/*
* This is called to attempt to lock the inode associated with this
* inode log item, in preparation for the push routine which does the actual
* iflush. Don't sleep on the inode lock or the flush lock.
*
* If the flush lock is already held, indicating that the inode has
* been or is in the process of being flushed, then (ideally) we'd like to
* see if the inode's buffer is still incore, and if so give it a nudge.
* We delay doing so until the pushbuf routine, though, to avoid holding
* the AIL lock across a call to the blackhole which is the buffer cache.
* Also we don't want to sleep in any device strategy routines, which can happen
* if we do the subsequent bawrite in here.
*/
STATIC uint
xfs_inode_item_trylock(
xfs_inode_log_item_t *iip)
{
register xfs_inode_t *ip;
ip = iip->ili_inode;
if (xfs_ipincount(ip) > 0) {
return XFS_ITEM_PINNED;
}
if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) {
return XFS_ITEM_LOCKED;
}
if (!xfs_iflock_nowait(ip)) {
/*
* If someone else isn't already trying to push the inode
* buffer, we get to do it.
*/
if (iip->ili_pushbuf_flag == 0) {
iip->ili_pushbuf_flag = 1;
#ifdef DEBUG
iip->ili_push_owner = current_pid();
#endif
/*
* Inode is left locked in shared mode.
* Pushbuf routine gets to unlock it.
*/
return XFS_ITEM_PUSHBUF;
} else {
/*
* We hold the AIL lock, so we must specify the
* NONOTIFY flag so that we won't double trip.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED|XFS_IUNLOCK_NONOTIFY);
return XFS_ITEM_FLUSHING;
}
/* NOTREACHED */
}
/* Stale items should force out the iclog */
if (ip->i_flags & XFS_ISTALE) {
xfs_ifunlock(ip);
xfs_iunlock(ip, XFS_ILOCK_SHARED|XFS_IUNLOCK_NONOTIFY);
return XFS_ITEM_PINNED;
}
#ifdef DEBUG
if (!XFS_FORCED_SHUTDOWN(ip->i_mount)) {
ASSERT(iip->ili_format.ilf_fields != 0);
ASSERT(iip->ili_logged == 0);
ASSERT(iip->ili_item.li_flags & XFS_LI_IN_AIL);
}
#endif
return XFS_ITEM_SUCCESS;
}
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;
}
/*
* This is called to attempt to lock the inode associated with this
* inode log item, in preparation for the push routine which does the actual
* iflush. Don't sleep on the inode lock or the flush lock.
*
* If the flush lock is already held, indicating that the inode has
* been or is in the process of being flushed, then (ideally) we'd like to
* see if the inode's buffer is still incore, and if so give it a nudge.
* We delay doing so until the pushbuf routine, though, to avoid holding
* the AIL lock across a call to the blackhole which is the buffer cache.
* Also we don't want to sleep in any device strategy routines, which can happen
* if we do the subsequent bawrite in here.
*/
STATIC uint
xfs_inode_item_trylock(
struct xfs_log_item *lip)
{
struct xfs_inode_log_item *iip = INODE_ITEM(lip);
struct xfs_inode *ip = iip->ili_inode;
if (xfs_ipincount(ip) > 0)
return XFS_ITEM_PINNED;
if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
return XFS_ITEM_LOCKED;
if (!xfs_iflock_nowait(ip)) {
/*
* inode has already been flushed to the backing buffer,
* leave it locked in shared mode, pushbuf routine will
* unlock it.
*/
return XFS_ITEM_PUSHBUF;
}
/* Stale items should force out the iclog */
if (ip->i_flags & XFS_ISTALE) {
xfs_ifunlock(ip);
/*
* we hold the AIL lock - notify the unlock routine of this
* so it doesn't try to get the lock again.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED|XFS_IUNLOCK_NONOTIFY);
return XFS_ITEM_PINNED;
}
#ifdef DEBUG
if (!XFS_FORCED_SHUTDOWN(ip->i_mount)) {
ASSERT(iip->ili_format.ilf_fields != 0);
ASSERT(iip->ili_logged == 0);
ASSERT(lip->li_flags & XFS_LI_IN_AIL);
}
#endif
return XFS_ITEM_SUCCESS;
}
/*
* 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 uint
xfs_inode_item_trylock(
struct xfs_log_item *lip)
{
struct xfs_inode_log_item *iip = INODE_ITEM(lip);
struct xfs_inode *ip = iip->ili_inode;
if (xfs_ipincount(ip) > 0)
return XFS_ITEM_PINNED;
if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
return XFS_ITEM_LOCKED;
if (!xfs_iflock_nowait(ip)) {
/*
*/
return XFS_ITEM_PUSHBUF;
}
/* */
if (ip->i_flags & XFS_ISTALE) {
xfs_ifunlock(ip);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return XFS_ITEM_PINNED;
}
#ifdef DEBUG
if (!XFS_FORCED_SHUTDOWN(ip->i_mount)) {
ASSERT(iip->ili_fields != 0);
ASSERT(iip->ili_logged == 0);
ASSERT(lip->li_flags & XFS_LI_IN_AIL);
}
#endif
return XFS_ITEM_SUCCESS;
}
/*
* Inodes in different states need to be treated differently. The following
* table lists the inode states and the reclaim actions necessary:
*
* inode state iflush ret required action
* --------------- ---------- ---------------
* bad - reclaim
* shutdown EIO unpin and reclaim
* clean, unpinned 0 reclaim
* stale, unpinned 0 reclaim
* clean, pinned(*) 0 requeue
* stale, pinned EAGAIN requeue
* dirty, async - requeue
* dirty, sync 0 reclaim
*
* (*) dgc: I don't think the clean, pinned state is possible but it gets
* handled anyway given the order of checks implemented.
*
* Also, because we get the flush lock first, we know that any inode that has
* been flushed delwri has had the flush completed by the time we check that
* the inode is clean.
*
* Note that because the inode is flushed delayed write by AIL pushing, the
* flush lock may already be held here and waiting on it can result in very
* long latencies. Hence for sync reclaims, where we wait on the flush lock,
* the caller should push the AIL first before trying to reclaim inodes to
* minimise the amount of time spent waiting. For background relaim, we only
* bother to reclaim clean inodes anyway.
*
* Hence the order of actions after gaining the locks should be:
* bad => reclaim
* shutdown => unpin and reclaim
* pinned, async => requeue
* pinned, sync => unpin
* stale => reclaim
* clean => reclaim
* dirty, async => requeue
* dirty, sync => flush, wait and reclaim
*/
STATIC int
xfs_reclaim_inode(
struct xfs_inode *ip,
struct xfs_perag *pag,
int sync_mode)
{
struct xfs_buf *bp = NULL;
xfs_ino_t ino = ip->i_ino; /* for radix_tree_delete */
int error;
restart:
error = 0;
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (!xfs_iflock_nowait(ip)) {
if (!(sync_mode & SYNC_WAIT))
goto out;
xfs_iflock(ip);
}
if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
xfs_iunpin_wait(ip);
/* xfs_iflush_abort() drops the flush lock */
xfs_iflush_abort(ip, false);
goto reclaim;
}
if (xfs_ipincount(ip)) {
if (!(sync_mode & SYNC_WAIT))
goto out_ifunlock;
xfs_iunpin_wait(ip);
}
if (xfs_iflags_test(ip, XFS_ISTALE) || xfs_inode_clean(ip)) {
xfs_ifunlock(ip);
goto reclaim;
}
/*
* Never flush out dirty data during non-blocking reclaim, as it would
* just contend with AIL pushing trying to do the same job.
*/
if (!(sync_mode & SYNC_WAIT))
goto out_ifunlock;
/*
* Now we have an inode that needs flushing.
*
* Note that xfs_iflush will never block on the inode buffer lock, as
* xfs_ifree_cluster() can lock the inode buffer before it locks the
* ip->i_lock, and we are doing the exact opposite here. As a result,
* doing a blocking xfs_imap_to_bp() to get the cluster buffer would
* result in an ABBA deadlock with xfs_ifree_cluster().
*
* As xfs_ifree_cluser() must gather all inodes that are active in the
* cache to mark them stale, if we hit this case we don't actually want
* to do IO here - we want the inode marked stale so we can simply
* reclaim it. Hence if we get an EAGAIN error here, just unlock the
* inode, back off and try again. Hopefully the next pass through will
* see the stale flag set on the inode.
*/
error = xfs_iflush(ip, &bp);
if (error == -EAGAIN) {
xfs_iunlock(ip, XFS_ILOCK_EXCL);
/* backoff longer than in xfs_ifree_cluster */
delay(2);
goto restart;
}
if (!error) {
error = xfs_bwrite(bp);
xfs_buf_relse(bp);
}
reclaim:
ASSERT(!xfs_isiflocked(ip));
/*
* Because we use RCU freeing we need to ensure the inode always appears
* to be reclaimed with an invalid inode number when in the free state.
* We do this as early as possible under the ILOCK so that
* xfs_iflush_cluster() can be guaranteed to detect races with us here.
* By doing this, we guarantee that once xfs_iflush_cluster has locked
* XFS_ILOCK that it will see either a valid, flushable inode that will
* serialise correctly, or it will see a clean (and invalid) inode that
* it can skip.
*/
spin_lock(&ip->i_flags_lock);
ip->i_flags = XFS_IRECLAIM;
ip->i_ino = 0;
spin_unlock(&ip->i_flags_lock);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
XFS_STATS_INC(ip->i_mount, xs_ig_reclaims);
/*
* Remove the inode from the per-AG radix tree.
*
* Because radix_tree_delete won't complain even if the item was never
* added to the tree assert that it's been there before to catch
* problems with the inode life time early on.
*/
spin_lock(&pag->pag_ici_lock);
if (!radix_tree_delete(&pag->pag_ici_root,
XFS_INO_TO_AGINO(ip->i_mount, ino)))
ASSERT(0);
xfs_perag_clear_reclaim_tag(pag);
spin_unlock(&pag->pag_ici_lock);
/*
* Here we do an (almost) spurious inode lock in order to coordinate
* with inode cache radix tree lookups. This is because the lookup
* can reference the inodes in the cache without taking references.
*
* We make that OK here by ensuring that we wait until the inode is
* unlocked after the lookup before we go ahead and free it.
*/
xfs_ilock(ip, XFS_ILOCK_EXCL);
xfs_qm_dqdetach(ip);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
__xfs_inode_free(ip);
return error;
out_ifunlock:
xfs_ifunlock(ip);
out:
xfs_iflags_clear(ip, XFS_IRECLAIM);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
/*
* We could return -EAGAIN here to make reclaim rescan the inode tree in
* a short while. However, this just burns CPU time scanning the tree
* waiting for IO to complete and the reclaim work never goes back to
* the idle state. Instead, return 0 to let the next scheduled
* background reclaim attempt to reclaim the inode again.
*/
return 0;
}
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;
}
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 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);
}
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;
}
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;
}
/*
* xfs sync routine for internal use
*
* This routine supports all of the flags defined for the generic VFS_SYNC
* interface as explained above under xfs_sync. In the interests of not
* changing interfaces within the 6.5 family, additional internallly-
* required functions are specified within a separate xflags parameter,
* only available by calling this routine.
*
*/
STATIC int
xfs_sync_inodes(
xfs_mount_t *mp,
int flags,
int xflags,
int *bypassed)
{
xfs_inode_t *ip = NULL;
xfs_inode_t *ip_next;
xfs_buf_t *bp;
vnode_t *vp = NULL;
vmap_t vmap;
int error;
int last_error;
uint64_t fflag;
uint lock_flags;
uint base_lock_flags;
boolean_t mount_locked;
boolean_t vnode_refed;
int preempt;
xfs_dinode_t *dip;
xfs_iptr_t *ipointer;
#ifdef DEBUG
boolean_t ipointer_in = B_FALSE;
#define IPOINTER_SET ipointer_in = B_TRUE
#define IPOINTER_CLR ipointer_in = B_FALSE
#else
#define IPOINTER_SET
#define IPOINTER_CLR
#endif
/* Insert a marker record into the inode list after inode ip. The list
* must be locked when this is called. After the call the list will no
* longer be locked.
*/
#define IPOINTER_INSERT(ip, mp) { \
ASSERT(ipointer_in == B_FALSE); \
ipointer->ip_mnext = ip->i_mnext; \
ipointer->ip_mprev = ip; \
ip->i_mnext = (xfs_inode_t *)ipointer; \
ipointer->ip_mnext->i_mprev = (xfs_inode_t *)ipointer; \
preempt = 0; \
XFS_MOUNT_IUNLOCK(mp); \
mount_locked = B_FALSE; \
IPOINTER_SET; \
}
/* Remove the marker from the inode list. If the marker was the only item
* in the list then there are no remaining inodes and we should zero out
* the whole list. If we are the current head of the list then move the head
* past us.
*/
#define IPOINTER_REMOVE(ip, mp) { \
ASSERT(ipointer_in == B_TRUE); \
if (ipointer->ip_mnext != (xfs_inode_t *)ipointer) { \
ip = ipointer->ip_mnext; \
ip->i_mprev = ipointer->ip_mprev; \
ipointer->ip_mprev->i_mnext = ip; \
if (mp->m_inodes == (xfs_inode_t *)ipointer) { \
mp->m_inodes = ip; \
} \
} else { \
ASSERT(mp->m_inodes == (xfs_inode_t *)ipointer); \
mp->m_inodes = NULL; \
ip = NULL; \
} \
IPOINTER_CLR; \
}
#define XFS_PREEMPT_MASK 0x7f
if (bypassed)
*bypassed = 0;
if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
return 0;
error = 0;
last_error = 0;
preempt = 0;
/* Allocate a reference marker */
ipointer = (xfs_iptr_t *)kmem_zalloc(sizeof(xfs_iptr_t), KM_SLEEP);
fflag = XFS_B_ASYNC; /* default is don't wait */
if (flags & SYNC_BDFLUSH)
fflag = XFS_B_DELWRI;
if (flags & SYNC_WAIT)
fflag = 0; /* synchronous overrides all */
base_lock_flags = XFS_ILOCK_SHARED;
if (flags & (SYNC_DELWRI | SYNC_CLOSE)) {
/*
* We need the I/O lock if we're going to call any of
* the flush/inval routines.
*/
base_lock_flags |= XFS_IOLOCK_SHARED;
}
XFS_MOUNT_ILOCK(mp);
ip = mp->m_inodes;
mount_locked = B_TRUE;
vnode_refed = B_FALSE;
IPOINTER_CLR;
do {
ASSERT(ipointer_in == B_FALSE);
ASSERT(vnode_refed == B_FALSE);
lock_flags = base_lock_flags;
/*
* There were no inodes in the list, just break out
* of the loop.
*/
if (ip == NULL) {
break;
}
/*
* We found another sync thread marker - skip it
*/
if (ip->i_mount == NULL) {
ip = ip->i_mnext;
continue;
}
vp = XFS_ITOV_NULL(ip);
/*
* If the vnode is gone then this is being torn down,
* call reclaim if it is flushed, else let regular flush
* code deal with it later in the loop.
*/
if (vp == NULL) {
/* Skip ones already in reclaim */
if (ip->i_flags & XFS_IRECLAIM) {
ip = ip->i_mnext;
continue;
}
if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
ip = ip->i_mnext;
} else if ((xfs_ipincount(ip) == 0) &&
xfs_iflock_nowait(ip)) {
IPOINTER_INSERT(ip, mp);
xfs_finish_reclaim(ip, 1,
XFS_IFLUSH_DELWRI_ELSE_ASYNC);
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
IPOINTER_REMOVE(ip, mp);
} else {
xfs_iunlock(ip, XFS_ILOCK_EXCL);
ip = ip->i_mnext;
}
continue;
}
if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
XFS_MOUNT_IUNLOCK(mp);
kmem_free(ipointer, sizeof(xfs_iptr_t));
return 0;
}
/*
* If this is just vfs_sync() or pflushd() calling
* then we can skip inodes for which it looks like
* there is nothing to do. Since we don't have the
* inode locked this is racey, but these are periodic
* calls so it doesn't matter. For the others we want
* to know for sure, so we at least try to lock them.
*/
if (flags & SYNC_BDFLUSH) {
if (((ip->i_itemp == NULL) ||
!(ip->i_itemp->ili_format.ilf_fields &
XFS_ILOG_ALL)) &&
(ip->i_update_core == 0)) {
ip = ip->i_mnext;
continue;
}
}
/*
* Try to lock without sleeping. We're out of order with
* the inode list lock here, so if we fail we need to drop
* the mount lock and try again. If we're called from
* bdflush() here, then don't bother.
*
* The inode lock here actually coordinates with the
* almost spurious inode lock in xfs_ireclaim() to prevent
* the vnode we handle here without a reference from
* being freed while we reference it. If we lock the inode
* while it's on the mount list here, then the spurious inode
* lock in xfs_ireclaim() after the inode is pulled from
* the mount list will sleep until we release it here.
* This keeps the vnode from being freed while we reference
* it. It is also cheaper and simpler than actually doing
* a vn_get() for every inode we touch here.
*/
if (xfs_ilock_nowait(ip, lock_flags) == 0) {
if ((flags & SYNC_BDFLUSH) || (vp == NULL)) {
ip = ip->i_mnext;
continue;
}
/*
* We need to unlock the inode list lock in order
* to lock the inode. Insert a marker record into
* the inode list to remember our position, dropping
* the lock is now done inside the IPOINTER_INSERT
* macro.
*
* We also use the inode list lock to protect us
* in taking a snapshot of the vnode version number
* for use in calling vn_get().
*/
VMAP(vp, vmap);
IPOINTER_INSERT(ip, mp);
vp = vn_get(vp, &vmap);
if (vp == NULL) {
/*
* The vnode was reclaimed once we let go
* of the inode list lock. Skip to the
* next list entry. Remove the marker.
*/
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
vnode_refed = B_FALSE;
IPOINTER_REMOVE(ip, mp);
continue;
}
xfs_ilock(ip, lock_flags);
ASSERT(vp == XFS_ITOV(ip));
ASSERT(ip->i_mount == mp);
vnode_refed = B_TRUE;
}
/* From here on in the loop we may have a marker record
* in the inode list.
*/
if ((flags & SYNC_CLOSE) && (vp != NULL)) {
/*
* This is the shutdown case. We just need to
* flush and invalidate all the pages associated
* with the inode. Drop the inode lock since
* we can't hold it across calls to the buffer
* cache.
*
* We don't set the VREMAPPING bit in the vnode
* here, because we don't hold the vnode lock
* exclusively. It doesn't really matter, though,
* because we only come here when we're shutting
* down anyway.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (XFS_FORCED_SHUTDOWN(mp)) {
VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
} else {
VOP_FLUSHINVAL_PAGES(vp, 0, -1, FI_REMAPF);
}
xfs_ilock(ip, XFS_ILOCK_SHARED);
} else if ((flags & SYNC_DELWRI) && (vp != NULL)) {
if (VN_DIRTY(vp)) {
/* We need to have dropped the lock here,
* so insert a marker if we have not already
* done so.
*/
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
/*
* Drop the inode lock since we can't hold it
* across calls to the buffer cache.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED);
VOP_FLUSH_PAGES(vp, (xfs_off_t)0, -1,
fflag, FI_NONE, error);
xfs_ilock(ip, XFS_ILOCK_SHARED);
}
}
if (flags & SYNC_BDFLUSH) {
if ((flags & SYNC_ATTR) &&
((ip->i_update_core) ||
((ip->i_itemp != NULL) &&
(ip->i_itemp->ili_format.ilf_fields != 0)))) {
/* Insert marker and drop lock if not already
* done.
*/
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
/*
* We don't want the periodic flushing of the
* inodes by vfs_sync() to interfere with
* I/O to the file, especially read I/O
* where it is only the access time stamp
* that is being flushed out. To prevent
* long periods where we have both inode
* locks held shared here while reading the
* inode's buffer in from disk, we drop the
* inode lock while reading in the inode
* buffer. We have to release the buffer
* and reacquire the inode lock so that they
* are acquired in the proper order (inode
* locks first). The buffer will go at the
* end of the lru chain, though, so we can
* expect it to still be there when we go
* for it again in xfs_iflush().
*/
if ((xfs_ipincount(ip) == 0) &&
xfs_iflock_nowait(ip)) {
xfs_ifunlock(ip);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
error = xfs_itobp(mp, NULL, ip,
&dip, &bp, 0);
if (!error) {
xfs_buf_relse(bp);
} else {
/* Bailing out, remove the
* marker and free it.
*/
XFS_MOUNT_ILOCK(mp);
IPOINTER_REMOVE(ip, mp);
XFS_MOUNT_IUNLOCK(mp);
ASSERT(!(lock_flags &
XFS_IOLOCK_SHARED));
kmem_free(ipointer,
sizeof(xfs_iptr_t));
return (0);
}
/*
* Since we dropped the inode lock,
* the inode may have been reclaimed.
* Therefore, we reacquire the mount
* lock and check to see if we were the
* inode reclaimed. If this happened
* then the ipointer marker will no
* longer point back at us. In this
* case, move ip along to the inode
* after the marker, remove the marker
* and continue.
*/
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
if (ip != ipointer->ip_mprev) {
IPOINTER_REMOVE(ip, mp);
ASSERT(!vnode_refed);
ASSERT(!(lock_flags &
XFS_IOLOCK_SHARED));
continue;
}
ASSERT(ip->i_mount == mp);
if (xfs_ilock_nowait(ip,
XFS_ILOCK_SHARED) == 0) {
ASSERT(ip->i_mount == mp);
/*
* We failed to reacquire
* the inode lock without
* sleeping, so just skip
* the inode for now. We
* clear the ILOCK bit from
* the lock_flags so that we
* won't try to drop a lock
* we don't hold below.
*/
lock_flags &= ~XFS_ILOCK_SHARED;
IPOINTER_REMOVE(ip_next, mp);
} else if ((xfs_ipincount(ip) == 0) &&
xfs_iflock_nowait(ip)) {
ASSERT(ip->i_mount == mp);
/*
* Since this is vfs_sync()
* calling we only flush the
* inode out if we can lock
* it without sleeping and
* it is not pinned. Drop
* the mount lock here so
* that we don't hold it for
* too long. We already have
* a marker in the list here.
*/
XFS_MOUNT_IUNLOCK(mp);
mount_locked = B_FALSE;
error = xfs_iflush(ip,
XFS_IFLUSH_DELWRI);
} else {
ASSERT(ip->i_mount == mp);
IPOINTER_REMOVE(ip_next, mp);
}
}
}
} else {
if ((flags & SYNC_ATTR) &&
((ip->i_update_core) ||
((ip->i_itemp != NULL) &&
(ip->i_itemp->ili_format.ilf_fields != 0)))) {
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
if (flags & SYNC_WAIT) {
xfs_iflock(ip);
error = xfs_iflush(ip,
XFS_IFLUSH_SYNC);
} else {
/*
* If we can't acquire the flush
* lock, then the inode is already
* being flushed so don't bother
* waiting. If we can lock it then
* do a delwri flush so we can
* combine multiple inode flushes
* in each disk write.
*/
if (xfs_iflock_nowait(ip)) {
error = xfs_iflush(ip,
XFS_IFLUSH_DELWRI);
}
else if (bypassed)
(*bypassed)++;
}
}
}
if (lock_flags != 0) {
xfs_iunlock(ip, lock_flags);
}
if (vnode_refed) {
/*
* If we had to take a reference on the vnode
* above, then wait until after we've unlocked
* the inode to release the reference. This is
* because we can be already holding the inode
* lock when VN_RELE() calls xfs_inactive().
*
* Make sure to drop the mount lock before calling
* VN_RELE() so that we don't trip over ourselves if
* we have to go for the mount lock again in the
* inactive code.
*/
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
VN_RELE(vp);
vnode_refed = B_FALSE;
}
if (error) {
last_error = error;
}
/*
* bail out if the filesystem is corrupted.
*/
if (error == EFSCORRUPTED) {
if (!mount_locked) {
XFS_MOUNT_ILOCK(mp);
IPOINTER_REMOVE(ip, mp);
}
XFS_MOUNT_IUNLOCK(mp);
ASSERT(ipointer_in == B_FALSE);
kmem_free(ipointer, sizeof(xfs_iptr_t));
return XFS_ERROR(error);
}
/* Let other threads have a chance at the mount lock
* if we have looped many times without dropping the
* lock.
*/
if ((++preempt & XFS_PREEMPT_MASK) == 0) {
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
}
if (mount_locked == B_FALSE) {
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
IPOINTER_REMOVE(ip, mp);
continue;
}
ASSERT(ipointer_in == B_FALSE);
ip = ip->i_mnext;
} while (ip != mp->m_inodes);
XFS_MOUNT_IUNLOCK(mp);
ASSERT(ipointer_in == B_FALSE);
kmem_free(ipointer, sizeof(xfs_iptr_t));
return XFS_ERROR(last_error);
}
STATIC int
xfs_reclaim_inode(
struct xfs_inode *ip,
struct xfs_perag *pag,
int sync_mode)
{
int error;
restart:
error = 0;
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (!xfs_iflock_nowait(ip)) {
if (!(sync_mode & SYNC_WAIT))
goto out;
xfs_promote_inode(ip);
xfs_iflock(ip);
}
if (is_bad_inode(VFS_I(ip)))
goto reclaim;
if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
xfs_iunpin_wait(ip);
goto reclaim;
}
if (xfs_ipincount(ip)) {
if (!(sync_mode & SYNC_WAIT)) {
xfs_ifunlock(ip);
goto out;
}
xfs_iunpin_wait(ip);
}
if (xfs_iflags_test(ip, XFS_ISTALE))
goto reclaim;
if (xfs_inode_clean(ip))
goto reclaim;
error = xfs_iflush(ip, SYNC_TRYLOCK | sync_mode);
if (sync_mode & SYNC_WAIT) {
if (error == EAGAIN) {
xfs_iunlock(ip, XFS_ILOCK_EXCL);
delay(2);
goto restart;
}
xfs_iflock(ip);
goto reclaim;
}
if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(ip->i_mount)) {
xfs_warn(ip->i_mount,
"inode 0x%llx background reclaim flush failed with %d",
(long long)ip->i_ino, error);
}
out:
xfs_iflags_clear(ip, XFS_IRECLAIM);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
return 0;
reclaim:
xfs_ifunlock(ip);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
XFS_STATS_INC(xs_ig_reclaims);
spin_lock(&pag->pag_ici_lock);
if (!radix_tree_delete(&pag->pag_ici_root,
XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
ASSERT(0);
__xfs_inode_clear_reclaim(pag, ip);
spin_unlock(&pag->pag_ici_lock);
xfs_ilock(ip, XFS_ILOCK_EXCL);
xfs_qm_dqdetach(ip);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs_inode_free(ip);
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;
}
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;
}
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;
}
/*
* Inodes in different states need to be treated differently, and the return
* value of xfs_iflush is not sufficient to get this right. The following table
* lists the inode states and the reclaim actions necessary for non-blocking
* reclaim:
*
*
* inode state iflush ret required action
* --------------- ---------- ---------------
* bad - reclaim
* shutdown EIO unpin and reclaim
* clean, unpinned 0 reclaim
* stale, unpinned 0 reclaim
* clean, pinned(*) 0 requeue
* stale, pinned EAGAIN requeue
* dirty, delwri ok 0 requeue
* dirty, delwri blocked EAGAIN requeue
* dirty, sync flush 0 reclaim
*
* (*) dgc: I don't think the clean, pinned state is possible but it gets
* handled anyway given the order of checks implemented.
*
* As can be seen from the table, the return value of xfs_iflush() is not
* sufficient to correctly decide the reclaim action here. The checks in
* xfs_iflush() might look like duplicates, but they are not.
*
* Also, because we get the flush lock first, we know that any inode that has
* been flushed delwri has had the flush completed by the time we check that
* the inode is clean. The clean inode check needs to be done before flushing
* the inode delwri otherwise we would loop forever requeuing clean inodes as
* we cannot tell apart a successful delwri flush and a clean inode from the
* return value of xfs_iflush().
*
* Note that because the inode is flushed delayed write by background
* writeback, the flush lock may already be held here and waiting on it can
* result in very long latencies. Hence for sync reclaims, where we wait on the
* flush lock, the caller should push out delayed write inodes first before
* trying to reclaim them to minimise the amount of time spent waiting. For
* background relaim, we just requeue the inode for the next pass.
*
* Hence the order of actions after gaining the locks should be:
* bad => reclaim
* shutdown => unpin and reclaim
* pinned, delwri => requeue
* pinned, sync => unpin
* stale => reclaim
* clean => reclaim
* dirty, delwri => flush and requeue
* dirty, sync => flush, wait and reclaim
*/
STATIC int
xfs_reclaim_inode(
struct xfs_inode *ip,
struct xfs_perag *pag,
int sync_mode)
{
int error;
restart:
error = 0;
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (!xfs_iflock_nowait(ip)) {
if (!(sync_mode & SYNC_WAIT))
goto out;
/*
* If we only have a single dirty inode in a cluster there is
* a fair chance that the AIL push may have pushed it into
* the buffer, but xfsbufd won't touch it until 30 seconds
* from now, and thus we will lock up here.
*
* Promote the inode buffer to the front of the delwri list
* and wake up xfsbufd now.
*/
xfs_promote_inode(ip);
xfs_iflock(ip);
}
if (is_bad_inode(VFS_I(ip)))
goto reclaim;
if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
xfs_iunpin_wait(ip);
goto reclaim;
}
if (xfs_ipincount(ip)) {
if (!(sync_mode & SYNC_WAIT)) {
xfs_ifunlock(ip);
goto out;
}
xfs_iunpin_wait(ip);
}
if (xfs_iflags_test(ip, XFS_ISTALE))
goto reclaim;
if (xfs_inode_clean(ip))
goto reclaim;
/*
* Now we have an inode that needs flushing.
*
* We do a nonblocking flush here even if we are doing a SYNC_WAIT
* reclaim as we can deadlock with inode cluster removal.
* xfs_ifree_cluster() can lock the inode buffer before it locks the
* ip->i_lock, and we are doing the exact opposite here. As a result,
* doing a blocking xfs_itobp() to get the cluster buffer will result
* in an ABBA deadlock with xfs_ifree_cluster().
*
* As xfs_ifree_cluser() must gather all inodes that are active in the
* cache to mark them stale, if we hit this case we don't actually want
* to do IO here - we want the inode marked stale so we can simply
* reclaim it. Hence if we get an EAGAIN error on a SYNC_WAIT flush,
* just unlock the inode, back off and try again. Hopefully the next
* pass through will see the stale flag set on the inode.
*/
error = xfs_iflush(ip, SYNC_TRYLOCK | sync_mode);
if (sync_mode & SYNC_WAIT) {
if (error == EAGAIN) {
xfs_iunlock(ip, XFS_ILOCK_EXCL);
/* backoff longer than in xfs_ifree_cluster */
delay(2);
goto restart;
}
xfs_iflock(ip);
goto reclaim;
}
/*
* When we have to flush an inode but don't have SYNC_WAIT set, we
* flush the inode out using a delwri buffer and wait for the next
* call into reclaim to find it in a clean state instead of waiting for
* it now. We also don't return errors here - if the error is transient
* then the next reclaim pass will flush the inode, and if the error
* is permanent then the next sync reclaim will reclaim the inode and
* pass on the error.
*/
if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(ip->i_mount)) {
xfs_warn(ip->i_mount,
"inode 0x%llx background reclaim flush failed with %d",
(long long)ip->i_ino, error);
}
out:
xfs_iflags_clear(ip, XFS_IRECLAIM);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
/*
* We could return EAGAIN here to make reclaim rescan the inode tree in
* a short while. However, this just burns CPU time scanning the tree
* waiting for IO to complete and xfssyncd never goes back to the idle
* state. Instead, return 0 to let the next scheduled background reclaim
* attempt to reclaim the inode again.
*/
return 0;
reclaim:
xfs_ifunlock(ip);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
XFS_STATS_INC(xs_ig_reclaims);
/*
* Remove the inode from the per-AG radix tree.
*
* Because radix_tree_delete won't complain even if the item was never
* added to the tree assert that it's been there before to catch
* problems with the inode life time early on.
*/
spin_lock(&pag->pag_ici_lock);
if (!radix_tree_delete(&pag->pag_ici_root,
XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
ASSERT(0);
__xfs_inode_clear_reclaim(pag, ip);
spin_unlock(&pag->pag_ici_lock);
/*
* Here we do an (almost) spurious inode lock in order to coordinate
* with inode cache radix tree lookups. This is because the lookup
* can reference the inodes in the cache without taking references.
*
* We make that OK here by ensuring that we wait until the inode is
* unlocked after the lookup before we go ahead and free it. We get
* both the ilock and the iolock because the code may need to drop the
* ilock one but will still hold the iolock.
*/
xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
xfs_qm_dqdetach(ip);
xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
xfs_inode_free(ip);
return error;
}
/*
* Handle logging requirements of various synchronous types of write.
*/
int
xfs_write_sync_logforce(
xfs_mount_t *mp,
xfs_inode_t *ip)
{
int error = 0;
/*
* 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) &&
!(ip->i_update_size)) {
xfs_inode_log_item_t *iip = ip->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(ip) > 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(ip, XFS_ILOCK_EXCL);
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, NULL);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
}
return error;
}
STATIC uint
xfs_inode_item_push(
struct xfs_log_item *lip,
struct list_head *buffer_list)
{
struct xfs_inode_log_item *iip = INODE_ITEM(lip);
struct xfs_inode *ip = iip->ili_inode;
struct xfs_buf *bp = NULL;
uint rval = XFS_ITEM_SUCCESS;
int error;
if (xfs_ipincount(ip) > 0)
return XFS_ITEM_PINNED;
if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
return XFS_ITEM_LOCKED;
/*
* Re-check the pincount now that we stabilized the value by
* taking the ilock.
*/
if (xfs_ipincount(ip) > 0) {
rval = XFS_ITEM_PINNED;
goto out_unlock;
}
/*
* Stale inode items should force out the iclog.
*/
if (ip->i_flags & XFS_ISTALE) {
rval = XFS_ITEM_PINNED;
goto out_unlock;
}
/*
* Someone else is already flushing the inode. Nothing we can do
* here but wait for the flush to finish and remove the item from
* the AIL.
*/
if (!xfs_iflock_nowait(ip)) {
rval = XFS_ITEM_FLUSHING;
goto out_unlock;
}
ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount));
ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount));
spin_unlock(&lip->li_ailp->xa_lock);
error = xfs_iflush(ip, &bp);
if (!error) {
if (!xfs_buf_delwri_queue(bp, buffer_list))
rval = XFS_ITEM_FLUSHING;
xfs_buf_relse(bp);
}
spin_lock(&lip->li_ailp->xa_lock);
out_unlock:
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return rval;
}
