The bug is from running xfstests 209 with autodefrag.

The race is as follows:
       t1                       t2(autodefrag)
   direct IO
     invalidate pagecache
     dio(old data)             add_inode_defrag
     invalidate pagecache
   endio

   direct IO
     invalidate pagecache
                                run_defrag
                                  readpage(old data)
                                  set page dirty (old data)
     dio(new data, rewrite)
     invalidate pagecache (*)
     endio

t2(autodefrag) will get old data into pagecache via readpage and set
pagecache dirty.  Meanwhile, invalidate pagecache(*) will fail due to
dirty flags in pages.  So the old data may be flushed into disk by
flush thread, which will lead to data loss.

And so does the case of user defragment progs.

The patch fixes this race by holding i_mutex when we readpage and set page dirty.
Signed-off-by: NLiu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

This deadlock comes from xfstests 251.

We'll hold the chunk_mutex throughout the whole of a chunk allocation.
But if we find that we've used up system chunk space, we need to allocate a
new system chunk, but this will lead to a recursion of chunk allocation and end
up with a deadlock on chunk_mutex.
So instead we need to allocate the system chunk first if we find we're in ENOSPC.
Signed-off-by: NLiu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

o For space info, the type of space info is useful for debug.
o For transaction handle, its transid is useful.
Signed-off-by: NLiu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

With the big metadata blocks, we can have crc items
that are much bigger than a page.  There are a few
places that we try to kmalloc memory to hold the
items during a split.

Items bigger than 4KB don't really have a huge benefit
in efficiency, but they do trigger larger order allocations.
This commits changes the csums to make sure they stay under
4KB.  This is not a format change, just a #define to limit
huge items.
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Btrfs puts the filesystem metadata into its own address space, and
somehow the block device address space isn't getting onto disk properly
before a mount.  The end result is that a loop of mkfs and mounting the
filesystem will sometimes find stale or incorrect data.

This commit should fix it by sprinkling fdatawrites and invalidate_bdev
calls around.  This is a short term measure to make sure it is fixed.
The block devices really should be flushed and cleaned up higher in the
stack.
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Conflicts:
	fs/btrfs/transaction.c
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Conflicts:
	fs/btrfs/ctree.c
	fs/btrfs/disk-io.c
	fs/btrfs/extent-tree.c
	fs/btrfs/extent_io.c
	fs/btrfs/extent_io.h
	fs/btrfs/inode.c
	fs/btrfs/scrub.c
Signed-off-by: NChris Mason <chris.mason@oracle.com>

With support for bigger metadata blocks, we must avoid mounting a
filesystem with different block size for mixed block groups, this causes
corruption (found by xfstests/083).
Signed-off-by: NDavid Sterba <dsterba@suse.cz>

Validate checksum algorithm during mount and prevent BUG_ON later in
btrfs_super_csum_size.
Signed-off-by: NDavid Sterba <dsterba@suse.cz>

Scrub used to be coded for nodesize == leafsize == sectorsize == PAGE_SIZE.
This is now changed to support sizes for nodesize and leafsize which are
N * PAGE_SIZE.
Signed-off-by: NStefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Just a minor cleanup commit in preparation for the big block changes.
Signed-off-by: NStefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Readahead already has a define for the max number of mirrors. Scrub
needs such a define now, the rest of the code will need something
like this soon. Therefore the define was added to ctree.h and removed
from the readahead code.
Signed-off-by: NStefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

If relocate of block group 0 fails with ENOSPC we end up infinitely
looping because key.offset -= 1 statement in that case brings us back to
where we started.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

init_ipath() allocates btrfs_data_container which is never freed.  Free
it in free_ipath() and nuke the comment for init_data_container() - we
can safely free it with kfree().
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Generally we don't allow dup for data, but mixed chunks are special and
people seem to think this has its use cases.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Do not run sanity checks on all target profiles unless they all will be
used.  This came up because alloc_profile_is_valid() is now more strict
than it used to be.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Currently if we don't have enough space allocated we go ahead and loop
though devices in the hopes of finding enough space for a chunk of the
*same* type as the one we are trying to relocate.  The problem with that
is that if we are trying to restripe the chunk its target type can be
more relaxed than the current one (eg require less devices or less
space).  So, when restriping, run checks against the target profile
instead of the current one.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Add __get_block_group_index() helper to be able to derive block group
index from an arbitary set of flags.  Implement get_block_group_index()
in terms of it.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Add get_restripe_target() helper and switch everybody to use it.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Header file is not a good place to define functions.  This also moves a
call to alloc_profile_is_valid() down the stack and removes a redundant
check from __btrfs_alloc_chunk() - alloc_profile_is_valid() takes it
into account.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

"0" is a valid value for an on-disk chunk profile, but it is not a valid
extended profile.  (We have a separate bit for single chunks in extended
case)

Also rename it to alloc_profile_is_valid() for clarity.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

Add functions to abstract the conversion between chunk and extended
allocation profile formats and switch everybody to use them.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

This has been causing a lot of confusion for quite a while now and a lot
of users were surprised by this (some of them were even stuck in a
ENOSPC situation which they couldn't easily get out of).  The addition
of restriper gives users a clear choice between raid0 and drive concat
setup so there's absolutely no excuse for us to keep doing this.
Signed-off-by: NIlya Dryomov <idryomov@gmail.com>

In commit 4692cf58 we introduced new backref walking code for btrfs. This
assumes we're searching live roots, which requires a transaction context.
While scrubbing, however, we must not join a transaction because this could
deadlock with the commit path. Additionally, what scrub really wants to do
is resolving a logical address in the commit root it's currently checking.

This patch adds support for logical to path resolving on commit roots and
makes scrub use that.
Signed-off-by: NJan Schmidt <list.btrfs@jan-o-sch.net>

The two helper functions commit_cowonly_roots() and
create_pending_snapshot() failed to check the return value from
btrfs_cow_block(), which could at least in theory fail with -ENOSPC from
btrfs_alloc_free_block(). This commit adds the missing checks.
Signed-off-by: NJan Schmidt <list.btrfs@jan-o-sch.net>

btrfs_init_lockdep only makes our lockdep class names look prettier, thus
it did never hurt we forgot to actually call it. This turns our lockdep
identifier strings from lockdep auto-set #[id] into really pretty
"btrfs-fs-01" or "btrfs-csum-03".
Signed-off-by: NJan Schmidt <list.btrfs@jan-o-sch.net>

Since we need to read and write extent buffers in their entirety we can't use
the normal bio_readpage_error stuff since it only works on a per page basis.  So
instead make it so that if we see an io error in endio we just mark the eb as
having an IO error and then in btree_read_extent_buffer_pages we will manually
try other mirrors and then overwrite the bad mirror if we find a good copy.
This works with larger than page size blocks.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

The metadata write IO completion code is now simple enough that we
don't need the threaded helpers anymore.
Signed-off-by: NChris Mason <chris.mason@oracle.com>

btrfs_search_slot sometimes needs write locks on high levels of
the tree.  It remembers the highest level that needs a write lock
and will use that for all future searches through the tree in a given
call.

But, very often we'll just cow the top level or the level below and we
won't really need write locks on the root again after that.  This patch
changes things to adjust the write lock requirement as it unlocks
levels.
Signed-off-by: NChris Mason <chris.mason@oracle.com>

lock_extent_buffer_for_io needs to loop around and make sure the
writeback bits are not set.
Signed-off-by: NChris Mason <chris.mason@oracle.com>

This cuts down on the CPU time used by map_private_extent_buffer
Signed-off-by: NChris Mason <chris.mason@oracle.com>

This patch simplifies how we track our extent buffers.  Previously we could exit
writepages with only having written half of an extent buffer, which meant we had
to track the state of the pages and the state of the extent buffers differently.
Now we only read in entire extent buffers and write out entire extent buffers,
this allows us to simply set bits in our bflags to indicate the state of the eb
and we no longer have to do things like track uptodate with our iotree.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Because an eb can have multiple pages we need to make sure that all pages within
the eb are markes as accessed, since releasepage can be called against any page
in the eb.  This will keep us from possibly evicting hot eb's when we're doing
larger than pagesize eb's.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>

Because btrfs cow's we can end up with extent buffers that are no longer
necessary just sitting around in memory.  So instead of evicting these pages, we
could end up evicting things we actually care about.  Thus we have
free_extent_buffer_stale for use when we are freeing tree blocks.  This will
make it so that the ref for the eb being in the radix tree is dropped as soon as
possible and then is freed when the refcount hits 0 instead of waiting to be
released by releasepage.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>

We can run into a problem where we find an eb for our existing page already on
the radix tree but it has a ref count of 0.  It hasn't yet been removed by RCU
yet so this can cause issues where we will use the EB after free.  So do
atomic_inc_not_zero on the exists->refs and if it is zero just do
synchronize_rcu() and try again.  We won't have to worry about new allocators
coming in since they will block on the page lock at this point.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>

We spend a lot of time looking up extent buffers from pages when we could just
store the pointer to the eb the page is associated with in page->private.  This
patch does just that, and it makes things a little simpler and reduces a bit of
CPU overhead involved with doing metadata IO.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>

A few years ago the btrfs code to support blocks lager than
the page size was disabled to fix a few corner cases in the
page cache handling.  This fixes the code to properly support
large metadata blocks again.

Since current kernels will crash early and often with larger
metadata blocks, this adds an incompat bit so that older kernels
can't mount it.

This also does away with different blocksizes for nodes and leaves.
You get a single block size for all tree blocks.
Signed-off-by: NChris Mason <chris.mason@oracle.com>

We have been passing nothing but (u64)-1 to find_free_extent for search_end in
all of the callers, so it's completely useless, and we've always been passing 0
in as search_start, so just remove them as function arguments and move
search_start into find_free_extent.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>

This is a relic from before we had the disk space cache and it was to make
bootup times when you had btrfs as root not be so damned slow.  Now that we have
the disk space cache this isn't a problem anymore and really having this code
casues uneeded fragmentation and complexity, so just remove it.  Thanks,
Signed-off-by: NJosef Bacik <josef@redhat.com>