STATIC int
xfs_file_fsync(
struct file *file,
loff_t start,
loff_t end,
int datasync)
{
struct inode *inode = file->f_mapping->host;
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
int error = 0;
int log_flushed = 0;
xfs_lsn_t lsn = 0;
trace_xfs_file_fsync(ip);
error = filemap_write_and_wait_range(inode->i_mapping, start, end);
if (error)
return error;
if (XFS_FORCED_SHUTDOWN(mp))
return -XFS_ERROR(EIO);
xfs_iflags_clear(ip, XFS_ITRUNCATED);
if (mp->m_flags & XFS_MOUNT_BARRIER) {
/*
* If we have an RT and/or log subvolume we need to make sure
* to flush the write cache the device used for file data
* first. This is to ensure newly written file data make
* it to disk before logging the new inode size in case of
* an extending write.
*/
if (XFS_IS_REALTIME_INODE(ip))
xfs_blkdev_issue_flush(mp->m_rtdev_targp);
else if (mp->m_logdev_targp != mp->m_ddev_targp)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
}
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (xfs_ipincount(ip)) {
if (!datasync ||
(ip->i_itemp->ili_fields & ~XFS_ILOG_TIMESTAMP))
lsn = ip->i_itemp->ili_last_lsn;
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (lsn)
error = _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, &log_flushed);
if ((mp->m_flags & XFS_MOUNT_BARRIER) &&
mp->m_logdev_targp == mp->m_ddev_targp &&
!XFS_IS_REALTIME_INODE(ip) &&
!log_flushed)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
return -error;
}
STATIC int
xfs_fs_nfs_commit_metadata(
struct inode *inode)
{
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
xfs_lsn_t lsn = 0;
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (xfs_ipincount(ip))
lsn = ip->i_itemp->ili_last_lsn;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (!lsn)
return 0;
return _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, NULL);
}
STATIC int
xfs_fs_nfs_commit_metadata(
struct inode *inode)
{
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
int error = 0;
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (xfs_ipincount(ip)) {
error = _xfs_log_force_lsn(mp, ip->i_itemp->ili_last_lsn,
XFS_LOG_SYNC, NULL);
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return error;
}
/*
* Fsync operations on directories are much simpler than on regular files,
* as there is no file data to flush, and thus also no need for explicit
* cache flush operations, and there are no non-transaction metadata updates
* on directories either.
*/
STATIC int
xfs_dir_fsync(
struct file *file,
loff_t start,
loff_t end,
int datasync)
{
struct xfs_inode *ip = XFS_I(file->f_mapping->host);
struct xfs_mount *mp = ip->i_mount;
xfs_lsn_t lsn = 0;
trace_xfs_dir_fsync(ip);
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (xfs_ipincount(ip))
lsn = ip->i_itemp->ili_last_lsn;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (!lsn)
return 0;
return _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, NULL);
}
STATIC int
xfs_file_fsync(
struct file *file,
loff_t start,
loff_t end,
int datasync)
{
struct inode *inode = file->f_mapping->host;
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
int error = 0;
int log_flushed = 0;
xfs_lsn_t lsn = 0;
trace_xfs_file_fsync(ip);
error = file_write_and_wait_range(file, start, end);
if (error)
return error;
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
xfs_iflags_clear(ip, XFS_ITRUNCATED);
/*
* If we have an RT and/or log subvolume we need to make sure to flush
* the write cache the device used for file data first. This is to
* ensure newly written file data make it to disk before logging the new
* inode size in case of an extending write.
*/
if (XFS_IS_REALTIME_INODE(ip))
xfs_blkdev_issue_flush(mp->m_rtdev_targp);
else if (mp->m_logdev_targp != mp->m_ddev_targp)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
/*
* All metadata updates are logged, which means that we just have to
* flush the log up to the latest LSN that touched the inode. If we have
* concurrent fsync/fdatasync() calls, we need them to all block on the
* log force before we clear the ili_fsync_fields field. This ensures
* that we don't get a racing sync operation that does not wait for the
* metadata to hit the journal before returning. If we race with
* clearing the ili_fsync_fields, then all that will happen is the log
* force will do nothing as the lsn will already be on disk. We can't
* race with setting ili_fsync_fields because that is done under
* XFS_ILOCK_EXCL, and that can't happen because we hold the lock shared
* until after the ili_fsync_fields is cleared.
*/
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (xfs_ipincount(ip)) {
if (!datasync ||
(ip->i_itemp->ili_fsync_fields & ~XFS_ILOG_TIMESTAMP))
lsn = ip->i_itemp->ili_last_lsn;
}
if (lsn) {
error = _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, &log_flushed);
ip->i_itemp->ili_fsync_fields = 0;
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
/*
* If we only have a single device, and the log force about was
* a no-op we might have to flush the data device cache here.
* This can only happen for fdatasync/O_DSYNC if we were overwriting
* an already allocated file and thus do not have any metadata to
* commit.
*/
if (!log_flushed && !XFS_IS_REALTIME_INODE(ip) &&
mp->m_logdev_targp == mp->m_ddev_targp)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
return error;
}
STATIC int
xfs_file_fsync(
struct file *file,
loff_t start,
loff_t end,
int datasync)
{
struct inode *inode = file->f_mapping->host;
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
int error = 0;
int log_flushed = 0;
xfs_lsn_t lsn = 0;
trace_xfs_file_fsync(ip);
error = filemap_write_and_wait_range(inode->i_mapping, start, end);
if (error)
return error;
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
xfs_iflags_clear(ip, XFS_ITRUNCATED);
if (mp->m_flags & XFS_MOUNT_BARRIER) {
/*
* If we have an RT and/or log subvolume we need to make sure
* to flush the write cache the device used for file data
* first. This is to ensure newly written file data make
* it to disk before logging the new inode size in case of
* an extending write.
*/
if (XFS_IS_REALTIME_INODE(ip))
xfs_blkdev_issue_flush(mp->m_rtdev_targp);
else if (mp->m_logdev_targp != mp->m_ddev_targp)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
}
/*
* All metadata updates are logged, which means that we just have
* to flush the log up to the latest LSN that touched the inode.
*/
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (xfs_ipincount(ip)) {
if (!datasync ||
(ip->i_itemp->ili_fields & ~XFS_ILOG_TIMESTAMP))
lsn = ip->i_itemp->ili_last_lsn;
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (lsn)
error = _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, &log_flushed);
/*
* If we only have a single device, and the log force about was
* a no-op we might have to flush the data device cache here.
* This can only happen for fdatasync/O_DSYNC if we were overwriting
* an already allocated file and thus do not have any metadata to
* commit.
*/
if ((mp->m_flags & XFS_MOUNT_BARRIER) &&
mp->m_logdev_targp == mp->m_ddev_targp &&
!XFS_IS_REALTIME_INODE(ip) &&
!log_flushed)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
return error;
}
STATIC int
xfs_file_fsync(
struct file *file,
struct dentry *dentry,
int datasync)
{
struct xfs_inode *ip = XFS_I(dentry->d_inode);
struct xfs_trans *tp;
int error = 0;
int log_flushed = 0;
xfs_itrace_entry(ip);
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -XFS_ERROR(EIO);
xfs_iflags_clear(ip, XFS_ITRUNCATED);
/*
* We always need to make sure that the required inode state is safe on
* disk. The inode might be clean but we still might need to force the
* log because of committed transactions that haven't hit the disk yet.
* Likewise, there could be unflushed non-transactional changes to the
* inode core that have to go to disk and this requires us to issue
* a synchronous transaction to capture these changes correctly.
*
* This code relies on the assumption that if the i_update_core field
* of the inode is clear and the inode is unpinned then it is clean
* and no action is required.
*/
xfs_ilock(ip, XFS_ILOCK_SHARED);
/*
* First check if the VFS inode is marked dirty. All the dirtying
* of non-transactional updates no goes through mark_inode_dirty*,
* which allows us to distinguish beteeen pure timestamp updates
* and i_size updates which need to be caught for fdatasync.
* After that also theck for the dirty state in the XFS inode, which
* might gets cleared when the inode gets written out via the AIL
* or xfs_iflush_cluster.
*/
if (((dentry->d_inode->i_state & I_DIRTY_DATASYNC) ||
((dentry->d_inode->i_state & I_DIRTY_SYNC) && !datasync)) &&
ip->i_update_core) {
/*
* Kick off a transaction to log the inode core to get the
* updates. The sync transaction will also force the log.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED);
tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_FSYNC_TS);
error = xfs_trans_reserve(tp, 0,
XFS_FSYNC_TS_LOG_RES(ip->i_mount), 0, 0, 0);
if (error) {
xfs_trans_cancel(tp, 0);
return -error;
}
xfs_ilock(ip, XFS_ILOCK_EXCL);
/*
* Note - it's possible that we might have pushed ourselves out
* of the way during trans_reserve which would flush the inode.
* But there's no guarantee that the inode buffer has actually
* gone out yet (it's delwri). Plus the buffer could be pinned
* anyway if it's part of an inode in another recent
* transaction. So we play it safe and fire off the
* transaction anyway.
*/
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, ip);
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = _xfs_trans_commit(tp, 0, &log_flushed);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
} else {
/*
* Timestamps/size haven't changed since last inode flush or
* inode transaction commit. That means either nothing got
* written or a transaction committed which caught the updates.
* If the latter happened and the transaction hasn't hit the
* disk yet, the inode will be still be pinned. If it is,
* force the log.
*/
if (xfs_ipincount(ip)) {
error = _xfs_log_force_lsn(ip->i_mount,
ip->i_itemp->ili_last_lsn,
XFS_LOG_SYNC, &log_flushed);
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
}
if (ip->i_mount->m_flags & XFS_MOUNT_BARRIER) {
/*
* If the log write didn't issue an ordered tag we need
* to flush the disk cache for the data device now.
*/
if (!log_flushed)
xfs_blkdev_issue_flush(ip->i_mount->m_ddev_targp);
/*
* If this inode is on the RT dev we need to flush that
* cache as well.
*/
if (XFS_IS_REALTIME_INODE(ip))
xfs_blkdev_issue_flush(ip->i_mount->m_rtdev_targp);
}
return -error;
}
STATIC int
xfs_file_fsync(
struct file *file,
loff_t start,
loff_t end,
int datasync)
{
struct inode *inode = file->f_mapping->host;
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
struct xfs_trans *tp;
int error = 0;
int log_flushed = 0;
trace_xfs_file_fsync(ip);
error = filemap_write_and_wait_range(inode->i_mapping, start, end);
if (error)
return error;
if (XFS_FORCED_SHUTDOWN(mp))
return -XFS_ERROR(EIO);
xfs_iflags_clear(ip, XFS_ITRUNCATED);
xfs_ilock(ip, XFS_IOLOCK_SHARED);
xfs_ioend_wait(ip);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
if (mp->m_flags & XFS_MOUNT_BARRIER) {
/*
* If we have an RT and/or log subvolume we need to make sure
* to flush the write cache the device used for file data
* first. This is to ensure newly written file data make
* it to disk before logging the new inode size in case of
* an extending write.
*/
if (XFS_IS_REALTIME_INODE(ip))
xfs_blkdev_issue_flush(mp->m_rtdev_targp);
else if (mp->m_logdev_targp != mp->m_ddev_targp)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
}
/*
* We always need to make sure that the required inode state is safe on
* disk. The inode might be clean but we still might need to force the
* log because of committed transactions that haven't hit the disk yet.
* Likewise, there could be unflushed non-transactional changes to the
* inode core that have to go to disk and this requires us to issue
* a synchronous transaction to capture these changes correctly.
*
* This code relies on the assumption that if the i_update_core field
* of the inode is clear and the inode is unpinned then it is clean
* and no action is required.
*/
xfs_ilock(ip, XFS_ILOCK_SHARED);
/*
* First check if the VFS inode is marked dirty. All the dirtying
* of non-transactional updates no goes through mark_inode_dirty*,
* which allows us to distinguish beteeen pure timestamp updates
* and i_size updates which need to be caught for fdatasync.
* After that also theck for the dirty state in the XFS inode, which
* might gets cleared when the inode gets written out via the AIL
* or xfs_iflush_cluster.
*/
if (((inode->i_state & I_DIRTY_DATASYNC) ||
((inode->i_state & I_DIRTY_SYNC) && !datasync)) &&
ip->i_update_core) {
/*
* Kick off a transaction to log the inode core to get the
* updates. The sync transaction will also force the log.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED);
tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
error = xfs_trans_reserve(tp, 0,
XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0);
if (error) {
xfs_trans_cancel(tp, 0);
return -error;
}
xfs_ilock(ip, XFS_ILOCK_EXCL);
/*
* Note - it's possible that we might have pushed ourselves out
* of the way during trans_reserve which would flush the inode.
* But there's no guarantee that the inode buffer has actually
* gone out yet (it's delwri). Plus the buffer could be pinned
* anyway if it's part of an inode in another recent
* transaction. So we play it safe and fire off the
* transaction anyway.
*/
xfs_trans_ijoin(tp, ip);
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = _xfs_trans_commit(tp, 0, &log_flushed);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
} else {
/*
* Timestamps/size haven't changed since last inode flush or
* inode transaction commit. That means either nothing got
* written or a transaction committed which caught the updates.
* If the latter happened and the transaction hasn't hit the
* disk yet, the inode will be still be pinned. If it is,
* force the log.
*/
if (xfs_ipincount(ip)) {
error = _xfs_log_force_lsn(mp,
ip->i_itemp->ili_last_lsn,
XFS_LOG_SYNC, &log_flushed);
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
}
/*
* If we only have a single device, and the log force about was
* a no-op we might have to flush the data device cache here.
* This can only happen for fdatasync/O_DSYNC if we were overwriting
* an already allocated file and thus do not have any metadata to
* commit.
*/
if ((mp->m_flags & XFS_MOUNT_BARRIER) &&
mp->m_logdev_targp == mp->m_ddev_targp &&
!XFS_IS_REALTIME_INODE(ip) &&
!log_flushed)
xfs_blkdev_issue_flush(mp->m_ddev_targp);
return -error;
}
/*ARGSUSED*/
int
_xfs_trans_commit(
xfs_trans_t *tp,
uint flags,
int *log_flushed)
{
xfs_log_iovec_t *log_vector;
int nvec;
xfs_mount_t *mp;
xfs_lsn_t commit_lsn;
/* REFERENCED */
int error;
int log_flags;
int sync;
#define XFS_TRANS_LOGVEC_COUNT 16
xfs_log_iovec_t log_vector_fast[XFS_TRANS_LOGVEC_COUNT];
struct xlog_in_core *commit_iclog;
int shutdown;
commit_lsn = -1;
/*
* Determine whether this commit is releasing a permanent
* log reservation or not.
*/
if (flags & XFS_TRANS_RELEASE_LOG_RES) {
ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
log_flags = XFS_LOG_REL_PERM_RESERV;
} else {
log_flags = 0;
}
mp = tp->t_mountp;
/*
* If there is nothing to be logged by the transaction,
* then unlock all of the items associated with the
* transaction and free the transaction structure.
* Also make sure to return any reserved blocks to
* the free pool.
*/
shut_us_down:
shutdown = XFS_FORCED_SHUTDOWN(mp) ? EIO : 0;
if (!(tp->t_flags & XFS_TRANS_DIRTY) || shutdown) {
xfs_trans_unreserve_and_mod_sb(tp);
/*
* It is indeed possible for the transaction to be
* not dirty but the dqinfo portion to be. All that
* means is that we have some (non-persistent) quota
* reservations that need to be unreserved.
*/
xfs_trans_unreserve_and_mod_dquots(tp);
if (tp->t_ticket) {
commit_lsn = xfs_log_done(mp, tp->t_ticket,
NULL, log_flags);
if (commit_lsn == -1 && !shutdown)
shutdown = XFS_ERROR(EIO);
}
current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
xfs_trans_free_items(tp, shutdown? XFS_TRANS_ABORT : 0);
xfs_trans_free_busy(tp);
xfs_trans_free(tp);
XFS_STATS_INC(xs_trans_empty);
return (shutdown);
}
ASSERT(tp->t_ticket != NULL);
/*
* If we need to update the superblock, then do it now.
*/
if (tp->t_flags & XFS_TRANS_SB_DIRTY)
xfs_trans_apply_sb_deltas(tp);
xfs_trans_apply_dquot_deltas(tp);
/*
* Ask each log item how many log_vector entries it will
* need so we can figure out how many to allocate.
* Try to avoid the kmem_alloc() call in the common case
* by using a vector from the stack when it fits.
*/
nvec = xfs_trans_count_vecs(tp);
if (nvec == 0) {
xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
goto shut_us_down;
} else if (nvec <= XFS_TRANS_LOGVEC_COUNT) {
log_vector = log_vector_fast;
} else {
log_vector = (xfs_log_iovec_t *)kmem_alloc(nvec *
sizeof(xfs_log_iovec_t),
KM_SLEEP);
}
/*
* Fill in the log_vector and pin the logged items, and
* then write the transaction to the log.
*/
xfs_trans_fill_vecs(tp, log_vector);
error = xfs_log_write(mp, log_vector, nvec, tp->t_ticket, &(tp->t_lsn));
/*
* The transaction is committed incore here, and can go out to disk
* at any time after this call. However, all the items associated
* with the transaction are still locked and pinned in memory.
*/
commit_lsn = xfs_log_done(mp, tp->t_ticket, &commit_iclog, log_flags);
tp->t_commit_lsn = commit_lsn;
if (nvec > XFS_TRANS_LOGVEC_COUNT) {
kmem_free(log_vector);
}
/*
* If we got a log write error. Unpin the logitems that we
* had pinned, clean up, free trans structure, and return error.
*/
if (error || commit_lsn == -1) {
current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
xfs_trans_uncommit(tp, flags|XFS_TRANS_ABORT);
return XFS_ERROR(EIO);
}
/*
* Once the transaction has committed, unused
* reservations need to be released and changes to
* the superblock need to be reflected in the in-core
* version. Do that now.
*/
xfs_trans_unreserve_and_mod_sb(tp);
sync = tp->t_flags & XFS_TRANS_SYNC;
/*
* Tell the LM to call the transaction completion routine
* when the log write with LSN commit_lsn completes (e.g.
* when the transaction commit really hits the on-disk log).
* After this call we cannot reference tp, because the call
* can happen at any time and the call will free the transaction
* structure pointed to by tp. The only case where we call
* the completion routine (xfs_trans_committed) directly is
* if the log is turned off on a debug kernel or we're
* running in simulation mode (the log is explicitly turned
* off).
*/
tp->t_logcb.cb_func = (void(*)(void*, int))xfs_trans_committed;
tp->t_logcb.cb_arg = tp;
/*
* We need to pass the iclog buffer which was used for the
* transaction commit record into this function, and attach
* the callback to it. The callback must be attached before
* the items are unlocked to avoid racing with other threads
* waiting for an item to unlock.
*/
shutdown = xfs_log_notify(mp, commit_iclog, &(tp->t_logcb));
/*
* Mark this thread as no longer being in a transaction
*/
current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
/*
* Once all the items of the transaction have been copied
* to the in core log and the callback is attached, the
* items can be unlocked.
*
* This will free descriptors pointing to items which were
* not logged since there is nothing more to do with them.
* For items which were logged, we will keep pointers to them
* so they can be unpinned after the transaction commits to disk.
* This will also stamp each modified meta-data item with
* the commit lsn of this transaction for dependency tracking
* purposes.
*/
xfs_trans_unlock_items(tp, commit_lsn);
/*
* If we detected a log error earlier, finish committing
* the transaction now (unpin log items, etc).
*
* Order is critical here, to avoid using the transaction
* pointer after its been freed (by xfs_trans_committed
* either here now, or as a callback). We cannot do this
* step inside xfs_log_notify as was done earlier because
* of this issue.
*/
if (shutdown)
xfs_trans_committed(tp, XFS_LI_ABORTED);
/*
* Now that the xfs_trans_committed callback has been attached,
* and the items are released we can finally allow the iclog to
* go to disk.
*/
error = xfs_log_release_iclog(mp, commit_iclog);
/*
* If the transaction needs to be synchronous, then force the
* log out now and wait for it.
*/
if (sync) {
if (!error) {
error = _xfs_log_force_lsn(mp, commit_lsn,
XFS_LOG_SYNC, log_flushed);
}
XFS_STATS_INC(xs_trans_sync);
} else {
XFS_STATS_INC(xs_trans_async);
}
return (error);
}
