_FSBULKSTAT_SINGLE:
+       case XFS_IOC_FSBULKSTAT:
+       case XFS_IOC_FSINUMBERS:
+               return xfs_ioc_bulkstat(mp, cmd, arg);
+
+       case XFS_IOC_FSGEOMETRY_V1:
+               return xfs_ioc_fsgeometry_v1(mp, arg);
+
+       case XFS_IOC_FSGEOMETRY:
+               return xfs_ioc_fsgeometry(mp, arg);
+
+       case XFS_IOC_GETVERSION:
+               return put_user(inode->i_generation, (int __user *)arg);
+
+       case XFS_IOC_FSGETXATTR:
+               return xfs_ioc_fsgetxattr(ip, 0, arg);
+       case XFS_IOC_FSGETXATTRA:
+               return xfs_ioc_fsgetxattr(ip, 1, arg);
+       case XFS_IOC_FSSETXATTR:
+               return xfs_ioc_fssetxattr(ip, filp, arg);
+       case XFS_IOC_GETXFLAGS:
+               return xfs_ioc_getxflags(ip, arg);
+       case XFS_IOC_SETXFLAGS:
+               return xfs_ioc_setxflags(ip, filp, arg);
+
+       case XFS_IOC_FSSETDM: {
+               struct fsdmidata        dmi;
+
+               if (copy_from_user(&dmi, arg, sizeof(dmi)))
+                       return -EFAULT;
+
+               error = mnt_want_write_file(filp);
+               if (error)
+                       return error;
+
+               error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
+                               dmi.fsd_dmstate);
+               mnt_drop_write_file(filp);
+               return error;
+       }
+
+       case XFS_IOC_GETBMAP:
+       case XFS_IOC_GETBMAPA:
+               return xfs_ioc_getbmap(ip, ioflags, cmd, arg);
+
+       case XFS_IOC_GETBMAPX:
+               return xfs_ioc_getbmapx(ip, arg);
+
+       case XFS_IOC_FD_TO_HANDLE:
+       case XFS_IOC_PATH_TO_HANDLE:
+       case XFS_IOC_PATH_TO_FSHANDLE: {
+               xfs_fsop_handlereq_t    hreq;
+
+               if (copy_from_user(&hreq, arg, sizeof(hreq)))
+                       return -EFAULT;
+               return xfs_find_handle(cmd, &hreq);
+       }
+       case XFS_IOC_OPEN_BY_HANDLE: {
+               xfs_fsop_handlereq_t    hreq;
+
+               if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
+                       return -EFAULT;
+               return xfs_open_by_handle(filp, &hreq);
+       }
+       case XFS_IOC_FSSETDM_BY_HANDLE:
+               return xfs_fssetdm_by_handle(filp, arg);
+
+       case XFS_IOC_READLINK_BY_HANDLE: {
+               xfs_fsop_handlereq_t    hreq;
+
+               if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
+                       return -EFAULT;
+               return xfs_readlink_by_handle(filp, &hreq);
+       }
+       case XFS_IOC_ATTRLIST_BY_HANDLE:
+               return xfs_attrlist_by_handle(filp, arg);
+
+       case XFS_IOC_ATTRMULTI_BY_HANDLE:
+               return xfs_attrmulti_by_handle(filp, arg);
+
+       case XFS_IOC_SWAPEXT: {
+               struct xfs_swapext      sxp;
+
+               if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
+                       return -EFAULT;
+               error = mnt_want_write_file(filp);
+               if (error)
+                       return error;
+               error = xfs_ioc_swapext(&sxp);
+               mnt_drop_write_file(filp);
+               return error;
+       }
+
+       case XFS_IOC_FSCOUNTS: {
+               xfs_fsop_counts_t out;
+
+               error = xfs_fs_counts(mp, &out);
+               if (error)
+                       return error;
+
+               if (copy_to_user(arg, &out, sizeof(out)))
+                       return -EFAULT;
+               return 0;
+       }
+
+       case XFS_IOC_SET_RESBLKS: {
+               xfs_fsop_resblks_t inout;
+               __uint64_t         in;
+
+               if (!capable(CAP_SYS_ADMIN))
+                       return -EPERM;
+
+               if (mp->m_flags & XFS_MOUNT_RDONLY)
+                       return -EROFS;
+
+               if (copy_from_user(&inout, arg, sizeof(inout)))
+                       return -EFAULT;
+
+               error = mnt_want_write_file(filp);
+               if (error)
+                       return error;
+
+               /* input parameter is passed in resblks field of structure */
+               in = inout.resblks;
+               error = xfs_reserve_blocks(mp, &in, &inout);
+               mnt_drop_write_file(filp);
+               if (error)
+                       return error;
+
+               if (copy_to_user(arg, &inout, sizeof(inout)))
+                       return -EFAULT;
+               return 0;
+       }
+
+       case XFS_IOC_GET_RESBLKS: {
+               xfs_fsop_resblks_t out;
+
+               if (!capable(CAP_SYS_ADMIN))
+                       return -EPERM;
+
+               error = xfs_reserve_blocks(mp, NULL, &out);
+               if (error)
+                       return error;
+
+               if (copy_to_user(arg, &out, sizeof(out)))
+                       return -EFAULT;
+
+               return 0;
+       }
+
+       case XFS_IOC_FSGROWFSDATA: {
+               xfs_growfs_data_t in;
+
+               if (copy_from_user(&in, arg, sizeof(in)))
+                       return -EFAULT;
+
+               error = mnt_want_write_file(filp);
+               if (error)
+                       return error;
+               error = xfs_growfs_data(mp, &in);
+               mnt_drop_write_file(filp);
+               return error;
+       }
+
+       case XFS_IOC_FSGROWFSLOG: {
+               xfs_growfs_log_t in;
+
+               if (copy_from_user(&in, arg, sizeof(in)))
+                       return -EFAULT;
+
+               error = mnt_want_write_file(filp);
+               if (error)
+                       return error;
+               error = xfs_growfs_log(mp, &in);
+               mnt_drop_write_file(filp);
+               return error;
+       }
+
+       case XFS_IOC_FSGROWFSRT: {
+               xfs_growfs_rt_t in;
+
+               if (copy_from_user(&in, arg, sizeof(in)))
+                       return -EFAULT;
+
+               error = mnt_want_write_file(filp);
+               if (error)
+                       return error;
+               error = xfs_growfs_rt(mp, &in);
+               mnt_drop_write_file(filp);
+               return error;
+       }
+
+       case XFS_IOC_GOINGDOWN: {
+               __uint32_t in;
+
+               if (!capable(CAP_SYS_ADMIN))
+                       return -EPERM;
+
+               if (get_user(in, (__uint32_t __user *)arg))
+                       return -EFAULT;
+
+               return xfs_fs_goingdown(mp, in);
+       }
+
+       case XFS_IOC_ERROR_INJECTION: {
+               xfs_error_injection_t in;
+
+               if (!capable(CAP_SYS_ADMIN))
+                       return -EPERM;
+
+               if (copy_from_user(&in, arg, sizeof(in)))
+                       return -EFAULT;
+
+               return xfs_errortag_add(in.errtag, mp);
+       }
+
+       case XFS_IOC_ERROR_CLEARALL:
+               if (!capable(CAP_SYS_ADMIN))
+                       return -EPERM;
+
+               return xfs_errortag_clearall(mp, 1);
+
+       case XFS_IOC_FREE_EOFBLOCKS: {
+               struct xfs_fs_eofblocks eofb;
+               struct xfs_eofblocks keofb;
+
+               if (!capable(CAP_SYS_ADMIN))
+                       return -EPERM;
+
+               if (mp->m_flags & XFS_MOUNT_RDONLY)
+                       return -EROFS;
+
+               if (copy_from_user(&eofb, arg, sizeof(eofb)))
+                       return -EFAULT;
+
+               error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
+               if (error)
+                       return error;
+
+               return xfs_icache_free_eofblocks(mp, &keofb);
+       }
+
+       default:
+               return -ENOTTY;
+       }
+}