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);
}
STATIC int
xfs_ioctl_setattr(
xfs_inode_t *ip,
struct fsxattr *fa,
int mask)
{
struct xfs_mount *mp = ip->i_mount;
struct xfs_trans *tp;
unsigned int lock_flags = 0;
struct xfs_dquot *udqp = NULL;
struct xfs_dquot *gdqp = NULL;
struct xfs_dquot *olddquot = NULL;
int code;
xfs_itrace_entry(ip);
if (mp->m_flags & XFS_MOUNT_RDONLY)
return XFS_ERROR(EROFS);
if (XFS_FORCED_SHUTDOWN(mp))
return XFS_ERROR(EIO);
/*
* 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) && (mask & FSX_PROJID)) {
code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
ip->i_d.di_gid, fa->fsx_projid,
XFS_QMOPT_PQUOTA, &udqp, &gdqp);
if (code)
return code;
}
/*
* For the other attributes, we acquire the inode lock and
* first do an error checking pass.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
code = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0);
if (code)
goto error_return;
lock_flags = XFS_ILOCK_EXCL;
xfs_ilock(ip, lock_flags);
/*
* 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 (current_fsuid() != ip->i_d.di_uid && !capable(CAP_FOWNER)) {
code = XFS_ERROR(EPERM);
goto error_return;
}
/*
* Do a quota reservation only if projid is actually going to change.
*/
if (mask & FSX_PROJID) {
if (XFS_IS_QUOTA_RUNNING(mp) &&
XFS_IS_PQUOTA_ON(mp) &&
ip->i_d.di_projid != fa->fsx_projid) {
ASSERT(tp);
code = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp,
capable(CAP_FOWNER) ?
XFS_QMOPT_FORCE_RES : 0);
if (code) /* out of quota */
goto error_return;
}
}
if (mask & FSX_EXTSIZE) {
/*
* Can't change extent size if any extents are allocated.
*/
if (ip->i_d.di_nextents &&
((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) !=
fa->fsx_extsize)) {
code = XFS_ERROR(EINVAL); /* EFBIG? */
goto error_return;
}
/*
* Extent size must be a multiple of the appropriate block
* size, if set at all.
*/
if (fa->fsx_extsize != 0) {
xfs_extlen_t size;
if (XFS_IS_REALTIME_INODE(ip) ||
((mask & FSX_XFLAGS) &&
(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 (fa->fsx_extsize % size) {
code = XFS_ERROR(EINVAL);
goto error_return;
}
}
}
if (mask & FSX_XFLAGS) {
/*
* 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)) {
code = XFS_ERROR(EINVAL); /* EFBIG? */
goto error_return;
}
/*
* If realtime flag is set then must have realtime data.
*/
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)) {
code = XFS_ERROR(EINVAL);
goto error_return;
}
}
/*
* 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)) {
code = XFS_ERROR(EPERM);
goto error_return;
}
}
xfs_trans_ijoin(tp, ip, lock_flags);
xfs_trans_ihold(tp, ip);
/*
* Change file ownership. Must be the owner or privileged.
*/
if (mask & FSX_PROJID) {
/*
* 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(CAP_FSETID))
ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
/*
* Change the ownerships and register quota modifications
* in the transaction.
*/
if (ip->i_d.di_projid != fa->fsx_projid) {
if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
olddquot = xfs_qm_vop_chown(tp, ip,
&ip->i_gdquot, gdqp);
}
ip->i_d.di_projid = fa->fsx_projid;
/*
* We may have to rev the inode as well as
* the superblock version number since projids didn't
* exist before DINODE_VERSION_2 and SB_VERSION_NLINK.
*/
if (ip->i_d.di_version == 1)
xfs_bump_ino_vers2(tp, ip);
}
}
if (mask & FSX_EXTSIZE)
ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
if (mask & FSX_XFLAGS) {
xfs_set_diflags(ip, fa->fsx_xflags);
xfs_diflags_to_linux(ip);
}
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
xfs_ichgtime(ip, XFS_ICHGTIME_CHG);
XFS_STATS_INC(xs_ig_attrchg);
/*
* If this is a synchronous mount, make sure that the
* transaction goes to disk before returning to the user.
* This is slightly sub-optimal in that truncates require
* two sync transactions instead of one for wsync filesystems.
* One for the truncate and one for the timestamps since we
* don't want to change the timestamps unless we're sure the
* truncate worked. Truncates are less than 1% of the laddis
* mix so this probably isn't worth the trouble to optimize.
*/
if (mp->m_flags & XFS_MOUNT_WSYNC)
xfs_trans_set_sync(tp);
code = xfs_trans_commit(tp, 0);
xfs_iunlock(ip, lock_flags);
/*
* Release any dquot(s) the inode had kept before chown.
*/
xfs_qm_dqrele(olddquot);
xfs_qm_dqrele(udqp);
xfs_qm_dqrele(gdqp);
if (code)
return code;
if (DM_EVENT_ENABLED(ip, DM_EVENT_ATTRIBUTE)) {
XFS_SEND_NAMESP(mp, DM_EVENT_ATTRIBUTE, ip, DM_RIGHT_NULL,
NULL, DM_RIGHT_NULL, NULL, NULL, 0, 0,
(mask & FSX_NONBLOCK) ? DM_FLAGS_NDELAY : 0);
}
return 0;
error_return:
xfs_qm_dqrele(udqp);
xfs_qm_dqrele(gdqp);
xfs_trans_cancel(tp, 0);
if (lock_flags)
xfs_iunlock(ip, lock_flags);
return code;
}
/*
* 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;
}
/*
* xfs_rename
*/
int
xfs_rename(
xfs_inode_t *src_dp,
struct xfs_name *src_name,
xfs_inode_t *src_ip,
xfs_inode_t *target_dp,
struct xfs_name *target_name,
xfs_inode_t *target_ip)
{
xfs_trans_t *tp = NULL;
xfs_mount_t *mp = src_dp->i_mount;
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 spaceres;
int num_inodes;
xfs_itrace_entry(src_dp);
xfs_itrace_entry(target_dp);
if (DM_EVENT_ENABLED(src_dp, DM_EVENT_RENAME) ||
DM_EVENT_ENABLED(target_dp, DM_EVENT_RENAME)) {
error = XFS_SEND_NAMESP(mp, DM_EVENT_RENAME,
src_dp, DM_RIGHT_NULL,
target_dp, DM_RIGHT_NULL,
src_name->name, target_name->name,
0, 0, 0);
if (error)
return error;
}
/* Return through std_return after this point. */
new_parent = (src_dp != target_dp);
src_is_directory = ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR);
if (src_is_directory) {
/*
* Check for link count overflow on target_dp
*/
if (target_ip == NULL && new_parent &&
target_dp->i_d.di_nlink >= XFS_MAXLINK) {
error = XFS_ERROR(EMLINK);
goto std_return;
}
}
xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip,
inodes, &num_inodes);
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_name->len);
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 std_return;
}
/*
* Attach the dquots to the inodes
*/
if ((error = XFS_QM_DQVOPRENAME(mp, inodes))) {
xfs_trans_cancel(tp, cancel_flags);
goto std_return;
}
/*
* 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_inodes(inodes, num_inodes, 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.
*/
IHOLD(src_dp);
xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL);
if (new_parent) {
IHOLD(target_dp);
xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL);
}
IHOLD(src_ip);
xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL);
if (target_ip) {
IHOLD(target_ip);
xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL);
}
/*
* 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);
goto error_return;
}
/*
* Set up the target.
*/
if (target_ip == NULL) {
/*
* If there's no space reservation, check the entry will
* fit before actually inserting it.
*/
error = xfs_dir_canenter(tp, target_dp, target_name, spaceres);
if (error)
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,
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,
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;
if (src_is_directory) {
/*
* Drop the link from the old "." entry.
*/
error = xfs_droplink(tp, target_ip);
if (error)
goto abort_return;
}
} /* 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, &xfs_name_dotdot,
target_dp->i_ino,
&first_block, &free_list, spaceres);
ASSERT(error != EEXIST);
if (error)
goto abort_return;
}
/*
* We always want to hit the ctime on the source inode.
*
* 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_ip->i_ino,
&first_block, &free_list, spaceres);
if (error)
goto abort_return;
xfs_ichgtime(src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE);
if (new_parent)
xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);
/*
* 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);
}
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));
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);
/* Fall through to std_return with error = 0 or errno from
* xfs_trans_commit */
std_return:
if (DM_EVENT_ENABLED(src_dp, DM_EVENT_POSTRENAME) ||
DM_EVENT_ENABLED(target_dp, DM_EVENT_POSTRENAME)) {
(void) XFS_SEND_NAMESP (mp, DM_EVENT_POSTRENAME,
src_dp, DM_RIGHT_NULL,
target_dp, DM_RIGHT_NULL,
src_name->name, target_name->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;
}
STATIC ssize_t
xfs_file_aio_write(
struct kiocb *iocb,
const struct iovec *iovp,
unsigned long nr_segs,
loff_t pos)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
ssize_t ret = 0, error = 0;
int ioflags = 0;
xfs_fsize_t isize, new_size;
int iolock;
int eventsent = 0;
size_t ocount = 0, count;
int need_i_mutex;
XFS_STATS_INC(xs_write_calls);
BUG_ON(iocb->ki_pos != pos);
if (unlikely(file->f_flags & O_DIRECT))
ioflags |= IO_ISDIRECT;
if (file->f_mode & FMODE_NOCMTIME)
ioflags |= IO_INVIS;
error = generic_segment_checks(iovp, &nr_segs, &ocount, VERIFY_READ);
if (error)
return error;
count = ocount;
if (count == 0)
return 0;
xfs_wait_for_freeze(mp, SB_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
relock:
if (ioflags & IO_ISDIRECT) {
iolock = XFS_IOLOCK_SHARED;
need_i_mutex = 0;
} else {
iolock = XFS_IOLOCK_EXCL;
need_i_mutex = 1;
mutex_lock(&inode->i_mutex);
}
xfs_ilock(ip, XFS_ILOCK_EXCL|iolock);
start:
error = -generic_write_checks(file, &pos, &count,
S_ISBLK(inode->i_mode));
if (error) {
xfs_iunlock(ip, XFS_ILOCK_EXCL|iolock);
goto out_unlock_mutex;
}
if ((DM_EVENT_ENABLED(ip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
int dmflags = FILP_DELAY_FLAG(file);
if (need_i_mutex)
dmflags |= DM_FLAGS_IMUX;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(ip->i_mount, DM_EVENT_WRITE, ip,
pos, count, dmflags, &iolock);
if (error) {
goto out_unlock_internal;
}
xfs_ilock(ip, 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) && pos != ip->i_size)
goto start;
}
if (ioflags & IO_ISDIRECT) {
xfs_buftarg_t *target =
XFS_IS_REALTIME_INODE(ip) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((pos & target->bt_smask) || (count & target->bt_smask)) {
xfs_iunlock(ip, XFS_ILOCK_EXCL|iolock);
return XFS_ERROR(-EINVAL);
}
if (!need_i_mutex && (mapping->nrpages || pos > ip->i_size)) {
xfs_iunlock(ip, XFS_ILOCK_EXCL|iolock);
iolock = XFS_IOLOCK_EXCL;
need_i_mutex = 1;
mutex_lock(&inode->i_mutex);
xfs_ilock(ip, XFS_ILOCK_EXCL|iolock);
goto start;
}
}
new_size = pos + count;
if (new_size > ip->i_size)
ip->i_new_size = new_size;
if (likely(!(ioflags & IO_INVIS)))
file_update_time(file);
/*
* 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 > ip->i_size) {
error = xfs_zero_eof(ip, pos, ip->i_size);
if (error) {
xfs_iunlock(ip, XFS_ILOCK_EXCL);
goto out_unlock_internal;
}
}
xfs_iunlock(ip, 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.
*/
error = -file_remove_suid(file);
if (unlikely(error))
goto out_unlock_internal;
/* We can write back this queue in page reclaim */
current->backing_dev_info = mapping->backing_dev_info;
if ((ioflags & IO_ISDIRECT)) {
if (mapping->nrpages) {
WARN_ON(need_i_mutex == 0);
error = xfs_flushinval_pages(ip,
(pos & PAGE_CACHE_MASK),
-1, FI_REMAPF_LOCKED);
if (error)
goto out_unlock_internal;
}
if (need_i_mutex) {
/* demote the lock now the cached pages are gone */
xfs_ilock_demote(ip, XFS_IOLOCK_EXCL);
mutex_unlock(&inode->i_mutex);
iolock = XFS_IOLOCK_SHARED;
need_i_mutex = 0;
}
trace_xfs_file_direct_write(ip, count, iocb->ki_pos, ioflags);
ret = generic_file_direct_write(iocb, iovp,
&nr_segs, pos, &iocb->ki_pos, 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;
ioflags &= ~IO_ISDIRECT;
xfs_iunlock(ip, iolock);
goto relock;
}
} else {
int enospc = 0;
ssize_t ret2 = 0;
write_retry:
trace_xfs_file_buffered_write(ip, count, iocb->ki_pos, ioflags);
ret2 = generic_file_buffered_write(iocb, iovp, nr_segs,
pos, &iocb->ki_pos, count, ret);
/*
* if we just got an ENOSPC, flush the inode now we
* aren't holding any page locks and retry *once*
*/
if (ret2 == -ENOSPC && !enospc) {
error = xfs_flush_pages(ip, 0, -1, 0, FI_NONE);
if (error)
goto out_unlock_internal;
enospc = 1;
goto write_retry;
}
ret = ret2;
}
current->backing_dev_info = NULL;
isize = i_size_read(inode);
if (unlikely(ret < 0 && ret != -EFAULT && iocb->ki_pos > isize))
iocb->ki_pos = isize;
if (iocb->ki_pos > ip->i_size) {
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (iocb->ki_pos > ip->i_size)
ip->i_size = iocb->ki_pos;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
if (ret == -ENOSPC &&
DM_EVENT_ENABLED(ip, DM_EVENT_NOSPACE) && !(ioflags & IO_INVIS)) {
xfs_iunlock(ip, iolock);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error = XFS_SEND_NAMESP(ip->i_mount, DM_EVENT_NOSPACE, ip,
DM_RIGHT_NULL, ip, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_ilock(ip, iolock);
if (error)
goto out_unlock_internal;
goto start;
}
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_DSYNC) || IS_SYNC(inode)) {
loff_t end = pos + ret - 1;
int error2;
xfs_iunlock(ip, iolock);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error2 = filemap_write_and_wait_range(mapping, pos, end);
if (!error)
error = error2;
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_ilock(ip, iolock);
error2 = -xfs_file_fsync(file, file->f_path.dentry,
(file->f_flags & __O_SYNC) ? 0 : 1);
if (!error)
error = error2;
}
out_unlock_internal:
if (ip->i_new_size) {
xfs_ilock(ip, XFS_ILOCK_EXCL);
ip->i_new_size = 0;
/*
* If this was a direct or synchronous I/O that failed (such
* as ENOSPC) then part of the I/O may have been written to
* disk before the error occured. In this case the on-disk
* file size may have been adjusted beyond the in-memory file
* size and now needs to be truncated back.
*/
if (ip->i_d.di_size > ip->i_size)
ip->i_d.di_size = ip->i_size;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
xfs_iunlock(ip, iolock);
out_unlock_mutex:
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
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;
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(
struct xfs_inode *xip,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int nsegs,
loff_t *offset,
int ioflags)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
unsigned long segs = nsegs;
xfs_mount_t *mp;
ssize_t ret = 0, error = 0;
xfs_fsize_t isize, new_size;
int iolock;
int eventsent = 0;
size_t ocount = 0, count;
loff_t pos;
int need_i_mutex;
XFS_STATS_INC(xs_write_calls);
error = generic_segment_checks(iovp, &segs, &ocount, VERIFY_READ);
if (error)
return error;
count = ocount;
pos = *offset;
if (count == 0)
return 0;
mp = xip->i_mount;
xfs_wait_for_freeze(mp, SB_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
relock:
if (ioflags & IO_ISDIRECT) {
iolock = XFS_IOLOCK_SHARED;
need_i_mutex = 0;
} else {
iolock = XFS_IOLOCK_EXCL;
need_i_mutex = 1;
mutex_lock(&inode->i_mutex);
}
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
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;
}
if ((DM_EVENT_ENABLED(xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
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, xip,
pos, count, dmflags, &iolock);
if (error) {
goto out_unlock_internal;
}
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) && pos != xip->i_size)
goto start;
}
if (ioflags & IO_ISDIRECT) {
xfs_buftarg_t *target =
XFS_IS_REALTIME_INODE(xip) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((pos & target->bt_smask) || (count & target->bt_smask)) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return XFS_ERROR(-EINVAL);
}
if (!need_i_mutex && (mapping->nrpages || pos > xip->i_size)) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
iolock = XFS_IOLOCK_EXCL;
need_i_mutex = 1;
mutex_lock(&inode->i_mutex);
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
goto start;
}
}
new_size = pos + count;
if (new_size > xip->i_size)
xip->i_new_size = new_size;
if (likely(!(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 (pos > xip->i_size) {
error = xfs_zero_eof(xip, pos, xip->i_size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL);
goto out_unlock_internal;
}
}
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 = -file_remove_suid(file);
if (unlikely(error)) {
goto out_unlock_internal;
}
}
/* We can write back this queue in page reclaim */
current->backing_dev_info = mapping->backing_dev_info;
if ((ioflags & IO_ISDIRECT)) {
if (mapping->nrpages) {
WARN_ON(need_i_mutex == 0);
xfs_inval_cached_trace(xip, pos, -1,
(pos & PAGE_CACHE_MASK), -1);
error = xfs_flushinval_pages(xip,
(pos & PAGE_CACHE_MASK),
-1, FI_REMAPF_LOCKED);
if (error)
goto out_unlock_internal;
}
if (need_i_mutex) {
/* demote the lock now the cached pages are gone */
xfs_ilock_demote(xip, XFS_IOLOCK_EXCL);
mutex_unlock(&inode->i_mutex);
iolock = XFS_IOLOCK_SHARED;
need_i_mutex = 0;
}
xfs_rw_enter_trace(XFS_DIOWR_ENTER, xip, (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;
ioflags &= ~IO_ISDIRECT;
xfs_iunlock(xip, iolock);
goto relock;
}
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER, xip, (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);
isize = i_size_read(inode);
if (unlikely(ret < 0 && ret != -EFAULT && *offset > isize))
*offset = isize;
if (*offset > xip->i_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_size)
xip->i_size = *offset;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
if (ret == -ENOSPC &&
DM_EVENT_ENABLED(xip, DM_EVENT_NOSPACE) && !(ioflags & IO_INVIS)) {
xfs_iunlock(xip, iolock);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, xip,
DM_RIGHT_NULL, xip, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_ilock(xip, iolock);
if (error)
goto out_unlock_internal;
goto start;
}
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)) {
int error2;
xfs_iunlock(xip, iolock);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error2 = sync_page_range(inode, mapping, pos, ret);
if (!error)
error = error2;
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_ilock(xip, iolock);
error2 = xfs_write_sync_logforce(mp, xip);
if (!error)
error = error2;
}
out_unlock_internal:
if (xip->i_new_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
xip->i_new_size = 0;
/*
* If this was a direct or synchronous I/O that failed (such
* as ENOSPC) then part of the I/O may have been written to
* disk before the error occured. In this case the on-disk
* file size may have been adjusted beyond the in-memory file
* size and now needs to be truncated back.
*/
if (xip->i_d.di_size > xip->i_size)
xip->i_d.di_size = xip->i_size;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
xfs_iunlock(xip, iolock);
out_unlock_mutex:
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
return -error;
}
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);
}