RCU free the struct inode. This will allow:

- Subsequent store-free path walking patch. The inode must be consulted for
  permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
  to take i_lock no longer need to take sb_inode_list_lock to walk the list in
  the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
  page lock to follow page->mapping.

The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.

In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.

The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.
Signed-off-by: NNick Piggin <npiggin@kernel.dk>

Make d_count non-atomic and protect it with d_lock. This allows us to ensure a
0 refcount dentry remains 0 without dcache_lock. It is also fairly natural when
we start protecting many other dentry members with d_lock.
Signed-off-by: NNick Piggin <npiggin@kernel.dk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

change nilfs_dat_commit_free and nilfs_inode_cachep static
to fix following warnings

fs/nilfs2/super.c:72:19: warning: symbol 'nilfs_inode_cachep' was not declared. Should it be static?
fs/nilfs2/dat.c:106:6: warning: symbol 'nilfs_dat_commit_free' was not declared. Should it be static?
Signed-off-by: NJiro SEKIBA <jir@unicus.jp>
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

Nilfs now can use sb->s_bdi to get backing_dev_info, so we use it
instead of ns_bdi on the nilfs object and remove ns_bdi.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

Nilfs hasn't supported the freeze/thaw feature because it didn't work
due to the peculiar design that multiple super block instances could
be allocated for a device.  This limitation was removed by the patch
"nilfs2: do not allocate multiple super block instances for a device".

So now this adds the freeze/thaw support to nilfs.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

The current implementation doesn't mount snapshots with checkpoint
numbers larger than INT_MAX since it uses match_int() for parsing
"cp=" mount option.

This uses simple_strtoull() for the conversion to resolve the issue.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This finally removes own inode allocator and destructor functions for
metadata files.  Several routines, nilfs_mdt_new(),
nilfs_mdt_new_common(), nilfs_mdt_clear(), nilfs_mdt_destroy(), and
nilfs_alloc_inode_common() will be gone.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

After applied the patch that unified sb instances, root dentry of
snapshots can be left in dcache even after their trees are unmounted.

The orphan root dentry/inode keeps a root object, and this causes
false positive of nilfs_checkpoint_is_mounted function.

This resolves the issue by having nilfs_checkpoint_is_mounted test
whether the root dentry is busy or not.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This makes use of iget5_locked to allocate or get inode for metadata
files to stop using own inode allocator.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This stops pre-allocating nilfs object in nilfs_get_sb routine, and
stops managing its life cycle by reference counting.

nilfs_find_or_create_nilfs() function, nilfs->ns_mount_mutex,
nilfs_objects list, and the reference counter will be removed through
the simplification.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This stops allocating multiple super block instances for a device.

All snapshots and a current mode mount (i.e. latest tree) will be
controlled with nilfs_root objects that are kept within an sb
instance.

nilfs_get_sb() is rewritten so that it always has a root object for
the latest tree and snapshots make additional root objects.

The root dentry of the latest tree is binded to sb->s_root even if it
isn't attached on a directory.  Root dentries of snapshots or the
latest tree are binded to mnt->mnt_root on which they are mounted.

With this patch, nilfs_find_sbinfo() function, nilfs->ns_supers list,
and nilfs->ns_current back pointer, are deleted.  In addition,
init_nilfs() and load_nilfs() are simplified since they will be called
once for a device, not repeatedly called for mount points.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This splits the code to attach snapshots into a separate routine for
convenience sake.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This splits the code to allocate root dentry into a separate routine
for convenience in successive changes.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This moves sbi->s_inodes_count and sbi->s_blocks_count into nilfs_root
object.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This rewrites functions using ifile so that they get ifile from
nilfs_root object, and will remove sbi->s_ifile.  Some functions that
don't know the root object are extended to receive it from caller.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

The previous export operations cannot handle multiple versions of
a filesystem if they belong to the same sb instance.

This adds a new type of file handle and extends export operations so
that they can get the inode specified by a checkpoint number as well
as an inode number and a generation number.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This puts a pointer to nilfs_root object in the private part of
on-memory inode, and makes nilfs_iget function pick up the inode with
the same root object.

Non-root inodes inherit its nilfs_root object from parent inode.  That
of the root inode is allocated through nilfs_attach_checkpoint()
function.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This uses iget5_locked instead of iget_locked so that gc cache can
look up inodes with an inode number and an optional checkpoint number.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

The current nilfs_destroy_inode() doesn't handle metadata file inodes
including gc inodes (dummy inodes used for garbage collection).

This allows nilfs_destroy_inode() to destroy inodes of metadata files.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

The BKL is only used in put_super, fill_super and remount_fs that are all
three protected by the superblocks s_umount rw_semaphore. Therefore it is
safe to remove the BKL entirely.
Signed-off-by: NJan Blunck <jblunck@infradead.org>
Signed-off-by: NArnd Bergmann <arnd@arndb.de>

This patch is a preparation necessary to remove the BKL from do_new_mount().
It explicitly adds calls to lock_kernel()/unlock_kernel() around
get_sb/fill_super operations for filesystems that still uses the BKL.

I've read through all the code formerly covered by the BKL inside
do_kern_mount() and have satisfied myself that it doesn't need the BKL
any more.

do_kern_mount() is already called without the BKL when mounting the rootfs
and in nfsctl. do_kern_mount() calls vfs_kern_mount(), which is called
from various places without BKL: simple_pin_fs(), nfs_do_clone_mount()
through nfs_follow_mountpoint(), afs_mntpt_do_automount() through
afs_mntpt_follow_link(). Both later functions are actually the filesystems
follow_link inode operation. vfs_kern_mount() is calling the specified
get_sb function and lets the filesystem do its job by calling the given
fill_super function.

Therefore I think it is safe to push down the BKL from the VFS to the
low-level filesystems get_sb/fill_super operation.

[arnd: do not add the BKL to those file systems that already
       don't use it elsewhere]
Signed-off-by: NJan Blunck <jblunck@infradead.org>
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Christoph Hellwig <hch@infradead.org>

Switch to the WRITE_FLUSH_FUA flag for log writes, remove the EOPNOTSUPP
detection for barriers and stop setting the barrier flag for discards.

tj: nilfs is now fixed to wait for discard completion.  Updated this
    patch accordingly and dropped warning about it.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <jaxboe@fusionio.com>

Instead of abusing a buffer_head flag just add a variant of
sync_dirty_buffer which allows passing the exact type of write
flag required.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

If nilfs_attach_checkpoint() gets a memory allocation failure during
creation of ifile, it will return without removing nilfs_sb_info
struct from ns_supers list.  When a concurrently mounted snapshot is
unmounted or another new snapshot is mounted after that, this causes
kernel oops as below:

> BUG: unable to handle kernel NULL pointer dereference at (null)
> IP: [<f83662ff>] nilfs_find_sbinfo+0x74/0xa4 [nilfs2]
> *pde = 00000000
> Oops: 0000 [#1] SMP
<snip>
> Call Trace:
>  [<f835dc29>] ? nilfs_get_sb+0x165/0x532 [nilfs2]
>  [<c1173c87>] ? ida_get_new_above+0x16d/0x187
>  [<c109a7f8>] ? alloc_vfsmnt+0x7e/0x10a
>  [<c1070790>] ? kstrdup+0x2c/0x40
>  [<c1089041>] ? vfs_kern_mount+0x96/0x14e
>  [<c108913d>] ? do_kern_mount+0x32/0xbd
>  [<c109b331>] ? do_mount+0x642/0x6a1
>  [<c101a415>] ? do_page_fault+0x0/0x2d1
>  [<c1099c00>] ? copy_mount_options+0x80/0xe2
>  [<c10705d8>] ? strndup_user+0x48/0x67
>  [<c109b3f1>] ? sys_mount+0x61/0x90
>  [<c10027cc>] ? sysenter_do_call+0x12/0x22

This fixes the problem.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Tested-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Cc: stable@kernel.org

[folded build fix from sfr]
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

This forces nilfs to check compatibility of feature flags so as to
reject a filesystem with unknown features when it mounts or remounts
the filesystem.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This removes nilfs_bmap_union and finally unifies three structures and
the union in bmap/btree code into one.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This adds is_remount argument to the parse_options() function that
obtains mount options from strings.

Previously, parse_options did not distinguish context whether it's
called for a new mount or remount, so the caller needed additional
verifications outside the function.

This allows parse_options to verify options and print messages
depending on the context.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This replaces seq_printf() with seq_puts() in nilfs_show_options for
mount options which have no argument.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

Nilfs has "discard" mount option which issues discard/TRIM commands to
underlying block device, but it lacks a complementary option and has
no way to disable the feature through remount.

This adds "nodiscard" option to resolve this imbalance.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

Nilfs enables write barriers by default and has "nobarrier" mount
option to disable this feature.  But it lacks the complementary option
and has no way to re-enable the feature on remount.

This adds "barrier" option to resolve this imbalance.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This will sync super blocks in turns instead of syncing duplicate
super blocks at the time.  This will help searching valid super root
when super block is written into disk before log is written, which is
happen when barrier-less block devices are unmounted uncleanly.  In
the situation, old super block likely points to valid log.

This patch introduces ns_sbwcount member to the nilfs object and adds
nilfs_sb_will_flip() function; ns_sbwcount counts how many times super
blocks write back to the disk.  And, nilfs_sb_will_flip() decides
whether flipping required or not based on the count of ns_sbwcount to
sync super blocks asymmetrically.

The following functions are also changed:

 - nilfs_prepare_super(): flips super blocks according to the
   argument.  The argument is calculated by nilfs_sb_will_flip()
   function.

 - nilfs_cleanup_super(): sets "clean" flag to both super blocks if
   they point to the same checkpoint.

To update both of super block information, caller of
nilfs_commit_super must set the information on both super blocks.
Signed-off-by: NJiro SEKIBA <jir@unicus.jp>
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This function checks validity of super block pointers.
If first super block is invalid, it will swap the super blocks.
The function should be called before any super block information updates.
Caller must obtain nilfs->ns_sem.
Signed-off-by: NJiro SEKIBA <jir@unicus.jp>
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This moves out section that updates information of the recent log
position stored in super blocks from nilfs_commit_super to a new
routine named nilfs_set_log_cursor.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This function marks error state and write it on super blocks.  This is
a preparation for making super block writeback alternately.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This function write out filesystem state to super blocks in order to
share the same cleanup work.  This is a preparation for making super
block writeback alternately.

Cc: Jiro SEKIBA <jir@unicus.jp>
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

Mount time field in super block is wrongly updated when nilfs remounts
the partition from read-write to read-only.  This fixes the issue.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

This gets rid of unwanted space chars in front of conditional
sentences of nilfs_destroy_cachep().
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>

Snapshots and regular ro/rw mounts are essentially-different within
the meaning whether the checkpoint is static or not and is marked with
a snapshot flag or not.

The current implemenation, however, allows to remount a snapshot to a
regular rw-mount if the checkpoint number equals the latest one.

This transition is actually impossible since changing a checkpoint to
a snapshot makes another checkpoint, thus the condition is never
satisfied.

This fixes the weird state of affairs, and specifically separates
snapshots and regular rw/ro-mounts.
Signed-off-by: NRyusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>