Decoding a merge dir, whose origin's parent is under a redirected
lower dir is not always possible. As a simple aproximation, we do
not encode lower dir file handles when overlay has multiple lower
layers and origin is below the topmost lower layer.

We should later relax this condition and copy up only the parent
that is under a redirected lower.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

We only need to encode origin if there is a chance that the same object was
encoded pre copy up and then we need to stay consistent with the same
encoding also after copy up.

In case a non-pure upper is not indexed, then it was copied up before NFS
export support was enabled. In that case, we don't need to worry about
staying consistent with pre copy up encoding and we encode an upper file
handle.

This mitigates the problem that with no index, we cannot find an upper
inode from origin inode, so we cannot decode a non-indexed upper from
origin file handle.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.

To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.

Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.

An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.

The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
  delicate re-factoring of ovl_lookup() to get a dentry with overlay
  parent info. It was preferred to avoid doing that re-factoring unless
  it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
  trivial) code path of d_splice_alias() could be traveled and that
  meant writing more tests and introduces race cases that are very hard
  to hit on purpose. Taking the path of connecting overlay dentry by
  forward lookup is therefore the safe and boring way to avoid surprises.

The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
  the overlay dentry by real dentry path. These subtleties are usually
  handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
  interrupted by rename and restarting connect forever.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Decoding an upper file handle is done by decoding the upper dentry from
underlying upper fs, finding or allocating an overlay inode that is
hashed by the real upper inode and instantiating an overlay dentry with
that inode.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Encode overlay file handles as struct ovl_fh containing the file handle
encoding of the real upper inode.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Document NFS export design.
Followup patches will implement this design.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Those helpers are going to be used by overlayfs to implement
NFS export decode.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

We need to make some room in struct ovl_entry to store information
about redirected ancestors for NFS export, so cram two booleans as
bit flags.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

With NFS export, some operations on decoded file handles (e.g. open,
link, setattr, xattr_set) may call copy up with a disconnected non-dir.
In this case, we will copy up lower inode to index dir without
linking it to upper dir.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

This is required for NFS export.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

This is needed for using ovl_get_inode() for decoding file handles
for NFS export.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

The helper is needed to lookup an index by file handle for NFS export.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Orphan index entries are non-dir index entries whose union nlink count
dropped to zero. With index=on, orphan index entries are removed on
mount. With NFS export feature enabled, orphan index entries are replaced
with white out index entries to block future open by handle from opening
the lower file.

When dir index has a stale 'upper' xattr, we assume that the upper dir
was removed and we treat the dir index as orphan entry that needs to be
whited out or removed.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

With NFS export feature enabled, when overlay inode nlink drops to
zero, instead of removing the index entry, replace it with a whiteout
index entry.

This is needed for NFS export in order to prevent future open by handle
from opening the lower file directly.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When non-dir index union nlink drops to zero the non-dir index
is cleaned. Do the same for directory type index entries when
union directory is removed.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

With the NFS export feature enabled, all dirs are indexed on copy up.
Non-dir files are copied up directly to indexdir and then hardlinked
to upper dir.

Directories are copied up to indexdir, then an index entry is created
in indexdir with 'upper' xattr pointing to the copied up dir and then
the copied up dir is moved to upper dir.

Directory index is also used for consistency verification, like
detecting multiple redirected dirs to the same lower dir on lookup.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

With the NFS export feature enabled, all non-dir are indexed on copy up.
The copy up origin inode of an indexed non-dir can be used as a unique
identifier of the overlay object.

The full index is also used for consistency verfication, like detecting
multiple non-hardlink uppers with the same 'origin' on lookup.

Directory index on copy up will be implemented by following patch.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

The helper determines which lower file needs to be indexed
on copy up and before nlink changes.

For index=on, the helper evaluates to true for lower hardlinks.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

A previous failed attempt to create or whiteout a directory index may
leave index entries named '#%x' in the index dir. Cleanup those temp
entries on mount instead of failing the mount.

In the future, we may drop 'work' dir and use 'index' dir instead.
This change is enough for cleaning up copy up leftovers 'from the future',
but it is not enough for cleaning up rmdir leftovers 'from the future'
(i.e. temp dir containing whiteouts).
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Directory index entries should have 'upper' xattr pointing to the real
upper dir. Verifying that the upper dir file handle is not stale is
expensive, so only verify stale directory index entries on mount if
NFS export feature is enabled.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Whiteout index entries are used as an indication that an exported
overlay file handle should be treated as stale (i.e. after unlink
of the overlay inode).

Check on mount that whiteout index entries have a name that looks like
a valid file handle and cleanup invalid index entries.

For whiteout index entries, do not check that they also have valid
origin fh and nlink xattr, because those xattr do not exist for a
whiteout index entry.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

A directory index is a directory type entry in index dir with a
"trusted.overlay.upper" xattr containing an encoded ovl_fh of the merge
directory upper dir inode.

On lookup of non-dir files, lower file is followed by origin file handle.
On lookup of dir entries, lower dir is found by name and then compared
to origin file handle. We only trust dir index if we verified that lower
dir matches origin file handle, otherwise index may be inconsistent and
we ignore it.

If we find an indexed non-upper dir or an indexed merged dir, whose
index 'upper' xattr points to a different upper dir, that means that the
lower directory may be also referenced by another upper dir via redirect,
so we fail the lookup on inconsistency error.

To be consistent with directory index entries format, the association of
index dir to upper root dir, that was stored by older kernels in
"trusted.overlay.origin" xattr is now stored in "trusted.overlay.upper"
xattr. This also serves as an indication that overlay was mounted with a
kernel that support index directory entries. For backward compatibility,
if an 'origin' xattr exists on the index dir we also verify it on mount.

Directory index entries are going to be used for NFS export.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

On a malformed overlay, several redirected dirs can point to the same
dir on a lower layer. This presents a similar challenge as broken
hardlinks, because different objects in the overlay can return the same
st_ino/st_dev pair from stat(2).

For broken hardlinks, we do not provide constant st_ino on copy up to
avoid this inconsistency. When NFS export feature is enabled, apply
the same logic to files and directories with unverified lower origin.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When the NFS export feature is enabled, overlayfs implicitly enables the
feature "verify_lower". When the "verify_lower" feature is enabled, a
directory inode found in lower layer by name or by redirect_dir is
verified against the file handle of the copy up origin that is stored in
the upper layer.

This introduces a change of behavior for the case of lower layer
modification while overlay is offline. A lower directory created or
moved offline under an exisitng upper directory, will not be merged with
that upper directory.

The NFS export feature should not be used after copying layers, because
the new lower directory inodes would fail verification and won't be
merged with upper directories.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Introduce the "nfs_export" config, module and mount options.

The NFS export feature depends on the "index" feature and enables two
implicit overlayfs features: "index_all" and "verify_lower".
The "index_all" feature creates an index on copy up of every file and
directory. The "verify_lower" feature uses the full index to detect
overlay filesystems inconsistencies on lookup, like redirect from
multiple upper dirs to the same lower dir.

NFS export can be enabled for non-upper mount with no index. However,
because lower layer redirects cannot be verified with the index, enabling
NFS export support on an overlay with no upper layer requires turning off
redirect follow (e.g. "redirect_dir=nofollow").

The full index may incur some overhead on mount time, especially when
verifying that lower directory file handles are not stale.

NFS export support, full index and consistency verification will be
implemented by following patches.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Document that inode index feature solves breaking hard links on
copy up.

Simplify Kconfig backward compatibility disclaimer.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Remove the "origin" language from the functions that handle set, get
and verify of "origin" xattr and pass the xattr name as an argument.

The same helpers are going to be used for NFS export to get, get and
verify the "upper" xattr for directory index entries.

ovl_verify_origin() is now a helper used only to verify non upper
file handle stored in "origin" xattr of upper inode.

The upper root dir file handle is still stored in "origin" xattr on
the index dir for backward compatibility. This is going to be changed
by the patch that adds directory index entries support.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Pass the fs instance with lower_layers array instead of the dentry
lowerstack array to ovl_check_origin_fh(), because the dentry members
of lowerstack play no role in this helper.

This change simplifies the argument list of ovl_check_origin(),
ovl_cleanup_index() and ovl_verify_index().
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Re-factor ovl_check_origin() and ovl_get_origin(), so origin fh xattr is
read from upper inode only once during lookup with multiple lower layers
and only once when verifying index entry origin.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Store the fs root layer index inside ovl_layer struct, so we can
get the root fs layer index from merge dir lower layer instead of
find it with ovl_find_layer() helper.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When work dir creation fails, a warning is emitted and overlay is
mounted r/o. Trying to remount r/w will fail with no work dir.

When index dir creation fails, the same warning is emitted and overlay
is mounted r/o, but trying to remount r/w will succeed. This may cause
unintentional corruption of filesystem consistency.

Adjust the behavior of index dir creation failure to that of work dir
creation failure and do not allow to remount r/w. User needs to state
an explicitly intention to work without an index by mounting with
option 'index=off' to allow r/w mount with no index dir.

When mounting with option 'index=on' and no 'upperdir', index is
implicitly disabled, so do not warn about no file handle support.

The issue was introduced with inodes index feature in v4.13, but this
patch will not apply cleanly before ovl_fill_super() re-factoring in
v4.15.

Fixes: 02bcd157 ("ovl: introduce the inodes index dir feature")
Cc: <stable@vger.kernel.org> #v4.13
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Overlayfs falls back to index=off if lower/upper fs does not support
file handles. Do the same if upper fs does not support xattr.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

For a merge dir that was copied up before v4.12 or that was hand crafted
offline (e.g. mkdir {upper/lower}/dir), upper dir does not contain the
'trusted.overlay.origin' xattr.  In that case, stat(2) on the merge dir
returns the lower dir st_ino, but getdents(2) returns the upper dir d_ino.

After this change, on merge dir lookup, missing origin xattr on upper
dir will be fixed and 'impure' xattr will be fixed on parent of the legacy
merge dir.
Suggested-by: Nzhangyi (F) <yi.zhang@huawei.com>
Reviewed-by: Nzhangyi (F) <yi.zhang@huawei.com>
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

The optimization in ovl_cache_get_impure() that tries to remove an
unneeded "impure" xattr needs to take mnt_want_write() on upper fs.

Fixes: 4edb83bb ("ovl: constant d_ino for non-merge dirs")
Cc: <stable@vger.kernel.org> #v4.14
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

There are several write operations on upper fs not covered by
mnt_want_write():

- test set/remove OPAQUE xattr
- test create O_TMPFILE
- set ORIGIN xattr in ovl_verify_origin()
- cleanup of index entries in ovl_indexdir_cleanup()

Some of these go way back, but this patch only applies over the
v4.14 re-factoring of ovl_fill_super().

Cc: <stable@vger.kernel.org> #v4.14
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

The functions ovl_lower_positive() and ovl_check_empty_dir() both take
inode mutex on the real lower dir under ovl_want_write() which takes
the upper_mnt sb_writers lock.

While this is not a clear locking order or layering violation, it creates
an undesired lock dependency between two unrelated layers for no good
reason.

This lock dependency materializes to a false(?) positive lockdep warning
when calling rmdir() on a nested overlayfs, where both nested and
underlying overlayfs both use the same fs type as upper layer.

rmdir() on the nested overlayfs creates the lock chain:
  sb_writers of upper_mnt (e.g. tmpfs) in ovl_do_remove()
  ovl_i_mutex_dir_key[] of lower overlay dir in ovl_lower_positive()

rmdir() on the underlying overlayfs creates the lock chain in
reverse order:
  ovl_i_mutex_dir_key[] of lower overlay dir in vfs_rmdir()
  sb_writers of nested upper_mnt (e.g. tmpfs) in ovl_do_remove()

To rid of the unneeded locking dependency, move both ovl_lower_positive()
and ovl_check_empty_dir() to before ovl_want_write() in rmdir() and
rename() implementation.

This change spreads the pieces of ovl_check_empty_and_clear() directly
inside the rmdir()/rename() implementations so the helper is no longer
needed and removed.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

As a writable mount, it is not expected for overlayfs to return
EINVAL/EROFS for fsync, even if dir/file is not changed.

This commit fixes the case of fsync of directory, which is easier to
address, because overlayfs already implements fsync file operation for
directories.

The problem reported by Raphael is that new PostgreSQL 10.0 with a
database in overlayfs where lower layer in squashfs fails to start.
The failure is due to fsync error, when PostgreSQL does fsync on all
existing db directories on startup and a specific directory exists
lower layer with no changes.
Reported-by: NRaphael Hertzog <raphael@ouaza.com>
Cc: <stable@vger.kernel.org> # v3.18
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Tested-by: NRaphaël Hertzog <hertzog@debian.org>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

fsnotify pins a watched directory inode in cache, but if directory dentry
is released, new lookup will allocate a new dentry and a new inode.
Directory events will be notified on the new inode, while fsnotify listener
is watching the old pinned inode.

Hash all directory inodes to reuse the pinned inode on lookup. Pure upper
dirs are hashes by real upper inode, merge and lower dirs are hashed by
real lower inode.

The reference to lower inode was being held by the lower dentry object
in the overlay dentry (oe->lowerstack[0]). Releasing the overlay dentry
may drop lower inode refcount to zero. Add a refcount on behalf of the
overlay inode to prevent that.

As a by-product, hashing directory inodes also detects multiple
redirected dirs to the same lower dir and uncovered redirected dir
target on and returns -ESTALE on lookup.

The reported issue dates back to initial version of overlayfs, but this
patch depends on ovl_inode code that was introduced in kernel v4.13.

Cc: <stable@vger.kernel.org> #v4.13
Reported-by: NNiklas Cassel <niklas.cassel@axis.com>
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>
Tested-by: NNiklas Cassel <niklas.cassel@axis.com>