Enforcing exclusive ownership on upper/work dirs caused a docker
regression: https://github.com/moby/moby/issues/34672.

Euan spotted the regression and pointed to the offending commit.
Vivek has brought the regression to my attention and provided this
reproducer:

Terminal 1:

  mount -t overlay -o workdir=work,lowerdir=lower,upperdir=upper none
        merged/

Terminal 2:

  unshare -m

Terminal 1:

  umount merged
  mount -t overlay -o workdir=work,lowerdir=lower,upperdir=upper none
        merged/
  mount: /root/overlay-testing/merged: none already mounted or mount point
         busy

To fix the regression, I replaced the error with an alarming warning.
With index feature enabled, mount does fail, but logs a suggestion to
override exclusive dir protection by disabling index.
Note that index=off mount does take the inuse locks, so a concurrent
index=off will issue the warning and a concurrent index=on mount will fail.

Documentation was updated to reflect this change.

Fixes: 2cac0c00 ("ovl: get exclusive ownership on upper/work dirs")
Cc: <stable@vger.kernel.org> # v4.13
Reported-by: NEuan Kemp <euank@euank.com>
Reported-by: NVivek Goyal <vgoyal@redhat.com>
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Use the ovl_lock_rename_workdir() helper which requires
unlock_rename() only on lock success.

Fixes: ("fd210b7d ovl: move copy up lock out")
Cc: <stable@vger.kernel.org> # v4.13
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

index dentry was not released when breaking out of the loop
due to index verification error.

Fixes: 415543d5 ("ovl: cleanup bad and stale index entries on mount")
Cc: <stable@vger.kernel.org> # v4.13
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Fixes: caf70cb2 ("ovl: cleanup orphan index entries")
Cc: <stable@vger.kernel.org> # v4.13
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Fixes: 359f392c ("ovl: lookup index entry for copy up origin")
Cc: <stable@vger.kernel.org> # v4.13
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

GFP_TEMPORARY was introduced by commit e12ba74d ("Group short-lived
and reclaimable kernel allocations") along with __GFP_RECLAIMABLE.  It's
primary motivation was to allow users to tell that an allocation is
short lived and so the allocator can try to place such allocations close
together and prevent long term fragmentation.  As much as this sounds
like a reasonable semantic it becomes much less clear when to use the
highlevel GFP_TEMPORARY allocation flag.  How long is temporary? Can the
context holding that memory sleep? Can it take locks? It seems there is
no good answer for those questions.

The current implementation of GFP_TEMPORARY is basically GFP_KERNEL |
__GFP_RECLAIMABLE which in itself is tricky because basically none of
the existing caller provide a way to reclaim the allocated memory.  So
this is rather misleading and hard to evaluate for any benefits.

I have checked some random users and none of them has added the flag
with a specific justification.  I suspect most of them just copied from
other existing users and others just thought it might be a good idea to
use without any measuring.  This suggests that GFP_TEMPORARY just
motivates for cargo cult usage without any reasoning.

I believe that our gfp flags are quite complex already and especially
those with highlevel semantic should be clearly defined to prevent from
confusion and abuse.  Therefore I propose dropping GFP_TEMPORARY and
replace all existing users to simply use GFP_KERNEL.  Please note that
SLAB users with shrinkers will still get __GFP_RECLAIMABLE heuristic and
so they will be placed properly for memory fragmentation prevention.

I can see reasons we might want some gfp flag to reflect shorterm
allocations but I propose starting from a clear semantic definition and
only then add users with proper justification.

This was been brought up before LSF this year by Matthew [1] and it
turned out that GFP_TEMPORARY really doesn't have a clear semantic.  It
seems to be a heuristic without any measured advantage for most (if not
all) its current users.  The follow up discussion has revealed that
opinions on what might be temporary allocation differ a lot between
developers.  So rather than trying to tweak existing users into a
semantic which they haven't expected I propose to simply remove the flag
and start from scratch if we really need a semantic for short term
allocations.

[1] http://lkml.kernel.org/r/20170118054945.GD18349@bombadil.infradead.org

[akpm@linux-foundation.org: fix typo]
[akpm@linux-foundation.org: coding-style fixes]
[sfr@canb.auug.org.au: drm/i915: fix up]
  Link: http://lkml.kernel.org/r/20170816144703.378d4f4d@canb.auug.org.au
Link: http://lkml.kernel.org/r/20170728091904.14627-1-mhocko@kernel.orgSigned-off-by: NMichal Hocko <mhocko@suse.com>
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au>
Acked-by: NMel Gorman <mgorman@suse.de>
Acked-by: NVlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Neil Brown <neilb@suse.de>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit b9ac5c27 ("ovl: hash overlay non-dir inodes by copy up origin")
verifies that the origin lower inode stored in the overlayfs inode matched
the inode of a copy up origin dentry found by lookup.

There is a false positive result in that check when lower fs does not
support file handles and copy up origin cannot be followed by file handle
at lookup time.

The false negative happens when finding an overlay inode in cache on a
copied up overlay dentry lookup. The overlay inode still 'remembers' the
copy up origin inode, but the copy up origin dentry is not available for
verification.

Relax the check in case copy up origin dentry is not available.

Fixes: b9ac5c27 ("ovl: hash overlay non-dir inodes by copy up...")
Cc: <stable@vger.kernel.org> # v4.13
Reported-by: NJordi Pujol <jordipujolp@gmail.com>
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Need to treat non-regular overlayfs files the same as regular files when
checking for an atime update.

Add a d_real() flag to make it return the upper dentry for all file types.
Reported-by: N"zhangyi (F)" <yi.zhang@huawei.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Add a separate flags argument (in addition to the open flags) to control
the behavior of d_real().
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

d_real() is never called with a negative dentry.  So remove the
d_is_negative() check (which would never trigger anyway, since d_is_reg()
returns false for a negative dentry).
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

While we could replace the smp_mb__before_spinlock() with the new
smp_mb__after_spinlock(), the normal pattern is to use
smp_store_release() to publish an object that is used for
lockless_dereference() -- and mirrors the regular rcu_assign_pointer()
/ rcu_dereference() patterns.
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: NIngo Molnar <mingo@kernel.org>

Impure directories are ones which contain objects with origins (i.e. those
that have been copied up).  These are relevant to readdir operation only
because of the d_ino field, no other transformation is necessary.  Also a
directory can become impure between two getdents(2) calls.

This patch creates a cache for impure directories.  Unlike the cache for
merged directories, this one only contains entries with origin and is not
refcounted but has a its lifetime tied to that of the dentry.

Similarly to the merged cache, the impure cache is invalidated based on a
version number.  This version number is incremented when an entry with
origin is added or removed from the directory.

If the cache is empty, then the impure xattr is removed from the directory.

This patch also fixes up handling of d_ino for the ".." entry if the parent
directory is merged.
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When all layers are on the same fs, and iterating a directory which may
contain copy up entries, call vfs_getattr() on the overlay entries to make
sure that d_ino will be consistent with st_ino from stat(2).

There is an overhead of lookup per upper entry in readdir.

The overhead is minimal if the iterated entries are already in dcache.  It
is also quite useful for the common case of 'ls -l' that readdir() pre
populates the dcache with the listed entries, making the following stat()
calls faster.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

actor's return value is taken as a bool (filled/not filled) so we need to
return the error in the context.
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Index should always be of the same file type as origin, except for
the case of a whiteout index.  A whiteout index should only exist
if all lower aliases have been unlinked, which means that finding
a lower origin on lookup whose index is a whiteout should be treated
as a lookup error.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Directory index entries are going to be used for looking up
redirected upper dirs by lower dir fh when decoding an overlay
file handle of a merge dir.

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

We don't know the verification rules for directory and whiteout
index entries, because they have not been implemented yet, so fail
to mount overlay rw if those entries are found to avoid corrupting
an index that was created by a newer kernel.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

inode_doinit_with_dentry() in SELinux wants to read the upper inode's xattr
to get security label, and ovl_xattr_get() calls ovl_dentry_real(), which
depends on dentry->d_inode, but d_inode is null and not initialized yet at
this point resulting in an Oops.

Fix by getting the upperdentry info from the inode directly in this case.
Reported-by: NEryu Guan <eguan@redhat.com>
Fixes: 09d8b586 ("ovl: move __upperdentry to ovl_inode")
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Firstly by applying the following with coccinelle's spatch:

	@@ expression SB; @@
	-SB->s_flags & MS_RDONLY
	+sb_rdonly(SB)

to effect the conversion to sb_rdonly(sb), then by applying:

	@@ expression A, SB; @@
	(
	-(!sb_rdonly(SB)) && A
	+!sb_rdonly(SB) && A
	|
	-A != (sb_rdonly(SB))
	+A != sb_rdonly(SB)
	|
	-A == (sb_rdonly(SB))
	+A == sb_rdonly(SB)
	|
	-!(sb_rdonly(SB))
	+!sb_rdonly(SB)
	|
	-A && (sb_rdonly(SB))
	+A && sb_rdonly(SB)
	|
	-A || (sb_rdonly(SB))
	+A || sb_rdonly(SB)
	|
	-(sb_rdonly(SB)) != A
	+sb_rdonly(SB) != A
	|
	-(sb_rdonly(SB)) == A
	+sb_rdonly(SB) == A
	|
	-(sb_rdonly(SB)) && A
	+sb_rdonly(SB) && A
	|
	-(sb_rdonly(SB)) || A
	+sb_rdonly(SB) || A
	)

	@@ expression A, B, SB; @@
	(
	-(sb_rdonly(SB)) ? 1 : 0
	+sb_rdonly(SB)
	|
	-(sb_rdonly(SB)) ? A : B
	+sb_rdonly(SB) ? A : B
	)

to remove left over excess bracketage and finally by applying:

	@@ expression A, SB; @@
	(
	-(A & MS_RDONLY) != sb_rdonly(SB)
	+(bool)(A & MS_RDONLY) != sb_rdonly(SB)
	|
	-(A & MS_RDONLY) == sb_rdonly(SB)
	+(bool)(A & MS_RDONLY) == sb_rdonly(SB)
	)

to make comparisons against the result of sb_rdonly() (which is a bool)
work correctly.
Signed-off-by: NDavid Howells <dhowells@redhat.com>

ovl_workdir_create() returns a valid index dentry or NULL.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Commit 54fb347e ("ovl: verify index dir matches upper dir")
introduced a new ovl_fh flag OVL_FH_FLAG_PATH_UPPER to indicate
an upper file handle, but forgot to add the flag to the mask of
valid flags, so index dir origin verification always discards
existing origin and stores a new one.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When linking a file with copy up origin into a new parent, mark the
new parent dir "impure".

Fixes: ee1d6d37 ("ovl: mark upper dir with type origin entries "impure"")
Cc: <stable@vger.kernel.org> # v4.12
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

On failure to prepare_creds(), mount fails with a random
return value, as err was last set to an integer cast of
a valid lower mnt pointer or set to 0 if inodes index feature
is enabled.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Fixes: 3fe6e52f ("ovl: override creds with the ones from ...")
Cc: <stable@vger.kernel.org> # v4.7
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Signed-off-by: NAmir Goldstein <amir73il@gmail.com>

index entry should live only as long as there are upper or lower
hardlinks.

Cleanup orphan index entries on mount and when dropping the last
overlay inode nlink.

When about to cleanup or link up to orphan index and the index inode
nlink > 1, admit that something went wrong and adjust overlay nlink
to index inode nlink - 1 to prevent it from dropping below zero.
This could happen when adding lower hardlinks underneath a mounted
overlay and then trying to unlink them.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

With inodes index enabled, an overlay inode nlink counts the union of upper
and non-covered lower hardlinks. During the lifetime of a non-pure upper
inode, the following nlink modifying operations can happen:

1. Lower hardlink copy up
2. Upper hardlink created, unlinked or renamed over
3. Lower hardlink whiteout or renamed over

For the first, copy up case, the union nlink does not change, whether the
operation succeeds or fails, but the upper inode nlink may change.
Therefore, before copy up, we store the union nlink value relative to the
lower inode nlink in the index inode xattr trusted.overlay.nlink.

For the second, upper hardlink case, the union nlink should be incremented
or decremented IFF the operation succeeds, aligned with nlink change of the
upper inode. Therefore, before link/unlink/rename, we store the union nlink
value relative to the upper inode nlink in the index inode.

For the last, lower cover up case, we simplify things by preceding the
whiteout or cover up with copy up. This makes sure that there is an index
upper inode where the nlink xattr can be stored before the copied up upper
entry is unlink.

Return the overlay inode nlinks for indexed upper inodes on stat(2).
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Implement a copy up method for non-dir objects using index dir to
prevent breaking lower hardlinks on copy up.

This method requires that the inodes index dir feature was enabled and
that all underlying fs support file handle encoding/decoding.

On the first lower hardlink copy up, upper file is created in index dir,
named after the hex representation of the lower origin inode file handle.
On the second lower hardlink copy up, upper file is found in index dir,
by the same lower handle key.
On either case, the upper indexed inode is then linked to the copy up
upper path.

The index entry remains linked for future lower hardlink copy up and for
lower to upper inode map, that is needed for exporting overlayfs to NFS.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Move ovl_copy_up_start()/ovl_copy_up_end() out so that it's used for both
tempfile and workdir copy ups.
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Split up and rearrange copy up functions to make them better readable.
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

For rename, we need to ensure that an upper alias exists for hard links
before attempting the operation.  Introduce a flag in ovl_entry to track
the state of the upper alias.
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

This cleans up functions with too many arguments.
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Factor out helper for copying lower inode data and metadata to temp
upper inode, that is common to copy up using O_TMPFILE and workdir.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

On copy up of regular file using an O_TMPFILE, lock upper dir only
before linking the tempfile in place.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

Bad index entries are entries whose name does not match the
origin file handle stored in trusted.overlay.origin xattr.
Bad index entries could be a result of a system power off in
the middle of copy up.

Stale index entries are entries whose origin file handle is
stale. Stale index entries could be a result of copying layers
or removing lower entries while the overlay is not mounted.
The case of copying layers should be detected earlier by the
verification of upper root dir origin and index dir origin.

Both bad and stale index entries are detected and removed
on mount.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When inodes index feature is enabled, lookup in indexdir for the index
entry of lower real inode or copy up origin inode. The index entry name
is the hex representation of the lower inode file handle.

If the index dentry in negative, then either no lower aliases have been
copied up yet, or aliases have been copied up in older kernels and are
not indexed.

If the index dentry for a copy up origin inode is positive, but points
to an inode different than the upper inode, then either the upper inode
has been copied up and not indexed or it was indexed, but since then
index dir was cleared. Either way, that index cannot be used to indentify
the overlay inode.

If a positive dentry that matches the upper inode was found, then it is
safe to use the copy up origin st_ino for upper hardlinks, because all
indexed upper hardlinks are represented by the same overlay inode as the
copy up origin.

Set the INDEX type flag on an indexed upper dentry. A non-upper dentry
may also have a positive index from copy up of another lower hardlink.
This situation will be handled by following patches.

Index lookup is going to be used to prevent breaking hardlinks on copy up.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

An index dir contains persistent hardlinks to files in upper dir.
Therefore, we must never mount an existing index dir with a differnt
upper dir.

Store the upper root dir file handle in index dir inode when index
dir is created and verify the file handle before using an existing
index dir on mount.

Add an 'is_upper' flag to the overlay file handle encoding and set it
when encoding the upper root file handle. This is not critical for index
dir verification, but it is good practice towards a standard overlayfs
file handle format for NFS export.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>

When inodes index feature is enabled, verify that the file handle stored
in upper root dir matches the lower root dir or fail to mount.

If upper root dir has no stored file handle, encode and store the lower
root dir file handle in overlay.origin xattr.
Signed-off-by: NAmir Goldstein <amir73il@gmail.com>
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>