Pull btrfs fixes from David Sterba:

 - fix for improper fsync after hardlink

 - fix for a corruption during file deduplication

 - use after free fixes

 - RCU warning fix

 - fix for buffered write to nodatacow file

* tag 'for-4.19-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: Fix suspicious RCU usage warning in btrfs_debug_in_rcu
  btrfs: use after free in btrfs_quota_enable
  btrfs: btrfs_shrink_device should call commit transaction at the end
  btrfs: fix qgroup_free wrong num_bytes in btrfs_subvolume_reserve_metadata
  Btrfs: fix data corruption when deduplicating between different files
  Btrfs: sync log after logging new name
  Btrfs: fix unexpected failure of nocow buffered writes after snapshotting when low on space

Commit 672d599041c8 ("btrfs: Use wrapper macro for rcu string to remove
duplicate code") replaces some open coded RCU string handling with macro.

It turns out that btrfs_debug_in_rcu() is used for the first time and
the macro lacks lock/unlock of RCU string for non-debug case (i.e. when
the message is not printed), leading to suspicious RCU usage warning
when CONFIG_PROVE_RCU is on.

Fix this by adding a wrapper to call lock/unlock for the non-debug case
too.

Fixes: 672d599041c8 ("btrfs: Use wrapper macro for rcu string to remove duplicate code")
Reported-by: David Howells <dhowells@redhat.com>
Tested-by: David Howells <dhowells@redhat.com>
Signed-off-by: Misono Tomohiro <misono.tomohiro@jp.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

The issue here is that btrfs_commit_transaction() frees "trans" on both
the error and the success path.  So the problem would be if
btrfs_commit_transaction() succeeds, and then qgroup_rescan_init()
fails.  That means that "ret" is non-zero and "trans" is non-NULL and it
leads to a use after free inside the btrfs_end_transaction() macro.

Fixes: 340f1aa27f36 ("btrfs: qgroups: Move transaction management inside btrfs_quota_enable/disable")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Test case btrfs/164 reports use-after-free:

[ 6712.084324] general protection fault: 0000 [#1] PREEMPT SMP
..
[ 6712.195423]  btrfs_update_commit_device_size+0x75/0xf0 [btrfs]
[ 6712.201424]  btrfs_commit_transaction+0x57d/0xa90 [btrfs]
[ 6712.206999]  btrfs_rm_device+0x627/0x850 [btrfs]
[ 6712.211800]  btrfs_ioctl+0x2b03/0x3120 [btrfs]

Reason for this is that btrfs_shrink_device adds the resized device to
the fs_devices::resized_devices after it has called the last commit
transaction.

So the list fs_devices::resized_devices is not empty when
btrfs_shrink_device returns.  Now the parent function
btrfs_rm_device calls:

        btrfs_close_bdev(device);
        call_rcu(&device->rcu, free_device_rcu);

and then does the transactio ncommit. It goes through the
fs_devices::resized_devices in btrfs_update_commit_device_size and
leads to use-after-free.

Fix this by making sure btrfs_shrink_device calls the last needed
btrfs_commit_transaction before the return. This is consistent with what
the grow counterpart does and this makes sure the on-disk state is
persistent when the function returns.

Reported-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Tested-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>

After btrfs_qgroup_reserve_meta_prealloc(), num_bytes will be assigned
again by btrfs_calc_trans_metadata_size(). Once block_rsv fails, we
can't properly free the num_bytes of the previous qgroup_reserve. Use a
separate variable to store the num_bytes of the qgroup_reserve.

Delete the comment for the qgroup_reserved that does not exist and add a
comment about use_global_rsv.

Fixes: c4c129db5da8 ("btrfs: drop unused parameter qgroup_reserved")
CC: stable@vger.kernel.org # 4.18+
Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

If we deduplicate extents between two different files we can end up
corrupting data if the source range ends at the size of the source file,
the source file's size is not aligned to the filesystem's block size
and the destination range does not go past the size of the destination
file size.

Example:

  $ mkfs.btrfs -f /dev/sdb
  $ mount /dev/sdb /mnt

  $ xfs_io -f -c "pwrite -S 0x6b 0 2518890" /mnt/foo
  # The first byte with a value of 0xae starts at an offset (2518890)
  # which is not a multiple of the sector size.
  $ xfs_io -c "pwrite -S 0xae 2518890 102398" /mnt/foo

  # Confirm the file content is full of bytes with values 0x6b and 0xae.
  $ od -t x1 /mnt/foo
  0000000 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b
  *
  11467540 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b ae ae ae ae ae ae
  11467560 ae ae ae ae ae ae ae ae ae ae ae ae ae ae ae ae
  *
  11777540 ae ae ae ae ae ae ae ae
  11777550

  # Create a second file with a length not aligned to the sector size,
  # whose bytes all have the value 0x6b, so that its extent(s) can be
  # deduplicated with the first file.
  $ xfs_io -f -c "pwrite -S 0x6b 0 557771" /mnt/bar

  # Now deduplicate the entire second file into a range of the first file
  # that also has all bytes with the value 0x6b. The destination range's
  # end offset must not be aligned to the sector size and must be less
  # then the offset of the first byte with the value 0xae (byte at offset
  # 2518890).
  $ xfs_io -c "dedupe /mnt/bar 0 1957888 557771" /mnt/foo

  # The bytes in the range starting at offset 2515659 (end of the
  # deduplication range) and ending at offset 2519040 (start offset
  # rounded up to the block size) must all have the value 0xae (and not
  # replaced with 0x00 values). In other words, we should have exactly
  # the same data we had before we asked for deduplication.
  $ od -t x1 /mnt/foo
  0000000 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b
  *
  11467540 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b ae ae ae ae ae ae
  11467560 ae ae ae ae ae ae ae ae ae ae ae ae ae ae ae ae
  *
  11777540 ae ae ae ae ae ae ae ae
  11777550

  # Unmount the filesystem and mount it again. This guarantees any file
  # data in the page cache is dropped.
  $ umount /dev/sdb
  $ mount /dev/sdb /mnt

  $ od -t x1 /mnt/foo
  0000000 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b
  *
  11461300 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 00 00 00 00 00
  11461320 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  *
  11470000 ae ae ae ae ae ae ae ae ae ae ae ae ae ae ae ae
  *
  11777540 ae ae ae ae ae ae ae ae
  11777550

  # The bytes in range 2515659 to 2519040 have a value of 0x00 and not a
  # value of 0xae, data corruption happened due to the deduplication
  # operation.

So fix this by rounding down, to the sector size, the length used for the
deduplication when the following conditions are met:

  1) Source file's range ends at its i_size;
  2) Source file's i_size is not aligned to the sector size;
  3) Destination range does not cross the i_size of the destination file.

Fixes: e1d227a42ea2 ("btrfs: Handle unaligned length in extent_same")
CC: stable@vger.kernel.org # 4.2+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

When we add a new name for an inode which was logged in the current
transaction, we update the inode in the log so that its new name and
ancestors are added to the log. However when we do this we do not persist
the log, so the changes remain in memory only, and as a consequence, any
ancestors that were created in the current transaction are updated such
that future calls to btrfs_inode_in_log() return true. This leads to a
subsequent fsync against such new ancestor directories returning
immediately, without persisting the log, therefore after a power failure
the new ancestor directories do not exist, despite fsync being called
against them explicitly.

Example:

  $ mkfs.btrfs -f /dev/sdb
  $ mount /dev/sdb /mnt

  $ mkdir /mnt/A
  $ mkdir /mnt/B
  $ mkdir /mnt/A/C
  $ touch /mnt/B/foo
  $ xfs_io -c "fsync" /mnt/B/foo
  $ ln /mnt/B/foo /mnt/A/C/foo
  $ xfs_io -c "fsync" /mnt/A
  <power failure>

After the power failure, directory "A" does not exist, despite the explicit
fsync on it.

Instead of fixing this by changing the behaviour of the explicit fsync on
directory "A" to persist the log instead of doing nothing, make the logging
of the new file name (which happens when creating a hard link or renaming)
persist the log. This approach not only is simpler, not requiring addition
of new fields to the inode in memory structure, but also gives us the same
behaviour as ext4, xfs and f2fs (possibly other filesystems too).

A test case for fstests follows soon.

Fixes: 12fcfd22fe5b ("Btrfs: tree logging unlink/rename fixes")
Reported-by: Vijay Chidambaram <vvijay03@gmail.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Pull overlayfs updates from Miklos Szeredi:
 "This contains two new features:

   - Stack file operations: this allows removal of several hacks from
     the VFS, proper interaction of read-only open files with copy-up,
     possibility to implement fs modifying ioctls properly, and others.

   - Metadata only copy-up: when file is on lower layer and only
     metadata is modified (except size) then only copy up the metadata
     and continue to use the data from the lower file"

* tag 'ovl-update-4.19' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs: (66 commits)
  ovl: Enable metadata only feature
  ovl: Do not do metacopy only for ioctl modifying file attr
  ovl: Do not do metadata only copy-up for truncate operation
  ovl: add helper to force data copy-up
  ovl: Check redirect on index as well
  ovl: Set redirect on upper inode when it is linked
  ovl: Set redirect on metacopy files upon rename
  ovl: Do not set dentry type ORIGIN for broken hardlinks
  ovl: Add an inode flag OVL_CONST_INO
  ovl: Treat metacopy dentries as type OVL_PATH_MERGE
  ovl: Check redirects for metacopy files
  ovl: Move some dir related ovl_lookup_single() code in else block
  ovl: Do not expose metacopy only dentry from d_real()
  ovl: Open file with data except for the case of fsync
  ovl: Add helper ovl_inode_realdata()
  ovl: Store lower data inode in ovl_inode
  ovl: Fix ovl_getattr() to get number of blocks from lower
  ovl: Add helper ovl_dentry_lowerdata() to get lower data dentry
  ovl: Copy up meta inode data from lowest data inode
  ovl: Modify ovl_lookup() and friends to lookup metacopy dentry
  ...

a_ops->readpages() is only ever used for read-ahead.  Ensure that we
pass this information down to the block layer.

Link: http://lkml.kernel.org/r/20180621010725.17813-4-axboe@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Chris Mason <clm@fb.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Commit e9894fd3e3b3 ("Btrfs: fix snapshot vs nocow writting") forced
nocow writes to fallback to COW, during writeback, when a snapshot is
created. This resulted in writes made before creating the snapshot to
unexpectedly fail with ENOSPC during writeback when success (0) was
returned to user space through the write system call.

The steps leading to this problem are:

1. When it's not possible to allocate data space for a write, the
   buffered write path checks if a NOCOW write is possible.  If it is,
   it will not reserve space and success (0) is returned to user space.

2. Then when a snapshot is created, the root's will_be_snapshotted
   atomic is incremented and writeback is triggered for all inode's that
   belong to the root being snapshotted. Incrementing that atomic forces
   all previous writes to fallback to COW during writeback (running
   delalloc).

3. This results in the writeback for the inodes to fail and therefore
   setting the ENOSPC error in their mappings, so that a subsequent
   fsync on them will report the error to user space. So it's not a
   completely silent data loss (since fsync will report ENOSPC) but it's
   a very unexpected and undesirable behaviour, because if a clean
   shutdown/unmount of the filesystem happens without previous calls to
   fsync, it is expected to have the data present in the files after
   mounting the filesystem again.

So fix this by adding a new atomic named snapshot_force_cow to the
root structure which prevents this behaviour and works the following way:

1. It is incremented when we start to create a snapshot after triggering
   writeback and before waiting for writeback to finish.

2. This new atomic is now what is used by writeback (running delalloc)
   to decide whether we need to fallback to COW or not. Because we
   incremented this new atomic after triggering writeback in the
   snapshot creation ioctl, we ensure that all buffered writes that
   happened before snapshot creation will succeed and not fallback to
   COW (which would make them fail with ENOSPC).

3. The existing atomic, will_be_snapshotted, is kept because it is used
   to force new buffered writes, that start after we started
   snapshotting, to reserve data space even when NOCOW is possible.
   This makes these writes fail early with ENOSPC when there's no
   available space to allocate, preventing the unexpected behaviour of
   writeback later failing with ENOSPC due to a fallback to COW mode.

Fixes: e9894fd3e3b3 ("Btrfs: fix snapshot vs nocow writting")
Signed-off-by: Robbie Ko <robbieko@synology.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>