void
xfs_initialize_vnode(
bhv_desc_t *bdp,
vnode_t *vp,
bhv_desc_t *inode_bhv,
int unlock)
{
xfs_inode_t *ip = XFS_BHVTOI(inode_bhv);
struct inode *inode = LINVFS_GET_IP(vp);
if (!inode_bhv->bd_vobj) {
vp->v_vfsp = bhvtovfs(bdp);
bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops);
bhv_insert(VN_BHV_HEAD(vp), inode_bhv);
}
vp->v_type = IFTOVT(ip->i_d.di_mode);
/* Have we been called during the new inode create process,
* in which case we are too early to fill in the Linux inode.
*/
if (vp->v_type == VNON)
return;
xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip);
/* For new inodes we need to set the ops vectors,
* and unlock the inode.
*/
if (unlock && (inode->i_state & I_NEW)) {
xfs_set_inodeops(inode);
unlock_new_inode(inode);
}
}
ssize_t
xfs_splice_write(
bhv_desc_t *bdp,
struct pipe_inode_info *pipe,
struct file *outfilp,
loff_t *ppos,
size_t count,
int flags,
int ioflags,
cred_t *credp)
{
xfs_inode_t *ip = XFS_BHVTOI(bdp);
xfs_mount_t *mp = ip->i_mount;
ssize_t ret;
struct inode *inode = outfilp->f_mapping->host;
xfs_fsize_t isize;
XFS_STATS_INC(xs_write_calls);
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -EIO;
xfs_ilock(ip, XFS_IOLOCK_EXCL);
if (DM_EVENT_ENABLED(BHV_TO_VNODE(bdp)->v_vfsp, ip, DM_EVENT_WRITE) &&
(!(ioflags & IO_INVIS))) {
bhv_vrwlock_t locktype = VRWLOCK_WRITE;
int error;
error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, BHV_TO_VNODE(bdp),
*ppos, count,
FILP_DELAY_FLAG(outfilp), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
return -error;
}
}
xfs_rw_enter_trace(XFS_SPLICE_WRITE_ENTER, &ip->i_iocore,
pipe, count, *ppos, ioflags);
ret = generic_file_splice_write(pipe, outfilp, ppos, count, flags);
if (ret > 0)
XFS_STATS_ADD(xs_write_bytes, ret);
isize = i_size_read(inode);
if (unlikely(ret < 0 && ret != -EFAULT && *ppos > isize))
*ppos = isize;
if (*ppos > ip->i_d.di_size) {
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (*ppos > ip->i_d.di_size) {
ip->i_d.di_size = *ppos;
i_size_write(inode, *ppos);
ip->i_update_core = 1;
ip->i_update_size = 1;
}
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
return ret;
}
ssize_t
xfs_sendfile(
bhv_desc_t *bdp,
struct file *filp,
loff_t *offset,
int ioflags,
size_t count,
read_actor_t actor,
void *target,
cred_t *credp)
{
ssize_t ret;
xfs_fsize_t n;
xfs_inode_t *ip;
xfs_mount_t *mp;
vnode_t *vp;
ip = XFS_BHVTOI(bdp);
vp = BHV_TO_VNODE(bdp);
mp = ip->i_mount;
vn_trace_entry(vp, "xfs_sendfile", (inst_t *)__return_address);
XFS_STATS_INC(xs_read_calls);
n = XFS_MAXIOFFSET(mp) - *offset;
if ((n <= 0) || (count == 0))
return 0;
if (n < count)
count = n;
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -EIO;
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ) &&
(!(ioflags & IO_INVIS))) {
vrwlock_t locktype = VRWLOCK_READ;
int error;
error = XFS_SEND_DATA(mp, DM_EVENT_READ, BHV_TO_VNODE(bdp), *offset, count,
FILP_DELAY_FLAG(filp), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return -error;
}
}
ret = generic_file_sendfile(filp, offset, count, actor, target);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
XFS_STATS_ADD(xs_read_bytes, ret);
xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
return ret;
}
/*
* Get a XFS inode from a given vnode.
*/
xfs_inode_t *
xfs_vtoi(
struct vnode *vp)
{
bhv_desc_t *bdp;
bdp = bhv_lookup_range(VN_BHV_HEAD(vp),
VNODE_POSITION_XFS, VNODE_POSITION_XFS);
if (unlikely(bdp == NULL))
return NULL;
return XFS_BHVTOI(bdp);
}
int
xfs_dir_lookup_int(
bhv_desc_t *dir_bdp,
uint lock_mode,
vname_t *dentry,
xfs_ino_t *inum,
xfs_inode_t **ipp)
{
vnode_t *dir_vp;
xfs_inode_t *dp;
int error;
dir_vp = BHV_TO_VNODE(dir_bdp);
vn_trace_entry(dir_vp, __FUNCTION__, (inst_t *)__return_address);
dp = XFS_BHVTOI(dir_bdp);
error = XFS_DIR_LOOKUP(dp->i_mount, NULL, dp,
VNAME(dentry), VNAMELEN(dentry), inum);
if (!error) {
/*
* Unlock the directory. We do this because we can't
* hold the directory lock while doing the vn_get()
* in xfs_iget(). Doing so could cause us to hold
* a lock while waiting for the inode to finish
* being inactive while it's waiting for a log
* reservation in the inactive routine.
*/
xfs_iunlock(dp, lock_mode);
error = xfs_iget(dp->i_mount, NULL, *inum, 0, 0, ipp, 0);
xfs_ilock(dp, lock_mode);
if (error) {
*ipp = NULL;
} else if ((*ipp)->i_d.di_mode == 0) {
/*
* The inode has been freed. Something is
* wrong so just get out of here.
*/
xfs_iunlock(dp, lock_mode);
xfs_iput_new(*ipp, 0);
*ipp = NULL;
xfs_ilock(dp, lock_mode);
error = XFS_ERROR(ENOENT);
}
}
return error;
}
int
xfs_bmap(bhv_desc_t *bdp,
xfs_off_t offset,
ssize_t count,
int flags,
xfs_iomap_t *iomapp,
int *niomaps)
{
xfs_inode_t *ip = XFS_BHVTOI(bdp);
xfs_iocore_t *io = &ip->i_iocore;
ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFREG);
ASSERT(((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != 0) ==
((ip->i_iocore.io_flags & XFS_IOCORE_RT) != 0));
return xfs_iomap(io, offset, count, flags, iomapp, niomaps);
}
/*
* xfs_get_dir_entry is used to get a reference to an inode given
* its parent directory inode and the name of the file. It does
* not lock the child inode, and it unlocks the directory before
* returning. The directory's generation number is returned for
* use by a later call to xfs_lock_dir_and_entry.
*/
int
xfs_get_dir_entry(
vname_t *dentry,
xfs_inode_t **ipp)
{
vnode_t *vp;
bhv_desc_t *bdp;
vp = VNAME_TO_VNODE(dentry);
bdp = vn_bhv_lookup_unlocked(VN_BHV_HEAD(vp), &xfs_vnodeops);
if (!bdp) {
*ipp = NULL;
return XFS_ERROR(ENOENT);
}
VN_HOLD(vp);
*ipp = XFS_BHVTOI(bdp);
return 0;
}
ssize_t
xfs_sendfile(
bhv_desc_t *bdp,
struct file *filp,
loff_t *offset,
int ioflags,
size_t count,
read_actor_t actor,
void *target,
cred_t *credp)
{
xfs_inode_t *ip = XFS_BHVTOI(bdp);
xfs_mount_t *mp = ip->i_mount;
ssize_t ret;
XFS_STATS_INC(xs_read_calls);
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(BHV_TO_VNODE(bdp)->v_vfsp, ip, DM_EVENT_READ) &&
(!(ioflags & IO_INVIS))) {
vrwlock_t locktype = VRWLOCK_READ;
int error;
error = XFS_SEND_DATA(mp, DM_EVENT_READ, BHV_TO_VNODE(bdp),
*offset, count,
FILP_DELAY_FLAG(filp), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return -error;
}
}
xfs_rw_enter_trace(XFS_SENDFILE_ENTER, &ip->i_iocore,
(void *)(unsigned long)target, count, *offset, ioflags);
ret = generic_file_sendfile(filp, offset, count, actor, target);
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return ret;
}
ssize_t
xfs_splice_write(
bhv_desc_t *bdp,
struct pipe_inode_info *pipe,
struct file *outfilp,
loff_t *ppos,
size_t count,
int flags,
int ioflags,
cred_t *credp)
{
xfs_inode_t *ip = XFS_BHVTOI(bdp);
xfs_mount_t *mp = ip->i_mount;
ssize_t ret;
XFS_STATS_INC(xs_write_calls);
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -EIO;
xfs_ilock(ip, XFS_IOLOCK_EXCL);
if (DM_EVENT_ENABLED(BHV_TO_VNODE(bdp)->v_vfsp, ip, DM_EVENT_WRITE) &&
(!(ioflags & IO_INVIS))) {
vrwlock_t locktype = VRWLOCK_WRITE;
int error;
error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, BHV_TO_VNODE(bdp),
*ppos, count,
FILP_DELAY_FLAG(outfilp), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
return -error;
}
}
xfs_rw_enter_trace(XFS_SPLICE_WRITE_ENTER, &ip->i_iocore,
pipe, count, *ppos, ioflags);
ret = generic_file_splice_write(pipe, outfilp, ppos, count, flags);
if (ret > 0)
XFS_STATS_ADD(xs_write_bytes, ret);
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
return ret;
}
void
xfs_page_trace(
int tag,
struct inode *inode,
struct page *page,
int mask)
{
xfs_inode_t *ip;
bhv_desc_t *bdp;
vnode_t *vp = LINVFS_GET_VP(inode);
loff_t isize = i_size_read(inode);
loff_t offset = page->index << PAGE_CACHE_SHIFT;
int delalloc = -1, unmapped = -1, unwritten = -1;
if (page_has_buffers(page))
xfs_count_page_state(page, &delalloc, &unmapped, &unwritten);
bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
ip = XFS_BHVTOI(bdp);
if (!ip->i_rwtrace)
return;
ktrace_enter(ip->i_rwtrace,
(void *)((unsigned long)tag),
(void *)ip,
(void *)inode,
(void *)page,
(void *)((unsigned long)mask),
(void *)((unsigned long)((ip->i_d.di_size >> 32) & 0xffffffff)),
(void *)((unsigned long)(ip->i_d.di_size & 0xffffffff)),
(void *)((unsigned long)((isize >> 32) & 0xffffffff)),
(void *)((unsigned long)(isize & 0xffffffff)),
(void *)((unsigned long)((offset >> 32) & 0xffffffff)),
(void *)((unsigned long)(offset & 0xffffffff)),
(void *)((unsigned long)delalloc),
(void *)((unsigned long)unmapped),
(void *)((unsigned long)unwritten),
(void *)NULL,
(void *)NULL);
}
ssize_t /* bytes read, or (-) error */
xfs_read(
bhv_desc_t *bdp,
struct file *file,
char *buf,
size_t size,
loff_t *offset,
int ioflags,
cred_t *credp)
{
ssize_t ret;
xfs_fsize_t n;
xfs_inode_t *ip;
xfs_mount_t *mp;
ip = XFS_BHVTOI(bdp);
mp = ip->i_mount;
XFS_STATS_INC(xs_read_calls);
if (unlikely(ioflags & IO_ISDIRECT)) {
if ((ssize_t)size < 0)
return -XFS_ERROR(EINVAL);
if (((__psint_t)buf & BBMASK) ||
(*offset & mp->m_blockmask) ||
(size & mp->m_blockmask)) {
if (*offset >= ip->i_d.di_size) {
return (0);
}
return -XFS_ERROR(EINVAL);
}
}
n = XFS_MAXIOFFSET(mp) - *offset;
if ((n <= 0) || (size == 0))
return 0;
if (n < size)
size = n;
if (XFS_FORCED_SHUTDOWN(mp)) {
return -EIO;
}
if (!(ioflags & IO_ISLOCKED))
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(BHV_TO_VNODE(bdp)->v_vfsp, ip, DM_EVENT_READ) &&
!(ioflags & IO_INVIS)) {
int error;
vrwlock_t locktype = VRWLOCK_READ;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
error = XFS_SEND_DATA(mp, DM_EVENT_READ, BHV_TO_VNODE(bdp), *offset, size,
dmflags, &locktype);
if (error) {
if (!(ioflags & IO_ISLOCKED))
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return -error;
}
}
if (unlikely(ioflags & IO_ISDIRECT)) {
xfs_rw_enter_trace(XFS_DIORD_ENTER, &ip->i_iocore,
buf, size, *offset, ioflags);
ret = (*offset < ip->i_d.di_size) ?
do_generic_direct_read(file, buf, size, offset) : 0;
UPDATE_ATIME(file->f_dentry->d_inode);
} else {
xfs_rw_enter_trace(XFS_READ_ENTER, &ip->i_iocore,
buf, size, *offset, ioflags);
ret = generic_file_read(file, buf, size, offset);
}
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
if (!(ioflags & IO_ISLOCKED))
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
if (unlikely(ioflags & IO_INVIS)) {
/* generic_file_read updates the atime but we need to
* undo that because this I/O was supposed to be invisible.
*/
struct inode *inode = LINVFS_GET_IP(BHV_TO_VNODE(bdp));
inode->i_atime = ip->i_d.di_atime.t_sec;
} else {
xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
}
return ret;
}
xfs_inode_t *
xfs_bhvtoi(bhv_desc_t *bhvp)
{
return XFS_BHVTOI(bhvp);
}
/*
* 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
*/
STATIC int
xfs_find_handle(
unsigned int cmd,
unsigned long arg)
{
int hsize;
xfs_handle_t handle;
xfs_fsop_handlereq_t hreq;
struct inode *inode;
struct vnode *vp;
if (copy_from_user(&hreq, (xfs_fsop_handlereq_t *)arg, sizeof(hreq)))
return -XFS_ERROR(EFAULT);
memset((char *)&handle, 0, sizeof(handle));
switch (cmd) {
case XFS_IOC_PATH_TO_FSHANDLE:
case XFS_IOC_PATH_TO_HANDLE: {
struct nameidata nd;
int error;
error = user_path_walk_link(hreq.path, &nd);
if (error)
return error;
ASSERT(nd.dentry);
ASSERT(nd.dentry->d_inode);
inode = igrab(nd.dentry->d_inode);
path_release(&nd);
break;
}
case XFS_IOC_FD_TO_HANDLE: {
struct file *file;
file = fget(hreq.fd);
if (!file)
return -EBADF;
ASSERT(file->f_dentry);
ASSERT(file->f_dentry->d_inode);
inode = igrab(file->f_dentry->d_inode);
fput(file);
break;
}
default:
ASSERT(0);
return -XFS_ERROR(EINVAL);
}
if (inode->i_sb->s_magic != XFS_SB_MAGIC) {
/* we're not in XFS anymore, Toto */
iput(inode);
return -XFS_ERROR(EINVAL);
}
/* we need the vnode */
vp = LINVFS_GET_VP(inode);
if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK) {
iput(inode);
return -XFS_ERROR(EBADF);
}
/* now we can grab the fsid */
memcpy(&handle.ha_fsid, vp->v_vfsp->vfs_altfsid, sizeof(xfs_fsid_t));
hsize = sizeof(xfs_fsid_t);
if (cmd != XFS_IOC_PATH_TO_FSHANDLE) {
xfs_inode_t *ip;
bhv_desc_t *bhv;
int lock_mode;
/* need to get access to the xfs_inode to read the generation */
bhv = vn_bhv_lookup_unlocked(VN_BHV_HEAD(vp), &xfs_vnodeops);
ASSERT(bhv);
ip = XFS_BHVTOI(bhv);
ASSERT(ip);
lock_mode = xfs_ilock_map_shared(ip);
/* fill in fid section of handle from inode */
handle.ha_fid.xfs_fid_len = sizeof(xfs_fid_t) -
sizeof(handle.ha_fid.xfs_fid_len);
handle.ha_fid.xfs_fid_pad = 0;
handle.ha_fid.xfs_fid_gen = ip->i_d.di_gen;
handle.ha_fid.xfs_fid_ino = ip->i_ino;
xfs_iunlock_map_shared(ip, lock_mode);
hsize = XFS_HSIZE(handle);
}
/* now copy our handle into the user buffer & write out the size */
if (copy_to_user((xfs_handle_t *)hreq.ohandle, &handle, hsize) ||
copy_to_user(hreq.ohandlen, &hsize, sizeof(__s32))) {
iput(inode);
return -XFS_ERROR(EFAULT);
}
iput(inode);
return 0;
}
int
xfs_ioctl(
bhv_desc_t *bdp,
struct inode *inode,
struct file *filp,
int ioflags,
unsigned int cmd,
unsigned long arg)
{
int error;
vnode_t *vp;
xfs_inode_t *ip;
xfs_mount_t *mp;
vp = LINVFS_GET_VP(inode);
vn_trace_entry(vp, "xfs_ioctl", (inst_t *)__return_address);
ip = XFS_BHVTOI(bdp);
mp = ip->i_mount;
switch (cmd) {
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:
/*
* Only allow the sys admin to reserve space unless
* unwritten extents are enabled.
*/
if (!XFS_SB_VERSION_HASEXTFLGBIT(&mp->m_sb) &&
!capable(CAP_SYS_ADMIN))
return -EPERM;
return xfs_ioc_space(bdp, vp, filp, ioflags, cmd, arg);
case XFS_IOC_DIOINFO: {
struct dioattr da;
da.d_miniosz = mp->m_sb.sb_blocksize;
da.d_mem = mp->m_sb.sb_blocksize;
/*
* this only really needs to be BBSIZE.
* it is set to the file system block size to
* avoid having to do block zeroing on short writes.
*/
da.d_maxiosz = XFS_FSB_TO_B(mp,
XFS_B_TO_FSBT(mp, KIO_MAX_ATOMIC_IO << 10));
if (copy_to_user((struct dioattr *)arg, &da, sizeof(da)))
return -XFS_ERROR(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:
case XFS_IOC_GETXFLAGS:
case XFS_IOC_SETXFLAGS:
case XFS_IOC_FSGETXATTR:
case XFS_IOC_FSSETXATTR:
case XFS_IOC_FSGETXATTRA:
return xfs_ioc_xattr(vp, ip, filp, cmd, arg);
case XFS_IOC_FSSETDM: {
struct fsdmidata dmi;
if (copy_from_user(&dmi, (struct fsdmidata *)arg, sizeof(dmi)))
return -XFS_ERROR(EFAULT);
error = xfs_set_dmattrs(bdp, dmi.fsd_dmevmask, dmi.fsd_dmstate,
NULL);
return -error;
}
case XFS_IOC_GETBMAP:
case XFS_IOC_GETBMAPA:
return xfs_ioc_getbmap(bdp, filp, ioflags, cmd, arg);
case XFS_IOC_GETBMAPX:
return xfs_ioc_getbmapx(bdp, arg);
case XFS_IOC_FD_TO_HANDLE:
case XFS_IOC_PATH_TO_HANDLE:
case XFS_IOC_PATH_TO_FSHANDLE:
return xfs_find_handle(cmd, arg);
case XFS_IOC_OPEN_BY_HANDLE:
return xfs_open_by_handle(mp, arg, filp, inode);
case XFS_IOC_FSSETDM_BY_HANDLE:
return xfs_fssetdm_by_handle(mp, arg, filp, inode);
case XFS_IOC_READLINK_BY_HANDLE:
return xfs_readlink_by_handle(mp, arg, filp, inode);
case XFS_IOC_ATTRLIST_BY_HANDLE:
return xfs_attrlist_by_handle(mp, arg, filp, inode);
case XFS_IOC_ATTRMULTI_BY_HANDLE:
return xfs_attrmulti_by_handle(mp, arg, filp, inode);
case XFS_IOC_SWAPEXT: {
error = xfs_swapext((struct xfs_swapext *)arg);
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((char *)arg, &out, sizeof(out)))
return -XFS_ERROR(EFAULT);
return 0;
}
case XFS_IOC_SET_RESBLKS: {
xfs_fsop_resblks_t inout;
__uint64_t in;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (copy_from_user(&inout, (char *)arg, sizeof(inout)))
return -XFS_ERROR(EFAULT);
/* input parameter is passed in resblks field of structure */
in = inout.resblks;
error = xfs_reserve_blocks(mp, &in, &inout);
if (error)
return -error;
if (copy_to_user((char *)arg, &inout, sizeof(inout)))
return -XFS_ERROR(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((char *)arg, &out, sizeof(out)))
return -XFS_ERROR(EFAULT);
return 0;
}
case XFS_IOC_FSGROWFSDATA: {
xfs_growfs_data_t in;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (copy_from_user(&in, (char *)arg, sizeof(in)))
return -XFS_ERROR(EFAULT);
error = xfs_growfs_data(mp, &in);
return -error;
}
case XFS_IOC_FSGROWFSLOG: {
xfs_growfs_log_t in;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (copy_from_user(&in, (char *)arg, sizeof(in)))
return -XFS_ERROR(EFAULT);
error = xfs_growfs_log(mp, &in);
return -error;
}
case XFS_IOC_FSGROWFSRT: {
xfs_growfs_rt_t in;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (copy_from_user(&in, (char *)arg, sizeof(in)))
return -XFS_ERROR(EFAULT);
error = xfs_growfs_rt(mp, &in);
return -error;
}
case XFS_IOC_FREEZE:
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (vp->v_vfsp->vfs_frozen == SB_UNFROZEN) {
freeze_bdev(mp->m_ddev_targp->pbr_bdev);
if (mp->m_rtdev_targp)
freeze_bdev(mp->m_rtdev_targp->pbr_bdev);
}
return 0;
case XFS_IOC_THAW:
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (vp->v_vfsp->vfs_frozen != SB_UNFROZEN) {
thaw_bdev(mp->m_ddev_targp->pbr_bdev, inode->i_sb);
if (mp->m_rtdev_targp)
thaw_bdev(mp->m_ddev_targp->pbr_bdev, NULL);
}
return 0;
case XFS_IOC_GOINGDOWN: {
__uint32_t in;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (get_user(in, (__uint32_t *)arg))
return -XFS_ERROR(EFAULT);
error = xfs_fs_goingdown(mp, in);
return -error;
}
case XFS_IOC_ERROR_INJECTION: {
xfs_error_injection_t in;
if (!capable(CAP_SYS_ADMIN))
return EPERM;
if (copy_from_user(&in, (char *)arg, sizeof(in)))
return -XFS_ERROR(EFAULT);
error = xfs_errortag_add(in.errtag, mp);
return -error;
}
case XFS_IOC_ERROR_CLEARALL:
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
error = xfs_errortag_clearall(mp);
return -error;
default:
return -ENOTTY;
}
}
/*
* xfs_rename
*/
int
xfs_rename(
bhv_desc_t *src_dir_bdp,
bhv_vname_t *src_vname,
bhv_vnode_t *target_dir_vp,
bhv_vname_t *target_vname,
cred_t *credp)
{
xfs_trans_t *tp;
xfs_inode_t *src_dp, *target_dp, *src_ip, *target_ip;
xfs_mount_t *mp;
int new_parent; /* moving to a new dir */
int src_is_directory; /* src_name is a directory */
int error;
xfs_bmap_free_t free_list;
xfs_fsblock_t first_block;
int cancel_flags;
int committed;
xfs_inode_t *inodes[4];
int target_ip_dropped = 0; /* dropped target_ip link? */
bhv_vnode_t *src_dir_vp;
int spaceres;
int target_link_zero = 0;
int num_inodes;
char *src_name = VNAME(src_vname);
char *target_name = VNAME(target_vname);
int src_namelen = VNAMELEN(src_vname);
int target_namelen = VNAMELEN(target_vname);
src_dir_vp = BHV_TO_VNODE(src_dir_bdp);
vn_trace_entry(src_dir_vp, "xfs_rename", (inst_t *)__return_address);
vn_trace_entry(target_dir_vp, "xfs_rename", (inst_t *)__return_address);
/*
* Find the XFS behavior descriptor for the target directory
* vnode since it was not handed to us.
*/
target_dp = xfs_vtoi(target_dir_vp);
if (target_dp == NULL) {
return XFS_ERROR(EXDEV);
}
src_dp = XFS_BHVTOI(src_dir_bdp);
mp = src_dp->i_mount;
if (DM_EVENT_ENABLED(src_dir_vp->v_vfsp, src_dp, DM_EVENT_RENAME) ||
DM_EVENT_ENABLED(target_dir_vp->v_vfsp,
target_dp, DM_EVENT_RENAME)) {
error = XFS_SEND_NAMESP(mp, DM_EVENT_RENAME,
src_dir_vp, DM_RIGHT_NULL,
target_dir_vp, DM_RIGHT_NULL,
src_name, target_name,
0, 0, 0);
if (error) {
return error;
}
}
/* Return through std_return after this point. */
/*
* Lock all the participating inodes. Depending upon whether
* the target_name exists in the target directory, and
* whether the target directory is the same as the source
* directory, we can lock from 2 to 4 inodes.
* xfs_lock_for_rename() will return ENOENT if src_name
* does not exist in the source directory.
*/
tp = NULL;
error = xfs_lock_for_rename(src_dp, target_dp, src_vname,
target_vname, &src_ip, &target_ip, inodes,
&num_inodes);
if (error) {
/*
* We have nothing locked, no inode references, and
* no transaction, so just get out.
*/
goto std_return;
}
ASSERT(src_ip != NULL);
if ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
/*
* Check for link count overflow on target_dp
*/
if (target_ip == NULL && (src_dp != target_dp) &&
target_dp->i_d.di_nlink >= XFS_MAXLINK) {
error = XFS_ERROR(EMLINK);
xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED);
goto rele_return;
}
}
/*
* If we are using project inheritance, we only allow renames
* into our tree when the project IDs are the same; else the
* tree quota mechanism would be circumvented.
*/
if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
(target_dp->i_d.di_projid != src_ip->i_d.di_projid))) {
error = XFS_ERROR(EXDEV);
xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED);
goto rele_return;
}
new_parent = (src_dp != target_dp);
src_is_directory = ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR);
/*
* Drop the locks on our inodes so that we can start the transaction.
*/
xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED);
XFS_BMAP_INIT(&free_list, &first_block);
tp = xfs_trans_alloc(mp, XFS_TRANS_RENAME);
cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
spaceres = XFS_RENAME_SPACE_RES(mp, target_namelen);
error = xfs_trans_reserve(tp, spaceres, XFS_RENAME_LOG_RES(mp), 0,
XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT);
if (error == ENOSPC) {
spaceres = 0;
error = xfs_trans_reserve(tp, 0, XFS_RENAME_LOG_RES(mp), 0,
XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT);
}
if (error) {
xfs_trans_cancel(tp, 0);
goto rele_return;
}
/*
* Attach the dquots to the inodes
*/
if ((error = XFS_QM_DQVOPRENAME(mp, inodes))) {
xfs_trans_cancel(tp, cancel_flags);
goto rele_return;
}
/*
* Reacquire the inode locks we dropped above.
*/
xfs_lock_inodes(inodes, num_inodes, 0, XFS_ILOCK_EXCL);
/*
* Join all the inodes to the transaction. From this point on,
* we can rely on either trans_commit or trans_cancel to unlock
* them. Note that we need to add a vnode reference to the
* directories since trans_commit & trans_cancel will decrement
* them when they unlock the inodes. Also, we need to be careful
* not to add an inode to the transaction more than once.
*/
VN_HOLD(src_dir_vp);
xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL);
if (new_parent) {
VN_HOLD(target_dir_vp);
xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL);
}
if ((src_ip != src_dp) && (src_ip != target_dp)) {
xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL);
}
if ((target_ip != NULL) &&
(target_ip != src_ip) &&
(target_ip != src_dp) &&
(target_ip != target_dp)) {
xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL);
}
/*
* Set up the target.
*/
if (target_ip == NULL) {
/*
* If there's no space reservation, check the entry will
* fit before actually inserting it.
*/
if (spaceres == 0 &&
(error = xfs_dir_canenter(tp, target_dp, target_name,
target_namelen)))
goto error_return;
/*
* If target does not exist and the rename crosses
* directories, adjust the target directory link count
* to account for the ".." reference from the new entry.
*/
error = xfs_dir_createname(tp, target_dp, target_name,
target_namelen, src_ip->i_ino,
&first_block, &free_list, spaceres);
if (error == ENOSPC)
goto error_return;
if (error)
goto abort_return;
xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
if (new_parent && src_is_directory) {
error = xfs_bumplink(tp, target_dp);
if (error)
goto abort_return;
}
} else { /* target_ip != NULL */
/*
* If target exists and it's a directory, check that both
* target and source are directories and that target can be
* destroyed, or that neither is a directory.
*/
if ((target_ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
/*
* Make sure target dir is empty.
*/
if (!(xfs_dir_isempty(target_ip)) ||
(target_ip->i_d.di_nlink > 2)) {
error = XFS_ERROR(EEXIST);
goto error_return;
}
}
/*
* Link the source inode under the target name.
* If the source inode is a directory and we are moving
* it across directories, its ".." entry will be
* inconsistent until we replace that down below.
*
* In case there is already an entry with the same
* name at the destination directory, remove it first.
*/
error = xfs_dir_replace(tp, target_dp, target_name,
target_namelen, src_ip->i_ino,
&first_block, &free_list, spaceres);
if (error)
goto abort_return;
xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
/*
* Decrement the link count on the target since the target
* dir no longer points to it.
*/
error = xfs_droplink(tp, target_ip);
if (error)
goto abort_return;
target_ip_dropped = 1;
if (src_is_directory) {
/*
* Drop the link from the old "." entry.
*/
error = xfs_droplink(tp, target_ip);
if (error)
goto abort_return;
}
/* Do this test while we still hold the locks */
target_link_zero = (target_ip)->i_d.di_nlink==0;
} /* target_ip != NULL */
/*
* Remove the source.
*/
if (new_parent && src_is_directory) {
/*
* Rewrite the ".." entry to point to the new
* directory.
*/
error = xfs_dir_replace(tp, src_ip, "..", 2, target_dp->i_ino,
&first_block, &free_list, spaceres);
ASSERT(error != EEXIST);
if (error)
goto abort_return;
xfs_ichgtime(src_ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
} else {
/*
* We always want to hit the ctime on the source inode.
* We do it in the if clause above for the 'new_parent &&
* src_is_directory' case, and here we get all the other
* cases. This isn't strictly required by the standards
* since the source inode isn't really being changed,
* but old unix file systems did it and some incremental
* backup programs won't work without it.
*/
xfs_ichgtime(src_ip, XFS_ICHGTIME_CHG);
}
/*
* Adjust the link count on src_dp. This is necessary when
* renaming a directory, either within one parent when
* the target existed, or across two parent directories.
*/
if (src_is_directory && (new_parent || target_ip != NULL)) {
/*
* Decrement link count on src_directory since the
* entry that's moved no longer points to it.
*/
error = xfs_droplink(tp, src_dp);
if (error)
goto abort_return;
}
error = xfs_dir_removename(tp, src_dp, src_name, src_namelen,
src_ip->i_ino, &first_block, &free_list, spaceres);
if (error)
goto abort_return;
xfs_ichgtime(src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
/*
* Update the generation counts on all the directory inodes
* that we're modifying.
*/
src_dp->i_gen++;
xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE);
if (new_parent) {
target_dp->i_gen++;
xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);
}
/*
* If there was a target inode, take an extra reference on
* it here so that it doesn't go to xfs_inactive() from
* within the commit.
*/
if (target_ip != NULL) {
IHOLD(target_ip);
}
/*
* If this is a synchronous mount, make sure that the
* rename transaction goes to disk before returning to
* the user.
*/
if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
xfs_trans_set_sync(tp);
}
/*
* Take refs. for vop_link_removed calls below. No need to worry
* about directory refs. because the caller holds them.
*
* Do holds before the xfs_bmap_finish since it might rele them down
* to zero.
*/
if (target_ip_dropped)
IHOLD(target_ip);
IHOLD(src_ip);
error = xfs_bmap_finish(&tp, &free_list, &committed);
if (error) {
xfs_bmap_cancel(&free_list);
xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES |
XFS_TRANS_ABORT));
if (target_ip != NULL) {
IRELE(target_ip);
}
if (target_ip_dropped) {
IRELE(target_ip);
}
IRELE(src_ip);
goto std_return;
}
/*
* trans_commit will unlock src_ip, target_ip & decrement
* the vnode references.
*/
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
if (target_ip != NULL) {
xfs_refcache_purge_ip(target_ip);
IRELE(target_ip);
}
/*
* Let interposed file systems know about removed links.
*/
if (target_ip_dropped) {
bhv_vop_link_removed(XFS_ITOV(target_ip), target_dir_vp,
target_link_zero);
IRELE(target_ip);
}
IRELE(src_ip);
/* Fall through to std_return with error = 0 or errno from
* xfs_trans_commit */
std_return:
if (DM_EVENT_ENABLED(src_dir_vp->v_vfsp, src_dp, DM_EVENT_POSTRENAME) ||
DM_EVENT_ENABLED(target_dir_vp->v_vfsp,
target_dp, DM_EVENT_POSTRENAME)) {
(void) XFS_SEND_NAMESP (mp, DM_EVENT_POSTRENAME,
src_dir_vp, DM_RIGHT_NULL,
target_dir_vp, DM_RIGHT_NULL,
src_name, target_name,
0, error, 0);
}
return error;
abort_return:
cancel_flags |= XFS_TRANS_ABORT;
/* FALLTHROUGH */
error_return:
xfs_bmap_cancel(&free_list);
xfs_trans_cancel(tp, cancel_flags);
goto std_return;
rele_return:
IRELE(src_ip);
if (target_ip != NULL) {
IRELE(target_ip);
}
goto std_return;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int nsegs,
loff_t *offset,
int ioflags,
cred_t *credp)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
unsigned long segs = nsegs;
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret = 0, error = 0;
xfs_fsize_t isize, new_size;
xfs_iocore_t *io;
vnode_t *vp;
unsigned long seg;
int iolock;
int eventsent = 0;
vrwlock_t locktype;
size_t ocount = 0, count;
loff_t pos;
int need_isem = 1, need_flush = 0;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
xip = XFS_BHVTOI(bdp);
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
ocount += iv->iov_len;
if (unlikely((ssize_t)(ocount|iv->iov_len) < 0))
return -EINVAL;
if (access_ok(VERIFY_READ, iv->iov_base, iv->iov_len))
continue;
if (seg == 0)
return -EFAULT;
segs = seg;
ocount -= iv->iov_len; /* This segment is no good */
break;
}
count = ocount;
pos = *offset;
if (count == 0)
return 0;
io = &xip->i_iocore;
mp = io->io_mount;
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
if (ioflags & IO_ISDIRECT) {
xfs_buftarg_t *target =
(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((pos & target->pbr_smask) || (count & target->pbr_smask))
return XFS_ERROR(-EINVAL);
if (!VN_CACHED(vp) && pos < i_size_read(inode))
need_isem = 0;
if (VN_CACHED(vp))
need_flush = 1;
}
relock:
if (need_isem) {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
down(&inode->i_sem);
} else {
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
}
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = i_size_read(inode);
if (file->f_flags & O_APPEND)
*offset = isize;
start:
error = -generic_write_checks(file, &pos, &count,
S_ISBLK(inode->i_mode));
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
goto out_unlock_isem;
}
new_size = pos + count;
if (new_size > isize)
io->io_new_size = new_size;
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = pos;
int dmflags = FILP_DELAY_FLAG(file);
if (need_isem)
dmflags |= DM_FLAGS_ISEM;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
pos, count,
dmflags, &locktype);
if (error) {
xfs_iunlock(xip, iolock);
goto out_unlock_isem;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) && savedsize != isize) {
pos = isize = xip->i_d.di_size;
goto start;
}
}
/*
* On Linux, generic_file_write updates the times even if
* no data is copied in so long as the write had a size.
*
* We must update xfs' times since revalidate will overcopy xfs.
*/
if (!(ioflags & IO_INVIS)) {
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
inode_update_time(inode, 1);
}
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (pos > isize) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, pos,
isize, pos + count);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
goto out_unlock_isem;
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (likely(!error))
error = -remove_suid(file->f_dentry);
if (unlikely(error)) {
xfs_iunlock(xip, iolock);
goto out_unlock_isem;
}
}
retry:
/* We can write back this queue in page reclaim */
current->backing_dev_info = mapping->backing_dev_info;
if ((ioflags & IO_ISDIRECT)) {
if (need_flush) {
xfs_inval_cached_trace(io, pos, -1,
ctooff(offtoct(pos)), -1);
VOP_FLUSHINVAL_PAGES(vp, ctooff(offtoct(pos)),
-1, FI_REMAPF_LOCKED);
}
if (need_isem) {
/* demote the lock now the cached pages are gone */
XFS_ILOCK_DEMOTE(mp, io, XFS_IOLOCK_EXCL);
up(&inode->i_sem);
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
need_isem = 0;
}
xfs_rw_enter_trace(XFS_DIOWR_ENTER, io, (void *)iovp, segs,
*offset, ioflags);
ret = generic_file_direct_write(iocb, iovp,
&segs, pos, offset, count, ocount);
/*
* direct-io write to a hole: fall through to buffered I/O
* for completing the rest of the request.
*/
if (ret >= 0 && ret != count) {
XFS_STATS_ADD(xs_write_bytes, ret);
pos += ret;
count -= ret;
need_isem = 1;
ioflags &= ~IO_ISDIRECT;
xfs_iunlock(xip, iolock);
goto relock;
}
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER, io, (void *)iovp, segs,
*offset, ioflags);
ret = generic_file_buffered_write(iocb, iovp, segs,
pos, offset, count, ret);
}
current->backing_dev_info = NULL;
if (ret == -EIOCBQUEUED)
ret = wait_on_sync_kiocb(iocb);
if ((ret == -ENOSPC) &&
DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
!(ioflags & IO_INVIS)) {
xfs_rwunlock(bdp, locktype);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (error)
goto out_unlock_isem;
xfs_rwlock(bdp, locktype);
pos = xip->i_d.di_size;
goto retry;
}
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
xip->i_d.di_size = *offset;
i_size_write(inode, *offset);
xip->i_update_core = 1;
xip->i_update_size = 1;
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
error = -ret;
if (ret <= 0)
goto out_unlock_internal;
XFS_STATS_ADD(xs_write_bytes, ret);
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC) &&
!(xip->i_update_size)) {
xfs_inode_log_item_t *iip = xip->i_itemp;
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
if (iip && iip->ili_last_lsn) {
xfs_log_force(mp, iip->ili_last_lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, NULL);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
if (error)
goto out_unlock_internal;
}
}
xfs_rwunlock(bdp, locktype);
if (need_isem)
up(&inode->i_sem);
error = sync_page_range(inode, mapping, pos, ret);
if (!error)
error = ret;
return error;
}
out_unlock_internal:
xfs_rwunlock(bdp, locktype);
out_unlock_isem:
if (need_isem)
up(&inode->i_sem);
return -error;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int nsegs,
loff_t *offset,
int ioflags,
cred_t *credp)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
unsigned long segs = nsegs;
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret = 0, error = 0;
xfs_fsize_t isize, new_size;
xfs_iocore_t *io;
bhv_vnode_t *vp;
unsigned long seg;
int iolock;
int eventsent = 0;
bhv_vrwlock_t locktype;
size_t ocount = 0, count;
loff_t pos;
int need_i_mutex = 1, need_flush = 0;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
xip = XFS_BHVTOI(bdp);
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
ocount += iv->iov_len;
if (unlikely((ssize_t)(ocount|iv->iov_len) < 0))
return -EINVAL;
if (access_ok(VERIFY_READ, iv->iov_base, iv->iov_len))
continue;
if (seg == 0)
return -EFAULT;
segs = seg;
ocount -= iv->iov_len; /* This segment is no good */
break;
}
count = ocount;
pos = *offset;
if (count == 0)
return 0;
io = &xip->i_iocore;
mp = io->io_mount;
vfs_wait_for_freeze(vp->v_vfsp, SB_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
if (ioflags & IO_ISDIRECT) {
xfs_buftarg_t *target =
(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((pos & target->bt_smask) || (count & target->bt_smask))
return XFS_ERROR(-EINVAL);
if (!VN_CACHED(vp) && pos < i_size_read(inode))
need_i_mutex = 0;
if (VN_CACHED(vp))
need_flush = 1;
}
relock:
if (need_i_mutex) {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
mutex_lock(&inode->i_mutex);
} else {
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
}
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = i_size_read(inode);
if (file->f_flags & O_APPEND)
*offset = isize;
start:
error = -generic_write_checks(file, &pos, &count,
S_ISBLK(inode->i_mode));
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
goto out_unlock_mutex;
}
new_size = pos + count;
if (new_size > isize)
io->io_new_size = new_size;
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = pos;
int dmflags = FILP_DELAY_FLAG(file);
if (need_i_mutex)
dmflags |= DM_FLAGS_IMUX;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
pos, count,
dmflags, &locktype);
if (error) {
xfs_iunlock(xip, iolock);
goto out_unlock_mutex;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reacquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) && savedsize != isize) {
pos = isize = xip->i_d.di_size;
goto start;
}
}
if (likely(!(ioflags & IO_INVIS))) {
file_update_time(file);
xfs_ichgtime_fast(xip, inode,
XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
}
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (pos > isize) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, pos, isize);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
goto out_unlock_mutex;
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (likely(!error))
error = -remove_suid(file->f_path.dentry);
if (unlikely(error)) {
xfs_iunlock(xip, iolock);
goto out_unlock_mutex;
}
}
retry:
/* We can write back this queue in page reclaim */
current->backing_dev_info = mapping->backing_dev_info;
if ((ioflags & IO_ISDIRECT)) {
if (need_flush) {
xfs_inval_cached_trace(io, pos, -1,
ctooff(offtoct(pos)), -1);
bhv_vop_flushinval_pages(vp, ctooff(offtoct(pos)),
-1, FI_REMAPF_LOCKED);
}
if (need_i_mutex) {
/* demote the lock now the cached pages are gone */
XFS_ILOCK_DEMOTE(mp, io, XFS_IOLOCK_EXCL);
mutex_unlock(&inode->i_mutex);
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
need_i_mutex = 0;
}
xfs_rw_enter_trace(XFS_DIOWR_ENTER, io, (void *)iovp, segs,
*offset, ioflags);
ret = generic_file_direct_write(iocb, iovp,
&segs, pos, offset, count, ocount);
/*
* direct-io write to a hole: fall through to buffered I/O
* for completing the rest of the request.
*/
if (ret >= 0 && ret != count) {
XFS_STATS_ADD(xs_write_bytes, ret);
pos += ret;
count -= ret;
need_i_mutex = 1;
ioflags &= ~IO_ISDIRECT;
xfs_iunlock(xip, iolock);
goto relock;
}
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER, io, (void *)iovp, segs,
*offset, ioflags);
ret = generic_file_buffered_write(iocb, iovp, segs,
pos, offset, count, ret);
}
current->backing_dev_info = NULL;
if (ret == -EIOCBQUEUED && !(ioflags & IO_ISAIO))
ret = wait_on_sync_kiocb(iocb);
if ((ret == -ENOSPC) &&
DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
!(ioflags & IO_INVIS)) {
xfs_rwunlock(bdp, locktype);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (error)
goto out_nounlocks;
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_rwlock(bdp, locktype);
pos = xip->i_d.di_size;
ret = 0;
goto retry;
}
isize = i_size_read(inode);
if (unlikely(ret < 0 && ret != -EFAULT && *offset > isize))
*offset = isize;
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
xip->i_d.di_size = *offset;
i_size_write(inode, *offset);
xip->i_update_core = 1;
xip->i_update_size = 1;
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
error = -ret;
if (ret <= 0)
goto out_unlock_internal;
XFS_STATS_ADD(xs_write_bytes, ret);
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
error = xfs_write_sync_logforce(mp, xip);
if (error)
goto out_unlock_internal;
xfs_rwunlock(bdp, locktype);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error = sync_page_range(inode, mapping, pos, ret);
if (!error)
error = ret;
return error;
}
out_unlock_internal:
xfs_rwunlock(bdp, locktype);
out_unlock_mutex:
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
out_nounlocks:
return -error;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
uio_t *uio,
int ioflag,
cred_t *credp)
{
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret = 0;
int error = 0;
xfs_fsize_t isize, new_size;
xfs_fsize_t n, limit;
xfs_fsize_t size;
xfs_iocore_t *io;
xfs_vnode_t *vp;
int iolock;
//int eventsent = 0;
vrwlock_t locktype;
xfs_off_t offset_c;
xfs_off_t *offset;
xfs_off_t pos;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
xip = XFS_BHVTOI(bdp);
io = &xip->i_iocore;
mp = io->io_mount;
if (XFS_FORCED_SHUTDOWN(xip->i_mount)) {
return EIO;
}
size = uio->uio_resid;
pos = offset_c = uio->uio_offset;
offset = &offset_c;
if (unlikely(ioflag & IO_ISDIRECT)) {
if (((__psint_t)buf & BBMASK) ||
(*offset & mp->m_blockmask) ||
(size & mp->m_blockmask)) {
return EINVAL;
}
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
} else {
if (io->io_flags & XFS_IOCORE_RT)
return EINVAL;
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
}
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = xip->i_d.di_size;
limit = XFS_MAXIOFFSET(mp);
if (ioflag & O_APPEND)
*offset = isize;
//start:
n = limit - *offset;
if (n <= 0) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return EFBIG;
}
if (n < size)
size = n;
new_size = *offset + size;
if (new_size > isize) {
io->io_new_size = new_size;
}
#ifdef RMC
/* probably be a long time before if ever that we do dmapi */
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = *offset;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
*offset, size,
dmflags, &locktype);
if (error) {
if (iolock) xfs_iunlock(xip, iolock);
return -error;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) &&
savedsize != xip->i_d.di_size) {
*offset = isize = xip->i_d.di_size;
goto start;
}
}
#endif
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (!(ioflag & IO_ISDIRECT) && (*offset > isize && isize)) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, *offset,
isize, *offset + size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return(-error);
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
#if 0
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (likely(!error))
error = -remove_suid(file->f_dentry);
if (unlikely(error)) {
xfs_iunlock(xip, iolock);
goto out_unlock_mutex;
}
}
#endif
//retry:
if (unlikely(ioflag & IO_ISDIRECT)) {
#ifdef RMC
xfs_off_t pos = *offset;
struct address_space *mapping = file->f_dentry->d_inode->i_mapping;
struct inode *inode = mapping->host;
ret = precheck_file_write(file, inode, &size, &pos);
if (ret || size == 0)
goto error;
xfs_inval_cached_pages(vp, io, pos, 1, 1);
inode->i_ctime = inode->i_mtime = CURRENT_TIME;
/* mark_inode_dirty_sync(inode); - we do this later */
xfs_rw_enter_trace(XFS_DIOWR_ENTER, io, buf, size, pos, ioflags);
ret = generic_file_direct_IO(WRITE, file, (char *)buf, size, pos);
xfs_inval_cached_pages(vp, io, pos, 1, 1);
if (ret > 0)
*offset += ret;
#endif
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER, io, buf, size, *offset, ioflags);
ret = xfs_write_file(xip,uio,ioflag);
}
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
//error:
if (ret <= 0) {
if (iolock)
xfs_rwunlock(bdp, locktype);
return ret;
}
XFS_STATS_ADD(xs_write_bytes, ret);
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
printf("xfs_write look at doing more here %s:%d\n",__FILE__,__LINE__);
#ifdef RMC
struct inode *inode = LINVFS_GET_IP(vp);
i_size_write(inode, *offset);
mark_inode_dirty_sync(inode);
#endif
xip->i_d.di_size = *offset;
xip->i_update_core = 1;
xip->i_update_size = 1;
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
/* Handle various SYNC-type writes */
#if 0
// if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
#endif
if (ioflag & IO_SYNC) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC) &&
!(xip->i_update_size)) {
xfs_inode_log_item_t *iip = xip->i_itemp;
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
if (iip && iip->ili_last_lsn) {
xfs_log_force(mp, iip->ili_last_lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, NULL);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
if (error)
goto out_unlock_internal;
}
xfs_rwunlock(bdp, locktype);
return ret;
} /* (ioflags & O_SYNC) */
out_unlock_internal:
xfs_rwunlock(bdp, locktype);
#if 0
out_unlock_mutex:
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
#endif
//out_nounlocks:
return -error;
}
ssize_t /* bytes read, or (-) error */
xfs_read(
bhv_desc_t *bdp,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int segs,
loff_t *offset,
int ioflags,
cred_t *credp)
{
struct file *file = iocb->ki_filp;
size_t size = 0;
ssize_t ret;
xfs_fsize_t n;
xfs_inode_t *ip;
xfs_mount_t *mp;
vnode_t *vp;
unsigned long seg;
ip = XFS_BHVTOI(bdp);
vp = BHV_TO_VNODE(bdp);
mp = ip->i_mount;
vn_trace_entry(vp, "xfs_read", (inst_t *)__return_address);
XFS_STATS_INC(xs_read_calls);
/* START copy & waste from filemap.c */
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
size += iv->iov_len;
if (unlikely((ssize_t)(size|iv->iov_len) < 0))
return XFS_ERROR(-EINVAL);
}
/* END copy & waste from filemap.c */
if (ioflags & IO_ISDIRECT) {
pb_target_t *target =
(ip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((*offset & target->pbr_smask) ||
(size & target->pbr_smask)) {
if (*offset == ip->i_d.di_size) {
return (0);
}
return -XFS_ERROR(EINVAL);
}
}
n = XFS_MAXIOFFSET(mp) - *offset;
if ((n <= 0) || (size == 0))
return 0;
if (n < size)
size = n;
if (XFS_FORCED_SHUTDOWN(mp)) {
return -EIO;
}
/* OK so we are holding the I/O lock for the duration
* of the submission, then what happens if the I/O
* does not really happen here, but is scheduled
* later?
*/
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ) &&
!(ioflags & IO_INVIS)) {
int error;
vrwlock_t locktype = VRWLOCK_READ;
error = XFS_SEND_DATA(mp, DM_EVENT_READ, BHV_TO_VNODE(bdp), *offset, size,
FILP_DELAY_FLAG(file), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return -error;
}
}
ret = __generic_file_aio_read(iocb, iovp, segs, offset);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
if (likely(!(ioflags & IO_INVIS)))
xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
return ret;
}
/*
* Syssgi interface for swapext
*/
int
xfs_swapext(
xfs_swapext_t __user *sxp)
{
xfs_swapext_t sx;
xfs_inode_t *ip=NULL, *tip=NULL, *ips[2];
xfs_trans_t *tp;
xfs_mount_t *mp;
xfs_bstat_t *sbp;
struct file *fp = NULL, *tfp = NULL;
vnode_t *vp, *tvp;
bhv_desc_t *bdp, *tbdp;
vn_bhv_head_t *bhp, *tbhp;
uint lock_flags=0;
int ilf_fields, tilf_fields;
int error = 0;
xfs_ifork_t tempif, *ifp, *tifp;
__uint64_t tmp;
int aforkblks = 0;
int taforkblks = 0;
int locked = 0;
if (copy_from_user(&sx, sxp, sizeof(sx)))
return XFS_ERROR(EFAULT);
/* Pull information for the target fd */
if (((fp = fget((int)sx.sx_fdtarget)) == NULL) ||
((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
bhp = VN_BHV_HEAD(vp);
bdp = vn_bhv_lookup(bhp, &xfs_vnodeops);
if (bdp == NULL) {
error = XFS_ERROR(EBADF);
goto error0;
} else {
ip = XFS_BHVTOI(bdp);
}
if (((tfp = fget((int)sx.sx_fdtmp)) == NULL) ||
((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
tbhp = VN_BHV_HEAD(tvp);
tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops);
if (tbdp == NULL) {
error = XFS_ERROR(EBADF);
goto error0;
} else {
tip = XFS_BHVTOI(tbdp);
}
if (ip->i_mount != tip->i_mount) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (ip->i_ino == tip->i_ino) {
error = XFS_ERROR(EINVAL);
goto error0;
}
mp = ip->i_mount;
sbp = &sx.sx_stat;
if (XFS_FORCED_SHUTDOWN(mp)) {
error = XFS_ERROR(EIO);
goto error0;
}
locked = 1;
/* Lock in i_ino order */
if (ip->i_ino < tip->i_ino) {
ips[0] = ip;
ips[1] = tip;
} else {
ips[0] = tip;
ips[1] = ip;
}
lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
xfs_lock_inodes(ips, 2, 0, lock_flags);
/* Check permissions */
error = xfs_iaccess(ip, S_IWUSR, NULL);
if (error)
goto error0;
error = xfs_iaccess(tip, S_IWUSR, NULL);
if (error)
goto error0;
/* Verify that both files have the same format */
if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify both files are either real-time or non-realtime */
if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
(tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Should never get a local format */
if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (VN_CACHED(tvp) != 0)
xfs_inval_cached_pages(XFS_ITOV(tip), &(tip->i_iocore),
(xfs_off_t)0, 0, 0);
/* Verify O_DIRECT for ftmp */
if (VN_CACHED(tvp) != 0) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify all data are being swapped */
if (sx.sx_offset != 0 ||
sx.sx_length != ip->i_d.di_size ||
sx.sx_length != tip->i_d.di_size) {
error = XFS_ERROR(EFAULT);
goto error0;
}
/*
* If the target has extended attributes, the tmp file
* must also in order to ensure the correct data fork
* format.
*/
if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/*
* Compare the current change & modify times with that
* passed in. If they differ, we abort this swap.
* This is the mechanism used to ensure the calling
* process that the file was not changed out from
* under it.
*/
if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
(sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
(sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
(sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
/* We need to fail if the file is memory mapped. Once we have tossed
* all existing pages, the page fault will have no option
* but to go to the filesystem for pages. By making the page fault call
* VOP_READ (or write in the case of autogrow) they block on the iolock
* until we have switched the extents.
*/
if (VN_MAPPED(vp)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs_iunlock(tip, XFS_ILOCK_EXCL);
/*
* There is a race condition here since we gave up the
* ilock. However, the data fork will not change since
* we have the iolock (locked for truncation too) so we
* are safe. We don't really care if non-io related
* fields change.
*/
VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
if ((error = xfs_trans_reserve(tp, 0,
XFS_ICHANGE_LOG_RES(mp), 0,
0, 0))) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
xfs_iunlock(tip, XFS_IOLOCK_EXCL);
xfs_trans_cancel(tp, 0);
return error;
}
xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
/*
* Count the number of extended attribute blocks
*/
if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
(ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
if (error) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
xfs_trans_cancel(tp, 0);
return error;
}
}
if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
(tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
&taforkblks);
if (error) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
xfs_trans_cancel(tp, 0);
return error;
}
}
/*
* Swap the data forks of the inodes
*/
ifp = &ip->i_df;
tifp = &tip->i_df;
tempif = *ifp; /* struct copy */
*ifp = *tifp; /* struct copy */
*tifp = tempif; /* struct copy */
/*
* Fix the on-disk inode values
*/
tmp = (__uint64_t)ip->i_d.di_nblocks;
ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
tmp = (__uint64_t) ip->i_d.di_nextents;
ip->i_d.di_nextents = tip->i_d.di_nextents;
tip->i_d.di_nextents = tmp;
tmp = (__uint64_t) ip->i_d.di_format;
ip->i_d.di_format = tip->i_d.di_format;
tip->i_d.di_format = tmp;
ilf_fields = XFS_ILOG_CORE;
switch(ip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
ifp->if_u1.if_extents =
ifp->if_u2.if_inline_ext;
}
ilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
ilf_fields |= XFS_ILOG_DBROOT;
break;
}
tilf_fields = XFS_ILOG_CORE;
switch(tip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
tifp->if_u1.if_extents =
tifp->if_u2.if_inline_ext;
}
tilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
tilf_fields |= XFS_ILOG_DBROOT;
break;
}
/*
* Increment vnode ref counts since xfs_trans_commit &
* xfs_trans_cancel will both unlock the inodes and
* decrement the associated ref counts.
*/
VN_HOLD(vp);
VN_HOLD(tvp);
xfs_trans_ijoin(tp, ip, lock_flags);
xfs_trans_ijoin(tp, tip, lock_flags);
xfs_trans_log_inode(tp, ip, ilf_fields);
xfs_trans_log_inode(tp, tip, tilf_fields);
/*
* If this is a synchronous mount, make sure that the
* transaction goes to disk before returning to the user.
*/
if (mp->m_flags & XFS_MOUNT_WSYNC) {
xfs_trans_set_sync(tp);
}
error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
fput(fp);
fput(tfp);
return error;
error0:
if (locked) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
}
if (fp != NULL) fput(fp);
if (tfp != NULL) fput(tfp);
return error;
}
struct xfs_inode *
xfs_vtoi(struct xfs_vnode *xvp)
{
return(XFS_BHVTOI(xvp->v_fbhv));
}
STATIC long
xfs_compat_ioctl(
int mode,
struct file *file,
unsigned cmd,
unsigned long arg)
{
struct inode *inode = file->f_path.dentry->d_inode;
bhv_vnode_t *vp = vn_from_inode(inode);
int error;
switch (cmd) {
case XFS_IOC_DIOINFO:
case XFS_IOC_FSGEOMETRY:
case XFS_IOC_GETVERSION:
case XFS_IOC_GETXFLAGS:
case XFS_IOC_SETXFLAGS:
case XFS_IOC_FSGETXATTR:
case XFS_IOC_FSSETXATTR:
case XFS_IOC_FSGETXATTRA:
case XFS_IOC_FSSETDM:
case XFS_IOC_GETBMAP:
case XFS_IOC_GETBMAPA:
case XFS_IOC_GETBMAPX:
/* not handled
case XFS_IOC_FSSETDM_BY_HANDLE:
case XFS_IOC_ATTRLIST_BY_HANDLE:
case XFS_IOC_ATTRMULTI_BY_HANDLE:
*/
case XFS_IOC_FSCOUNTS:
case XFS_IOC_SET_RESBLKS:
case XFS_IOC_GET_RESBLKS:
case XFS_IOC_FSGROWFSDATA:
case XFS_IOC_FSGROWFSLOG:
case XFS_IOC_FSGROWFSRT:
case XFS_IOC_FREEZE:
case XFS_IOC_THAW:
case XFS_IOC_GOINGDOWN:
case XFS_IOC_ERROR_INJECTION:
case XFS_IOC_ERROR_CLEARALL:
break;
#ifdef BROKEN_X86_ALIGNMENT
/* xfs_flock_t has wrong u32 vs u64 alignment */
case XFS_IOC_ALLOCSP_32:
case XFS_IOC_FREESP_32:
case XFS_IOC_ALLOCSP64_32:
case XFS_IOC_FREESP64_32:
case XFS_IOC_RESVSP_32:
case XFS_IOC_UNRESVSP_32:
case XFS_IOC_RESVSP64_32:
case XFS_IOC_UNRESVSP64_32:
arg = xfs_ioctl32_flock(arg);
cmd = _NATIVE_IOC(cmd, struct xfs_flock64);
break;
case XFS_IOC_FSGEOMETRY_V1_32:
arg = xfs_ioctl32_geom_v1(arg);
cmd = _NATIVE_IOC(cmd, struct xfs_fsop_geom_v1);
break;
#else /* These are handled fine if no alignment issues */
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_FSGEOMETRY_V1:
break;
/* xfs_bstat_t still has wrong u32 vs u64 alignment */
case XFS_IOC_SWAPEXT:
break;
#endif
case XFS_IOC_FSBULKSTAT_32:
case XFS_IOC_FSBULKSTAT_SINGLE_32:
case XFS_IOC_FSINUMBERS_32:
cmd = _NATIVE_IOC(cmd, struct xfs_fsop_bulkreq);
return xfs_ioc_bulkstat_compat(XFS_BHVTOI(VNHEAD(vp))->i_mount,
cmd, (void __user*)arg);
case XFS_IOC_FD_TO_HANDLE_32:
case XFS_IOC_PATH_TO_HANDLE_32:
case XFS_IOC_PATH_TO_FSHANDLE_32:
case XFS_IOC_OPEN_BY_HANDLE_32:
case XFS_IOC_READLINK_BY_HANDLE_32:
arg = xfs_ioctl32_fshandle(arg);
cmd = _NATIVE_IOC(cmd, struct xfs_fsop_handlereq);
break;
default:
return -ENOIOCTLCMD;
}
error = bhv_vop_ioctl(vp, inode, file, mode, cmd, (void __user *)arg);
VMODIFY(vp);
return error;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
struct file *file,
const char *buf,
size_t size,
loff_t *offset,
int ioflags,
cred_t *credp)
{
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret;
int error = 0;
xfs_fsize_t isize, new_size;
xfs_fsize_t n, limit;
xfs_iocore_t *io;
vnode_t *vp;
int iolock;
int eventsent = 0;
vrwlock_t locktype;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
xip = XFS_BHVTOI(bdp);
if (size == 0)
return 0;
io = &xip->i_iocore;
mp = io->io_mount;
fs_check_frozen(vp->v_vfsp, SB_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(xip->i_mount)) {
return -EIO;
}
if (unlikely(ioflags & IO_ISDIRECT)) {
if (((__psint_t)buf & BBMASK) ||
(*offset & mp->m_blockmask) ||
(size & mp->m_blockmask)) {
return XFS_ERROR(-EINVAL);
}
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
} else {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
}
if (ioflags & IO_ISLOCKED)
iolock = 0;
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = xip->i_d.di_size;
limit = XFS_MAXIOFFSET(mp);
if (file->f_flags & O_APPEND)
*offset = isize;
start:
n = limit - *offset;
if (n <= 0) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return -EFBIG;
}
if (n < size)
size = n;
new_size = *offset + size;
if (new_size > isize) {
io->io_new_size = new_size;
}
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = *offset;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
*offset, size,
dmflags, &locktype);
if (error) {
if (iolock) xfs_iunlock(xip, iolock);
return -error;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) &&
savedsize != xip->i_d.di_size) {
*offset = isize = xip->i_d.di_size;
goto start;
}
}
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (!(ioflags & IO_ISDIRECT) && (*offset > isize && isize)) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, *offset,
isize, *offset + size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return(-error);
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (error) {
xfs_iunlock(xip, iolock);
return -error;
}
}
if ((ssize_t) size < 0) {
ret = -EINVAL;
goto error;
}
if (!access_ok(VERIFY_READ, buf, size)) {
ret = -EINVAL;
goto error;
}
retry:
if (unlikely(ioflags & IO_ISDIRECT)) {
xfs_inval_cached_pages(vp, io, *offset, 1, 1);
xfs_rw_enter_trace(XFS_DIOWR_ENTER,
io, buf, size, *offset, ioflags);
ret = do_generic_direct_write(file, buf, size, offset);
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER,
io, buf, size, *offset, ioflags);
ret = do_generic_file_write(file, buf, size, offset);
}
if (unlikely(ioflags & IO_INVIS)) {
/* generic_file_write updates the mtime/ctime but we need
* to undo that because this I/O was supposed to be
* invisible.
*/
struct inode *inode = LINVFS_GET_IP(vp);
inode->i_mtime = xip->i_d.di_mtime.t_sec;
inode->i_ctime = xip->i_d.di_ctime.t_sec;
} else {
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
}
if ((ret == -ENOSPC) &&
DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
!(ioflags & IO_INVIS)) {
xfs_rwunlock(bdp, locktype);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (error)
return -error;
xfs_rwlock(bdp, locktype);
*offset = xip->i_d.di_size;
goto retry;
}
error:
if (ret <= 0) {
if (iolock)
xfs_rwunlock(bdp, locktype);
return ret;
}
XFS_STATS_ADD(xs_write_bytes, ret);
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
struct inode *inode = LINVFS_GET_IP(vp);
xip->i_d.di_size = *offset;
i_size_write(inode, *offset);
xip->i_update_core = 1;
xip->i_update_size = 1;
mark_inode_dirty_sync(inode);
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(file->f_dentry->d_inode)) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC)
&& !(xip->i_update_size)) {
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
xfs_inode_log_item_t *iip;
xfs_lsn_t lsn;
iip = xip->i_itemp;
if (iip && iip->ili_last_lsn) {
lsn = iip->ili_last_lsn;
xfs_log_force(mp, lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, NULL);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
}
} /* (ioflags & O_SYNC) */
/*
* If we are coming from an nfsd thread then insert into the
* reference cache.
*/
if (!strcmp(current->comm, "nfsd"))
xfs_refcache_insert(xip);
/* Drop lock this way - the old refcache release is in here */
if (iolock)
xfs_rwunlock(bdp, locktype);
return(ret);
}
ssize_t /* bytes read, or (-) error */
xfs_read(
bhv_desc_t *bdp,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int segs,
loff_t *offset,
int ioflags,
cred_t *credp)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file->f_mapping->host;
size_t size = 0;
ssize_t ret;
xfs_fsize_t n;
xfs_inode_t *ip;
xfs_mount_t *mp;
vnode_t *vp;
unsigned long seg;
ip = XFS_BHVTOI(bdp);
vp = BHV_TO_VNODE(bdp);
mp = ip->i_mount;
XFS_STATS_INC(xs_read_calls);
/* START copy & waste from filemap.c */
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
size += iv->iov_len;
if (unlikely((ssize_t)(size|iv->iov_len) < 0))
return XFS_ERROR(-EINVAL);
}
/* END copy & waste from filemap.c */
if (unlikely(ioflags & IO_ISDIRECT)) {
xfs_buftarg_t *target =
(ip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((*offset & target->pbr_smask) ||
(size & target->pbr_smask)) {
if (*offset == ip->i_d.di_size) {
return (0);
}
return -XFS_ERROR(EINVAL);
}
}
n = XFS_MAXIOFFSET(mp) - *offset;
if ((n <= 0) || (size == 0))
return 0;
if (n < size)
size = n;
if (XFS_FORCED_SHUTDOWN(mp)) {
return -EIO;
}
if (unlikely(ioflags & IO_ISDIRECT))
down(&inode->i_sem);
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ) &&
!(ioflags & IO_INVIS)) {
vrwlock_t locktype = VRWLOCK_READ;
ret = -XFS_SEND_DATA(mp, DM_EVENT_READ,
BHV_TO_VNODE(bdp), *offset, size,
FILP_DELAY_FLAG(file), &locktype);
if (ret) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
goto unlock_isem;
}
}
xfs_rw_enter_trace(XFS_READ_ENTER, &ip->i_iocore,
(void *)iovp, segs, *offset, ioflags);
ret = __generic_file_aio_read(iocb, iovp, segs, offset);
if (ret == -EIOCBQUEUED)
ret = wait_on_sync_kiocb(iocb);
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
if (likely(!(ioflags & IO_INVIS)))
xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
unlock_isem:
if (unlikely(ioflags & IO_ISDIRECT))
up(&inode->i_sem);
return ret;
}
ssize_t /* bytes read, or (-) error */
xfs_read(
bhv_desc_t *bdp,
uio_t *uio,
int ioflags,
cred_t *credp)
{
ssize_t ret, size;
xfs_fsize_t n;
xfs_inode_t *ip;
xfs_mount_t *mp;
ip = XFS_BHVTOI(bdp);
mp = ip->i_mount;
XFS_STATS_INC(xs_read_calls);
if (unlikely(ioflags & IO_ISDIRECT)) {
if (((__psint_t)buf & BBMASK) ||
(uio->uio_offset & mp->m_blockmask) ||
(uio->uio_resid & mp->m_blockmask)) {
if (uio->uio_offset >= ip->i_d.di_size) {
return (0);
}
return EINVAL;
}
}
if (uio->uio_resid == 0)
return 0;
n = XFS_MAXIOFFSET(mp) - uio->uio_offset;
if (n <= 0)
return EFBIG;
size = (n < uio->uio_resid)? n : uio->uio_resid;
if (XFS_FORCED_SHUTDOWN(mp)) {
return EIO;
}
xfs_ilock(ip, XFS_IOLOCK_SHARED);
#ifdef XXX
if (DM_EVENT_ENABLED(BHV_TO_VNODE(bdp)->v_vfsp, ip, DM_EVENT_READ) &&
!(ioflags & IO_INVIS)) {
int error;
vrwlock_t locktype = VRWLOCK_READ;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
error = XFS_SEND_DATA(mp, DM_EVENT_READ, BHV_TO_VNODE(bdp),
uio->uio_offset, size, dmflags, &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return (error);
}
}
#endif
ret = xfs_read_file(mp, ip, uio, ioflags);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
XFS_STATS_ADD(xs_read_bytes, ret);
if (likely((ioflags & IO_INVIS) == 0)) {
xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
}
return ret;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int segs,
loff_t *offset,
int ioflags,
cred_t *credp)
{
struct file *file = iocb->ki_filp;
size_t size = 0;
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret;
int error = 0;
xfs_fsize_t isize, new_size;
xfs_fsize_t n, limit;
xfs_iocore_t *io;
vnode_t *vp;
unsigned long seg;
int iolock;
int eventsent = 0;
vrwlock_t locktype;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
vn_trace_entry(vp, "xfs_write", (inst_t *)__return_address);
xip = XFS_BHVTOI(bdp);
/* START copy & waste from filemap.c */
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
size += iv->iov_len;
if (unlikely((ssize_t)(size|iv->iov_len) < 0))
return XFS_ERROR(-EINVAL);
}
/* END copy & waste from filemap.c */
if (size == 0)
return 0;
io = &(xip->i_iocore);
mp = io->io_mount;
xfs_check_frozen(mp, bdp, XFS_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(mp)) {
return -EIO;
}
if (ioflags & IO_ISDIRECT) {
pb_target_t *target =
(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((*offset & target->pbr_smask) ||
(size & target->pbr_smask)) {
return XFS_ERROR(-EINVAL);
}
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
} else {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
}
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = xip->i_d.di_size;
limit = XFS_MAXIOFFSET(mp);
if (file->f_flags & O_APPEND)
*offset = isize;
start:
n = limit - *offset;
if (n <= 0) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return -EFBIG;
}
if (n < size)
size = n;
new_size = *offset + size;
if (new_size > isize) {
io->io_new_size = new_size;
}
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = *offset;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
*offset, size,
FILP_DELAY_FLAG(file), &locktype);
if (error) {
xfs_iunlock(xip, iolock);
return -error;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) &&
savedsize != xip->i_d.di_size) {
*offset = isize = xip->i_d.di_size;
goto start;
}
}
/*
* On Linux, generic_file_write updates the times even if
* no data is copied in so long as the write had a size.
*
* We must update xfs' times since revalidate will overcopy xfs.
*/
if (size && !(ioflags & IO_INVIS))
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (!(ioflags & IO_ISDIRECT) && (*offset > isize && isize)) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, *offset,
isize, *offset + size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return(-error);
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (error) {
xfs_iunlock(xip, iolock);
return -error;
}
}
retry:
if (ioflags & IO_ISDIRECT) {
xfs_inval_cached_pages(vp, &xip->i_iocore, *offset, 1, 1);
}
ret = generic_file_aio_write_nolock(iocb, iovp, segs, offset);
if ((ret == -ENOSPC) &&
DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
!(ioflags & IO_INVIS)) {
xfs_rwunlock(bdp, locktype);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (error)
return -error;
xfs_rwlock(bdp, locktype);
*offset = xip->i_d.di_size;
goto retry;
}
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
struct inode *inode = LINVFS_GET_IP(vp);
xip->i_d.di_size = *offset;
i_size_write(inode, *offset);
xip->i_update_core = 1;
xip->i_update_size = 1;
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
if (ret <= 0) {
xfs_rwunlock(bdp, locktype);
return ret;
}
XFS_STATS_ADD(xs_write_bytes, ret);
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(file->f_dentry->d_inode)) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC)
&& !(xip->i_update_size)) {
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
xfs_inode_log_item_t *iip;
xfs_lsn_t lsn;
iip = xip->i_itemp;
if (iip && iip->ili_last_lsn) {
lsn = iip->ili_last_lsn;
xfs_log_force(mp, lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, (xfs_lsn_t)0);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
}
} /* (ioflags & O_SYNC) */
xfs_rwunlock(bdp, locktype);
return(ret);
}
