The majority of remaining references to struct xfs_defer_ops in XFS
are associated with xfs_defer_add(). At this point, there are no
more external xfs_defer_ops users left. All instances of
xfs_defer_ops are embedded in the transaction, which means we can
safely pass the transaction down to the dfops add interface.

Update xfs_defer_add() to receive the transaction as a parameter.
Various subsystems implement wrappers to allocate and construct the
context specific data structures for the associated deferred
operation type. Update these to also carry the transaction down as
needed and clean up unused dfops parameters along the way.

This removes most of the remaining references to struct
xfs_defer_ops throughout the code and facilitates removal of the
structure.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
[darrick: fix unused variable warnings with ftrace disabled]
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

The current semantics of xfs_defer_finish() require the caller to
call xfs_defer_cancel() on error. This is slightly inconsistent with
transaction commit error handling where a failed commit cleans up
the transaction before returning.

More significantly, the only requirement for exposure of
->dop_pending outside of xfs_defer_finish() is so that
xfs_defer_cancel() can drain it on error. Since the only recourse of
xfs_defer_finish() errors is cancellation, mirror the transaction
logic and cancel remaining dfops before returning from
xfs_defer_finish() with an error.

Beside simplifying xfs_defer_finish() semantics, this ensures that
xfs_defer_finish() always returns with an empty ->dop_pending and
thus facilitates removal of the list from xfs_defer_ops.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Inodes that are held across deferred operations are explicitly
joined to the dfops structure to ensure appropriate relogging.
While inodes are currently joined explicitly, we can detect the
conditions that require relogging at dfops finish time by inspecting
the transaction item list for inodes with ili_lock_flags == 0.

Replace the xfs_defer_ijoin() infrastructure with such detection and
automatic relogging of held inodes. This eliminates the need for the
per-dfops inode list, replaced by an on-stack variant in
xfs_defer_trans_roll().
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Log items that require relogging during deferred operations
processing are explicitly joined to the associated dfops via the
xfs_defer_*join() helpers. These calls imply that the associated
object is "held" by the transaction such that when rolled, the item
can be immediately joined to a follow up transaction. For buffers,
this means the buffer remains locked and held after each roll. For
inodes, this means that the inode remains locked.

Failure to join a held item to the dfops structure means the
associated object pins the tail of the log while dfops processing
completes, because the item never relogs and is not unlocked or
released until deferred processing completes.

Currently, all buffers that are held in transactions (XFS_BLI_HOLD)
with deferred operations are explicitly joined to the dfops. This is
not the case for inodes, however, as various contexts defer
operations to transactions with held inodes without explicit joins
to the associated dfops (and thus not relogging).

While this is not a catastrophic problem, it is not ideal. Given
that we want to eventually relog such items automatically during
dfops processing, start by explicitly adding these missing
xfs_defer_ijoin() calls. A call is added everywhere an inode is
joined to a transaction without transferring lock ownership and
said transaction runs deferred operations.

All xfs_defer_ijoin() calls will eventually be replaced by automatic
dfops inode relogging. This patch essentially implements the
behavior change that would otherwise occur due to automatic inode
dfops relogging.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

We have a few places that already check if an inode has actual data in
the COW fork to avoid work on reflink inodes that do not actually have
outstanding COW blocks.  There are a few more places that can avoid
working if doing the same check, so add a documented helper for this
condition and use it in all places where it makes sense.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Every caller of xfs_defer_finish() now passes the transaction and
its associated ->t_dfops. The xfs_defer_ops parameter is therefore
no longer necessary and can be removed.

Since most xfs_defer_finish() callers also have to consider
xfs_defer_cancel() on error, update the latter to also receive the
transaction for consistency. The log recovery code contains an
outlier case that cancels a dfops directly without an available
transaction. Retain an internal wrapper to support this outlier case
for the time being.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NBill O'Donnell <billodo@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

At this point, the transaction subsystem completely manages deferred
items internally such that the common and boilerplate
xfs_trans_alloc() -> xfs_defer_init() -> xfs_defer_finish() ->
xfs_trans_commit() sequence can be replaced with a simple
transaction allocation and commit.

Remove all such boilerplate deferred ops code. In doing so, we
change each case over to use the dfops in the transaction and
specifically eliminate:

- The on-stack dfops and associated xfs_defer_init() call, as the
  internal dfops is initialized on transaction allocation.
- xfs_bmap_finish() calls that precede a final xfs_trans_commit() of
  a transaction.
- xfs_defer_cancel() calls in error handlers that precede a
  transaction cancel.

The only deferred ops calls that remain are those that are
non-deterministic with respect to the final commit of the associated
transaction or are open-coded due to special handling.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NBill O'Donnell <billodo@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

All callers either explicitly initialize a dfops or pass a
transaction with an internal dfops. Drop the hacky old dfops
replacement logic and use the one associated with the transaction.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NBill O'Donnell <billodo@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

The error argument to xfs_btree_del_cursor already understands the
"nonzero for error" semantics, so remove pointless error testing in the
callers and pass it directly.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NCarlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

All but one caller of xfs_defer_init() passes in the ->t_firstblock
of the associated transaction. The one outlier is
xlog_recover_process_intents(), which simply passes a dummy value
because a valid pointer is required. This firstblock variable can
simply be removed.

At this point we could remove the xfs_defer_init() firstblock
parameter and initialize ->t_firstblock directly. Even that is not
necessary, however, because ->t_firstblock is automatically
reinitialized in the new transaction on a transaction roll. Since
xfs_defer_init() should never occur more than once on a particular
transaction (since the corresponding finish will roll it), replace
the reinit from xfs_defer_init() with an assert that verifies the
transaction has a NULLFSBLOCK firstblock.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

All callers pass ->t_firstblock from the current transaction.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

All callers pass ->t_firstblock from the current transaction.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Convert all xfs_bunmapi() callers to ->t_firstblock.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Convert all xfs_bmapi_write() users to ->t_firstblock.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

xfs_bmapi_write() always expects a valid firstblock pointer. It
immediately dereferences the pointer to help determine how to
initialize the bma.minleft field. The remaining accesses are
related to modifying btree format forks, which is only relevant for
!COW fork callers.

The reflink code passes a NULL transaction to xfs_bmapi_write() in a
couple places that do COW fork unwritten conversion. The purpose of
the firstblock field is to track the first block allocation in the
current transaction, so technically firstblock should not be
required for these callers either.

Tweak xfs_bmapi_write() to initialize the bma correctly without
accessing the firstblock pointer if no transaction is provided in
the first place. Update the reflink callers to pass NULL instead of
otherwise unused firstblock references.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Most callers of xfs_defer_init() immediately attach the dfops
structure to a transaction. Add a transaction parameter to eliminate
much of this boilerplate code. This also helps self-document the
fact that many codepaths now expect a dfops pointer implicitly via
xfs_trans->t_dfops.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Use ->t_dfops of the transaction from the caller. Reset it before we
return to avoid leaks of local stack memory.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

The xfs_btree_cur.bc_private.a.dfops field is only ever initialized
by the refcountbt cursor init function. The only caller of that
function with a non-NULL dfops is from deferred completion context,
which already has attached to ->t_dfops.

In addition to that, the only actual reference of a.dfops is the
cursor duplication function, which means the field is effectively
unused.

Remove the dfops field from the bc_private.a union. Any future users
can acquire the dfops from the transaction. This patch does not
change behavior.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Now that all xfs_bunmapi() callers use ->t_dfops, remove the
unnecessary parameter and access ->t_dfops directly. This patch does
not change behavior.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Use ->t_dfops for all remaining xfs_bunmapi() callers. This prepares
the latter to no longer require a dfops parameter.

Note that xfs_itruncate_extents_flags() associates a local dfops
with a transaction provided from the caller. Since there are
multiple callers, set and reset ->t_dfops before the function
returns to avoid exposure of stack memory to the caller.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Now that all callers use ->t_dfops, the xfs_bmapi_write() dfops
parameter is no longer necessary. Remove it and access ->t_dfops
directly. This patch does not change behavior.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Attach ->t_dfops for all remaining callers of xfs_bmapi_write().
This prepares the latter to no longer require a separate dfops
parameter.

Note that xfs_symlink() already uses ->t_dfops. Fix up the local
references for consistency.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

A couple COW fork unwritten extent conversion helpers pass an
uninitialized dfops pointer to xfs_bmapi_write(). This does not
cause problems because conversion does not use a transaction or the
dfops structure for the COW fork.  Drop the uninitialized usage of
dfops in these codepaths and pass NULL along to xfs_bmapi_write()
instead.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NCarlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

We only have one caller left, and open coding the simple extent list
lookup in it allows us to make the code both more understandable and
reuse calculations and variables already present.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

We already have to check for overlapping COW extents everytime we
come back to a page in xfs_writepage_map / xfs_map_cow, so this
additional trim is not required.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Remove the verbose license text from XFS files and replace them
with SPDX tags. This does not change the license of any of the code,
merely refers to the common, up-to-date license files in LICENSES/

This change was mostly scripted. fs/xfs/Makefile and
fs/xfs/libxfs/xfs_fs.h were modified by hand, the rest were detected
and modified by the following command:

for f in `git grep -l "GNU General" fs/xfs/` ; do
	echo $f
	cat $f | awk -f hdr.awk > $f.new
	mv -f $f.new $f
done

And the hdr.awk script that did the modification (including
detecting the difference between GPL-2.0 and GPL-2.0+ licenses)
is as follows:

$ cat hdr.awk
BEGIN {
	hdr = 1.0
	tag = "GPL-2.0"
	str = ""
}

/^ \* This program is free software/ {
	hdr = 2.0;
	next
}

/any later version./ {
	tag = "GPL-2.0+"
	next
}

/^ \*\// {
	if (hdr > 0.0) {
		print "// SPDX-License-Identifier: " tag
		print str
		print $0
		str=""
		hdr = 0.0
		next
	}
	print $0
	next
}

/^ \* / {
	if (hdr > 1.0)
		next
	if (hdr > 0.0) {
		if (str != "")
			str = str "\n"
		str = str $0
		next
	}
	print $0
	next
}

/^ \*/ {
	if (hdr > 0.0)
		next
	print $0
	next
}

// {
	if (hdr > 0.0) {
		if (str != "")
			str = str "\n"
		str = str $0
		next
	}
	print $0
}

END { }
$
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

There are two uses of dqget here -- one is to return the dquot for a
given type and id, and the other is to return the dquot for a given type
and inode.  Those are two separate things, so split them into two
smaller functions.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

The flags argument is always zero, get rid of it.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

xfs_reflink_clear_inode_flag double-joins an inode to a transaction,
which is not allowed.  Fix that and document that the caller must have
already joined it.
Signed-Off-By: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
[darrick: edit out trace for nonexistent ASSERT]
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

xfs_reflink_cancel_cow_range joins an inode twice to the same
transaction.  This is not allowed, so fix it and document that the
callers of xfs_reflink_cancel_cow_blocks() must have already joined the
inode to the permanent transaction passed in.
Signed-Off-By: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
[darrick: edited the commit log to remove trace for nonexistent ASSERT]
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Use xfs_iext_prev_extent to skip to the previous extent instead of
opencoding it.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

Now that we convert COW preallocations from unwritten to real on every
call this function needs to be called with the ilock held exclusively.

Fortunately we already do that, but update the assert to match.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

The AGFL perag reservation type accounts all allocations that feed
into (or are released from) the allocation group free list (agfl).
The purpose of the reservation is to support worst case conditions
for the reverse mapping btree (rmapbt). As such, the agfl
reservation usage accounting only considers rmapbt usage when the
in-core counters are initialized at mount time.

This implementation inconsistency leads to divergence of the in-core
and on-disk usage accounting over time. In preparation to resolve
this inconsistency and adjust the AGFL reservation into an rmapbt
specific reservation, rename the AGFL reservation type and
associated accounting fields to something more rmapbt-specific. Also
fix up a couple tracepoints that incorrectly use the AGFL
reservation type to pass the agfl state of the associated extent
where the raw reservation type is expected.

Note that this patch does not change perag reservation behavior.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

xfs_trans_alloc() does GFP_KERNEL allocation, and we can call it
while holding pages locked for writeback in the ->writepages path.
The memory allocation is allowed to wait on pages under writeback,
and so can wait on pages that are tagged as writeback by the
caller.

This affects both pre-IO submission and post-IO submission paths.
Hence xfs_setsize_trans_alloc(), xfs_reflink_end_cow(),
xfs_iomap_write_unwritten() and xfs_reflink_cancel_cow_range().
xfs_iomap_write_unwritten() already does the right thing, but the
others don't. Fix them.
Signed-Off-By: NDave Chinner <dchinner@redhat.com>
Fixes: 281627df ("xfs: log file size updates at I/O completion time")
Fixes: 43caeb18 ("xfs: move mappings from cow fork to data fork after copy-write)"
Reviewed-by: NEric Sandeen <sandeen@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>

There's a really bad bug in xfs_reflink_allocate_cow -- if bmapi_write
can return a zero error code but no mappings.  This happens if there's
an extent size hint (which causes allocation requests to be rounded to
extsz granularity internally), but there wasn't a big enough chunk of
free space to start filling at the extsz granularity and fill even one
block of the range that we actually requested.

In any case, if we got no mappings we can't possibly do anything useful
with the contents of imap, so we must bail out with ENOSPC here.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

Since the CoW fork only exists in memory, it is incorrect to update the
on-disk quota block counts when we modify the CoW fork.  Unlike the data
fork, even real extents in the CoW fork are only delalloc-style
reservations (on-disk they're owned by the refcountbt) so they must not
be tracked in the on disk quota info.  Ensure the i_delayed_blks
accounting reflects this too.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

Reflink and dedupe operations remap blocks from a source file into a
destination file.  The destination file needs exclusive locks on all
levels because we're updating its block map, but the source file isn't
undergoing any block map changes so we can use a shared lock.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

Refactor xfs_lock_two_inodes to take separate locking modes for each
inode.  Specifically, this enables us to take a SHARED lock on one inode
and an EXCL lock on the other.  The lock class (MMAPLOCK/ILOCK) must be
the same for each inode.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

Before we share blocks between files, we need to break the pnfs leases
on the layout before we start slicing and dicing the block map.  The
structure of this function sets us up for the lock contention reduction
in the next patch.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>