openwrt/staging/blogic.git
5 years agobtrfs: refactor the ticket wakeup code
Josef Bacik [Thu, 22 Aug 2019 19:10:58 +0000 (15:10 -0400)]
btrfs: refactor the ticket wakeup code

Now that btrfs_space_info_add_old_bytes simply checks if we can make the
reservation and updates bytes_may_use, there's no reason to have both
helpers in place.

Factor out the ticket wakeup logic into it's own helper, make
btrfs_space_info_add_old_bytes() update bytes_may_use and then call the
wakeup helper, and replace all calls to btrfs_space_info_add_new_bytes()
with the wakeup helper.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: stop partially refilling tickets when releasing space
Josef Bacik [Wed, 28 Aug 2019 15:15:24 +0000 (11:15 -0400)]
btrfs: stop partially refilling tickets when releasing space

btrfs_space_info_add_old_bytes is used when adding the extra space from
an existing reservation back into the space_info to be used by any
waiting tickets.  In order to keep us from overcommitting we check to
make sure that we can still use this space for our reserve ticket, and
if we cannot we'll simply subtract it from space_info->bytes_may_use.

However this is problematic, because it assumes that only changes to
bytes_may_use would affect our ability to make reservations.  Any
changes to bytes_reserved would be missed.  If we were unable to make a
reservation prior because of reserved space, but that reserved space was
free'd due to unlink or truncate and we were allowed to immediately
reclaim that metadata space we would still ENOSPC.

Consider the example where we create a file with a bunch of extents,
using up 2MiB of actual space for the new tree blocks.  Then we try to
make a reservation of 2MiB but we do not have enough space to make this
reservation.  The iput() occurs in another thread and we remove this
space, and since we did not write the blocks we simply do
space_info->bytes_reserved -= 2MiB.  We would never see this because we
do not check our space info used, we just try to re-use the freed
reservations.

To fix this problem, and to greatly simplify the wakeup code, do away
with this partial refilling nonsense.  Use
btrfs_space_info_add_old_bytes to subtract the reservation from
space_info->bytes_may_use, and then check the ticket against the total
used of the space_info the same way we do with the initial reservation
attempt.

This keeps the reservation logic consistent and solves the problem of
early ENOSPC in the case that we free up space in places other than
bytes_may_use and bytes_pinned.  Thanks,

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: add space reservation tracepoint for reserved bytes
Josef Bacik [Thu, 22 Aug 2019 19:10:56 +0000 (15:10 -0400)]
btrfs: add space reservation tracepoint for reserved bytes

I noticed when folding the trace_btrfs_space_reservation() tracepoint
into the btrfs_space_info_update_* helpers that we didn't emit a
tracepoint when doing btrfs_add_reserved_bytes().  I know this is
because we were swapping bytes_may_use for bytes_reserved, so in my mind
there was no reason to have the tracepoint there.  But now there is
because we always emit the unreserve for the bytes_may_use side, and
this would have broken if compression was on anyway.  Add a tracepoint
to cover the bytes_reserved counter so the math still comes out right.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: roll tracepoint into btrfs_space_info_update helper
Josef Bacik [Thu, 22 Aug 2019 19:10:55 +0000 (15:10 -0400)]
btrfs: roll tracepoint into btrfs_space_info_update helper

We duplicate this tracepoint everywhere we call these helpers, so update
the helper to have the tracepoint as well.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: do not allow reservations if we have pending tickets
Josef Bacik [Thu, 22 Aug 2019 19:10:54 +0000 (15:10 -0400)]
btrfs: do not allow reservations if we have pending tickets

If we already have tickets on the list we don't want to steal their
reservations.  This is a preparation patch for upcoming changes,
technically this shouldn't happen today because of the way we add bytes
to tickets before adding them to the space_info in most cases.

This does not change the FIFO nature of reserve tickets, it simply
allows us to enforce it in a different way.  Previously it was enforced
because any new space would be added to the first ticket on the list,
which would result in new reservations getting a reserve ticket.  This
replaces that mechanism by simply checking to see if we have outstanding
reserve tickets and skipping straight to adding a ticket for our
reservation.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: stop clearing EXTENT_DIRTY in inode I/O tree
Omar Sandoval [Thu, 15 Aug 2019 21:04:04 +0000 (14:04 -0700)]
btrfs: stop clearing EXTENT_DIRTY in inode I/O tree

Since commit fee187d9d9dd ("Btrfs: do not set EXTENT_DIRTY along with
EXTENT_DELALLOC"), we never set EXTENT_DIRTY in inode->io_tree, so we
can simplify and stop trying to clear it.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: treat RWF_{,D}SYNC writes as sync for CRCs
Omar Sandoval [Thu, 15 Aug 2019 21:04:03 +0000 (14:04 -0700)]
btrfs: treat RWF_{,D}SYNC writes as sync for CRCs

The VFS indicates a synchronous write to ->write_iter() via
iocb->ki_flags. The IOCB_{,D}SYNC flags may be set based on the file
(see iocb_flags()) or the RWF_* flags passed to a syscall like
pwritev2() (see kiocb_set_rw_flags()).

However, in btrfs_file_write_iter(), we're checking if a write is
synchronous based only on the file; we use this to decide when to bump
the sync_writers counter and thus do CRCs synchronously. Make sure we do
this for all synchronous writes as determined by the VFS.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ add const ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: use correct count in btrfs_file_write_iter()
Omar Sandoval [Thu, 15 Aug 2019 21:04:02 +0000 (14:04 -0700)]
btrfs: use correct count in btrfs_file_write_iter()

generic_write_checks() may modify iov_iter_count(), so we must get the
count after the call, not before. Using the wrong one has a couple of
consequences:

1. We check a longer range in check_can_nocow() for nowait than we're
   actually writing.
2. We create extra hole extent maps in btrfs_cont_expand(). As far as I
   can tell, this is harmless, but I might be missing something.

These issues are pretty minor, but let's fix it before something more
important trips on it.

Fixes: edf064e7c6fe ("btrfs: nowait aio support")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tie extent buffer and it's token together
David Sterba [Fri, 9 Aug 2019 15:48:21 +0000 (17:48 +0200)]
btrfs: tie extent buffer and it's token together

Further simplifaction of the get/set helpers is possible when the token
is uniquely tied to an extent buffer. A condition and an assignment can
be avoided.

The initializations are moved closer to the first use when the extent
buffer is valid. There's one exception in __push_leaf_left where the
token is reused.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: assume valid token for btrfs_set/get_token helpers
David Sterba [Fri, 9 Aug 2019 15:30:23 +0000 (17:30 +0200)]
btrfs: assume valid token for btrfs_set/get_token helpers

Now that we can safely assume that the token is always a valid pointer,
remove the branches that check that.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: define separate btrfs_set/get_XX helpers
David Sterba [Fri, 9 Aug 2019 15:12:38 +0000 (17:12 +0200)]
btrfs: define separate btrfs_set/get_XX helpers

There are helpers for all type widths defined via macro and optionally
can use a token which is a cached pointer to avoid repeated mapping of
the extent buffer.

The token value is known at compile time, when it's valid it's always
address of a local variable, otherwise it's NULL passed by the
token-less helpers.

This can be utilized to remove some branching as the helpers are used
frequenlty.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Make btrfs_find_name_in_ext_backref return struct btrfs_inode_extref
Nikolay Borisov [Tue, 27 Aug 2019 11:46:29 +0000 (14:46 +0300)]
btrfs: Make btrfs_find_name_in_ext_backref return struct btrfs_inode_extref

btrfs_find_name_in_ext_backref returns either 0/1 depending on whether it
found a backref for the given name. If it returns true then the actual
inode_ref struct is returned in one of its parameters. That's pointless,
instead refactor the function such that it returns either a pointer
to the btrfs_inode_extref or NULL it it didn't find anything. This
streamlines the function calling convention.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Make btrfs_find_name_in_backref return btrfs_inode_ref struct
Nikolay Borisov [Tue, 27 Aug 2019 11:46:28 +0000 (14:46 +0300)]
btrfs: Make btrfs_find_name_in_backref return btrfs_inode_ref struct

btrfs_find_name_in_backref returns either 0/1 depending on whether it
found a backref for the given name. If it returns true then the actual
inode_ref struct is returned in one of its parameters. That's pointless,
instead refactor the function such that it returns either a pointer
to the btrfs_inode_ref or NULL it it didn't find anything. This
streamlines the function calling convention.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move dev_stats helpers to volumes.c
David Sterba [Wed, 21 Aug 2019 18:05:32 +0000 (20:05 +0200)]
btrfs: move dev_stats helpers to volumes.c

The other dev stats functions are already there and the helpers are not
used by anything else.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move struct io_ctl to free-space-cache.h
David Sterba [Wed, 21 Aug 2019 17:57:04 +0000 (19:57 +0200)]
btrfs: move struct io_ctl to free-space-cache.h

The io_ctl structure is used for free space management, and used only by
the v1 space cache code, but unfortunatlly the full definition is
required by block-group.h so it can't be moved to free-space-cache.c
without additional changes.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move functions for tree compare to send.c
David Sterba [Wed, 21 Aug 2019 17:12:59 +0000 (19:12 +0200)]
btrfs: move functions for tree compare to send.c

Send is the only user of tree_compare, we can move it there along with
the other helpers and definitions.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: rename and export read_node_slot
David Sterba [Wed, 21 Aug 2019 17:16:27 +0000 (19:16 +0200)]
btrfs: rename and export read_node_slot

Preparatory work for code that will be moved out of ctree and uses this
function.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move private raid56 definitions from ctree.h
David Sterba [Wed, 21 Aug 2019 17:06:17 +0000 (19:06 +0200)]
btrfs: move private raid56 definitions from ctree.h

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move math functions to misc.h
David Sterba [Wed, 21 Aug 2019 16:54:28 +0000 (18:54 +0200)]
btrfs: move math functions to misc.h

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move cond_wake_up functions out of ctree
David Sterba [Wed, 21 Aug 2019 16:48:25 +0000 (18:48 +0200)]
btrfs: move cond_wake_up functions out of ctree

The file ctree.h serves as a header for everything and has become quite
bloated. Split some helpers that are generic and create a new file that
should be the catch-all for code that's not btrfs-specific.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: use proper error values on allocation failure in clone_fs_devices
Anand Jain [Tue, 27 Aug 2019 07:40:45 +0000 (15:40 +0800)]
btrfs: use proper error values on allocation failure in clone_fs_devices

Fix the fake ENOMEM return error code to the actual error in
clone_fs_devices().

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: proper error handling when invalid device is found in find_next_devid
Anand Jain [Tue, 27 Aug 2019 07:40:44 +0000 (15:40 +0800)]
btrfs: proper error handling when invalid device is found in find_next_devid

In a corrupted tree, if search for next devid finds the device with
devid = -1, then report the error -EUCLEAN back to the parent function
to fail gracefully.

The tree checker will not catch this in case the devids are created
using the following script:

  umount /btrfs
  dev1=/dev/sdb
  dev2=/dev/sdc
  mkfs.btrfs -fq -dsingle -msingle $dev1
  mount $dev1 /btrfs

  _fail()
  {
  echo $1
  exit 1
  }

  while true; do
  btrfs dev add -f $dev2 /btrfs || _fail "add failed"
  btrfs dev del $dev1 /btrfs || _fail "del failed"
  dev_tmp=$dev1
  dev1=$dev2
  dev2=$dev_tmp
  done

With output:

  BTRFS critical (device sdb): corrupt leaf: root=3 block=313739198464 slot=1 devid=1 invalid devid: has=507 expect=[0, 506]
  BTRFS error (device sdb): block=313739198464 write time tree block corruption detected
  BTRFS: error (device sdb) in btrfs_commit_transaction:2268: errno=-5 IO failure (Error while writing out transaction)
  BTRFS warning (device sdb): Skipping commit of aborted transaction.
  BTRFS: error (device sdb) in cleanup_transaction:1827: errno=-5 IO failure

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
[ add script and messages ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: fix allocation of free space cache v1 bitmap pages
Christophe Leroy [Wed, 21 Aug 2019 15:05:55 +0000 (15:05 +0000)]
btrfs: fix allocation of free space cache v1 bitmap pages

Various notifications of type "BUG kmalloc-4096 () : Redzone
overwritten" have been observed recently in various parts of the kernel.
After some time, it has been made a relation with the use of BTRFS
filesystem and with SLUB_DEBUG turned on.

[   22.809700] BUG kmalloc-4096 (Tainted: G        W        ): Redzone overwritten

[   22.810286] INFO: 0xbe1a5921-0xfbfc06cd. First byte 0x0 instead of 0xcc
[   22.810866] INFO: Allocated in __load_free_space_cache+0x588/0x780 [btrfs] age=22 cpu=0 pid=224
[   22.811193]  __slab_alloc.constprop.26+0x44/0x70
[   22.811345]  kmem_cache_alloc_trace+0xf0/0x2ec
[   22.811588]  __load_free_space_cache+0x588/0x780 [btrfs]
[   22.811848]  load_free_space_cache+0xf4/0x1b0 [btrfs]
[   22.812090]  cache_block_group+0x1d0/0x3d0 [btrfs]
[   22.812321]  find_free_extent+0x680/0x12a4 [btrfs]
[   22.812549]  btrfs_reserve_extent+0xec/0x220 [btrfs]
[   22.812785]  btrfs_alloc_tree_block+0x178/0x5f4 [btrfs]
[   22.813032]  __btrfs_cow_block+0x150/0x5d4 [btrfs]
[   22.813262]  btrfs_cow_block+0x194/0x298 [btrfs]
[   22.813484]  commit_cowonly_roots+0x44/0x294 [btrfs]
[   22.813718]  btrfs_commit_transaction+0x63c/0xc0c [btrfs]
[   22.813973]  close_ctree+0xf8/0x2a4 [btrfs]
[   22.814107]  generic_shutdown_super+0x80/0x110
[   22.814250]  kill_anon_super+0x18/0x30
[   22.814437]  btrfs_kill_super+0x18/0x90 [btrfs]
[   22.814590] INFO: Freed in proc_cgroup_show+0xc0/0x248 age=41 cpu=0 pid=83
[   22.814841]  proc_cgroup_show+0xc0/0x248
[   22.814967]  proc_single_show+0x54/0x98
[   22.815086]  seq_read+0x278/0x45c
[   22.815190]  __vfs_read+0x28/0x17c
[   22.815289]  vfs_read+0xa8/0x14c
[   22.815381]  ksys_read+0x50/0x94
[   22.815475]  ret_from_syscall+0x0/0x38

Commit 69d2480456d1 ("btrfs: use copy_page for copying pages instead of
memcpy") changed the way bitmap blocks are copied. But allthough bitmaps
have the size of a page, they were allocated with kzalloc().

Most of the time, kzalloc() allocates aligned blocks of memory, so
copy_page() can be used. But when some debug options like SLAB_DEBUG are
activated, kzalloc() may return unaligned pointer.

On powerpc, memcpy(), copy_page() and other copying functions use
'dcbz' instruction which provides an entire zeroed cacheline to avoid
memory read when the intention is to overwrite a full line. Functions
like memcpy() are writen to care about partial cachelines at the start
and end of the destination, but copy_page() assumes it gets pages. As
pages are naturally cache aligned, copy_page() doesn't care about
partial lines. This means that when copy_page() is called with a
misaligned pointer, a few leading bytes are zeroed.

To fix it, allocate bitmaps through kmem_cache instead of using kzalloc()
The cache pool is created with PAGE_SIZE alignment constraint.

Reported-by: Erhard F. <erhard_f@mailbox.org>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=204371
Fixes: 69d2480456d1 ("btrfs: use copy_page for copying pages instead of memcpy")
Cc: stable@vger.kernel.org # 4.19+
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Reviewed-by: David Sterba <dsterba@suse.com>
[ rename to btrfs_free_space_bitmap ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Detect unbalanced tree with empty leaf before crashing btree operations
Qu Wenruo [Thu, 22 Aug 2019 02:14:15 +0000 (10:14 +0800)]
btrfs: Detect unbalanced tree with empty leaf before crashing btree operations

[BUG]
With crafted image, btrfs will panic at btree operations:

  kernel BUG at fs/btrfs/ctree.c:3894!
  invalid opcode: 0000 [#1] SMP PTI
  CPU: 0 PID: 1138 Comm: btrfs-transacti Not tainted 5.0.0-rc8+ #9
  RIP: 0010:__push_leaf_left+0x6b6/0x6e0
  RSP: 0018:ffffc0bd4128b990 EFLAGS: 00010246
  RAX: 0000000000000000 RBX: ffffa0a4ab8f0e38 RCX: 0000000000000000
  RDX: ffffa0a280000000 RSI: 0000000000000000 RDI: ffffa0a4b3814000
  RBP: ffffc0bd4128ba38 R08: 0000000000001000 R09: ffffc0bd4128b948
  R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000240
  R13: ffffa0a4b556fb60 R14: ffffa0a4ab8f0af0 R15: ffffa0a4ab8f0af0
  FS: 0000000000000000(0000) GS:ffffa0a4b7a00000(0000) knlGS:0000000000000000
  CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007f2461c80020 CR3: 000000022b32a006 CR4: 00000000000206f0
  Call Trace:
  ? _cond_resched+0x1a/0x50
  push_leaf_left+0x179/0x190
  btrfs_del_items+0x316/0x470
  btrfs_del_csums+0x215/0x3a0
  __btrfs_free_extent.isra.72+0x5a7/0xbe0
  __btrfs_run_delayed_refs+0x539/0x1120
  btrfs_run_delayed_refs+0xdb/0x1b0
  btrfs_commit_transaction+0x52/0x950
  ? start_transaction+0x94/0x450
  transaction_kthread+0x163/0x190
  kthread+0x105/0x140
  ? btrfs_cleanup_transaction+0x560/0x560
  ? kthread_destroy_worker+0x50/0x50
  ret_from_fork+0x35/0x40
  Modules linked in:
  ---[ end trace c2425e6e89b5558f ]---

[CAUSE]
The offending csum tree looks like this:

  checksum tree key (CSUM_TREE ROOT_ITEM 0)
  node 29741056 level 1 items 14 free 107 generation 19 owner CSUM_TREE
  ...
  key (EXTENT_CSUM EXTENT_CSUM 85975040) block 29630464 gen 17
  key (EXTENT_CSUM EXTENT_CSUM 89911296) block 29642752 gen 17 <<<
  key (EXTENT_CSUM EXTENT_CSUM 92274688) block 29646848 gen 17
  ...

  leaf 29630464 items 6 free space 1 generation 17 owner CSUM_TREE
  item 0 key (EXTENT_CSUM EXTENT_CSUM 85975040) itemoff 3987 itemsize 8
  range start 85975040 end 85983232 length 8192
  ...
  leaf 29642752 items 0 free space 3995 generation 17 owner 0
      ^ empty leaf            invalid owner ^

  leaf 29646848 items 1 free space 602 generation 17 owner CSUM_TREE
  item 0 key (EXTENT_CSUM EXTENT_CSUM 92274688) itemoff 627 itemsize 3368
  range start 92274688 end 95723520 length 3448832

So we have a corrupted csum tree where one tree leaf is completely
empty, causing unbalanced btree, thus leading to unexpected btree
balance error.

[FIX]
For this particular case, we handle it in two directions to catch it:
- Check if the tree block is empty through btrfs_verify_level_key()
  So that invalid tree blocks won't be read out through
  btrfs_search_slot() and its variants.

- Check 0 tree owner in tree checker
  NO tree is using 0 as its tree owner, detect it and reject at tree
  block read time.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=202821
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Deprecate BTRFS_SUBVOL_CREATE_ASYNC flag
Nikolay Borisov [Mon, 26 Aug 2019 14:34:24 +0000 (17:34 +0300)]
btrfs: Deprecate BTRFS_SUBVOL_CREATE_ASYNC flag

Support for asynchronous snapshot creation was originally added in
72fd032e9424 ("Btrfs: add SNAP_CREATE_ASYNC ioctl") to cater for
ceph's backend needs. However, since Ceph has deprecated support for
btrfs there is no longer need for that support in btrfs. Additionally,
this was never supported by btrfs-progs, the official userspace tools.

Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: improve error handling in run_delalloc_nocow
Nikolay Borisov [Thu, 22 Aug 2019 14:24:20 +0000 (17:24 +0300)]
btrfs: improve error handling in run_delalloc_nocow

Correctly handle failure cases when adding an ordered extents in case
of REGULAR or PREALLOC extents. Remove the BUG_ON.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: comment and minor simplifications in run_delalloc_nocow
Nikolay Borisov [Thu, 22 Aug 2019 14:25:23 +0000 (17:25 +0300)]
btrfs: comment and minor simplifications in run_delalloc_nocow

Add a comment explaining why we keep the BUG also use the already read
and cached value of extent ram bytes stored in 'ram_bytes'.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: streamline code in run_delalloc_nocow in case of inline extents
Nikolay Borisov [Mon, 5 Aug 2019 14:47:06 +0000 (17:47 +0300)]
btrfs: streamline code in run_delalloc_nocow in case of inline extents

The extent range check right after the "out_check" label is redundant,
because the only way it can trigger is if we have an inline extent. In
this case it makes more sense to actually move it in the branch
explictly dealing with inlines extents.

What's more, the nested 'if (nocow)' can never be true because for
inline extents we always do COW and there is no chance 'nocow' can be
true, just remove that check.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: simplify extent type checks in run_delalloc_nocow
Nikolay Borisov [Mon, 5 Aug 2019 14:47:05 +0000 (17:47 +0300)]
btrfs: simplify extent type checks in run_delalloc_nocow

There is no point in checking the type of the extent again just to set
the 'type' variable, when this check has already been performed before.
Instead, extend the original if branch with an 'else' clause. This
allows to remove one local variable and make it obvious how the code
flow differs for prealloc/regular extents.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: improve comments around nocow path
Nikolay Borisov [Wed, 21 Aug 2019 07:42:57 +0000 (10:42 +0300)]
btrfs: improve comments around nocow path

run_delalloc_nocow contains numerous, somewhat subtle, checks when
figuring out whether a particular extent should be CoW'ed or not. This
patch explicitly states the assumptions those checks verify. As a
result also document 2 of the more subtle checks in check_committed_ref
as well.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: refactor variable scope in run_delalloc_nocow
Nikolay Borisov [Wed, 21 Aug 2019 07:42:03 +0000 (10:42 +0300)]
btrfs: refactor variable scope in run_delalloc_nocow

Of the 22 (!!!) local variables declared in this function only 9 have
function-wide context. Of the remaining 13, 12 are needed in the main
while loop of the function and 1 is needed in a tiny if branch, only in
case we have prealloc extent. This commit reduces the lifespan of every
variable to its bare minimum. It also renames the 'nolock' boolean to
freespace_inode to clearly indicate its purpose.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: only reserve metadata_size for inodes
Josef Bacik [Thu, 22 Aug 2019 19:14:34 +0000 (15:14 -0400)]
btrfs: only reserve metadata_size for inodes

Historically we reserved worst case for every btree operation, and
generally speaking we want to do that in cases where it could be the
worst case.  However for updating inodes we know the inode items are
already in the tree, so it will only be an update operation and never an
insert operation.  This allows us to always reserve only the
metadata_size amount for inode updates rather than the
insert_metadata_size amount.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: rename the btrfs_calc_*_metadata_size helpers
Josef Bacik [Thu, 22 Aug 2019 19:14:33 +0000 (15:14 -0400)]
btrfs: rename the btrfs_calc_*_metadata_size helpers

btrfs_calc_trunc_metadata_size differs from trans_metadata_size in that
it doesn't take into account any splitting at the levels, because
truncate will never split nodes.  However truncate _and_ changing will
never split nodes, so rename btrfs_calc_trunc_metadata_size to
btrfs_calc_metadata_size.  Also btrfs_calc_trans_metadata_size is purely
for inserting items, so rename this to btrfs_calc_insert_metadata_size.
Making these clearer will help when I start using them differently in
upcoming patches.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tree-checker: Add EXTENT_DATA_REF check
Qu Wenruo [Fri, 9 Aug 2019 01:24:24 +0000 (09:24 +0800)]
btrfs: tree-checker: Add EXTENT_DATA_REF check

EXTENT_DATA_REF is a little like DIR_ITEM which contains hash in its
key->offset.

This patch will check the following contents:
- Key->objectid
  Basic alignment check.

- Hash
  Hash of each extent_data_ref item must match key->offset.

- Offset
  Basic alignment check.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tree-checker: Add simple keyed refs check
Qu Wenruo [Fri, 9 Aug 2019 01:24:23 +0000 (09:24 +0800)]
btrfs: tree-checker: Add simple keyed refs check

For TREE_BLOCK_REF, SHARED_DATA_REF and SHARED_BLOCK_REF we need to
check:
              | TREE_BLOCK_REF | SHARED_BLOCK_REF | SHARED_BLOCK_REF
--------------+----------------+-----------------+------------------
key->objectid |    Alignment   |     Alignment    |    Alignment
key->offset   |    Any value   |     Alignment    |    Alignment
item_size     |        0       |        0         |   sizeof(le32) (*)

*: sizeof(struct btrfs_shared_data_ref)

So introduce a check to check all these 3 key types together.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tree-checker: Add EXTENT_ITEM and METADATA_ITEM check
Qu Wenruo [Fri, 9 Aug 2019 01:24:22 +0000 (09:24 +0800)]
btrfs: tree-checker: Add EXTENT_ITEM and METADATA_ITEM check

This patch introduces the ability to check extent items.

This check involves:
- key->objectid check
  Basic alignment check.

- key->type check
  Against btrfs_extent_item::type and SKINNY_METADATA feature.

- key->offset alignment check for EXTENT_ITEM

- key->offset check for METADATA_ITEM

- item size check
  Both against minimal size and stepping check.

- btrfs_extent_item check
  Checks its flags and generation.

- btrfs_extent_inline_ref checks
  Against 4 types inline ref.
  Checks bytenr alignment and tree level.

- btrfs_extent_item::refs check
  Check against total refs found in inline refs.

This check would be the most complex single item check due to its nature
of inlined items.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: fix error pointer check in __btrfs_map_block()
Dan Carpenter [Fri, 9 Aug 2019 14:07:39 +0000 (17:07 +0300)]
btrfs: fix error pointer check in __btrfs_map_block()

The btrfs_get_chunk_map() never returns NULL, it returns error pointers.

Fixes: 89b798ad1b42 ("btrfs: Use btrfs_get_io_geometry appropriately")
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: dev stat drop useless goto
Anand Jain [Wed, 21 Aug 2019 09:26:32 +0000 (17:26 +0800)]
btrfs: dev stat drop useless goto

In the function btrfs_init_dev_stats() goto out is not needed, because the
alloc has failed. So just return -ENOMEM.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: dev stats item key conversion per cpu type is not needed
Anand Jain [Wed, 21 Aug 2019 09:26:33 +0000 (17:26 +0800)]
btrfs: dev stats item key conversion per cpu type is not needed

%found_key is not used, drop it since it hasn't been used since the
beginning in 733f4fbbc108 ("Btrfs: read device stats on mount, write
modified ones during commit").

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Make reada_tree_block_flagged private
Nikolay Borisov [Wed, 21 Aug 2019 13:38:15 +0000 (16:38 +0300)]
btrfs: Make reada_tree_block_flagged private

This function is used only for the readahead machinery. It makes no
sense to keep it external to reada.c file. Place it above its sole
caller and make it static. No functional changes.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: compression: replace set_level callbacks by a common helper
David Sterba [Fri, 9 Aug 2019 14:49:06 +0000 (16:49 +0200)]
btrfs: compression: replace set_level callbacks by a common helper

The set_level callbacks do not do anything special and can be replaced
by a helper that uses the levels defined in the tables.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: define compression levels statically
David Sterba [Fri, 9 Aug 2019 14:25:34 +0000 (16:25 +0200)]
btrfs: define compression levels statically

The maximum and default levels do not change and can be defined
directly. The set_level callback was a temporary solution and will be
removed.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agoBtrfs: fix use-after-free when using the tree modification log
Filipe Manana [Mon, 12 Aug 2019 18:14:29 +0000 (19:14 +0100)]
Btrfs: fix use-after-free when using the tree modification log

At ctree.c:get_old_root(), we are accessing a root's header owner field
after we have freed the respective extent buffer. This results in an
use-after-free that can lead to crashes, and when CONFIG_DEBUG_PAGEALLOC
is set, results in a stack trace like the following:

  [ 3876.799331] stack segment: 0000 [#1] SMP DEBUG_PAGEALLOC PTI
  [ 3876.799363] CPU: 0 PID: 15436 Comm: pool Not tainted 5.3.0-rc3-btrfs-next-54 #1
  [ 3876.799385] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-0-ga698c8995f-prebuilt.qemu.org 04/01/2014
  [ 3876.799433] RIP: 0010:btrfs_search_old_slot+0x652/0xd80 [btrfs]
  (...)
  [ 3876.799502] RSP: 0018:ffff9f08c1a2f9f0 EFLAGS: 00010286
  [ 3876.799518] RAX: ffff8dd300000000 RBX: ffff8dd85a7a9348 RCX: 000000038da26000
  [ 3876.799538] RDX: 0000000000000000 RSI: ffffe522ce368980 RDI: 0000000000000246
  [ 3876.799559] RBP: dae1922adadad000 R08: 0000000008020000 R09: ffffe522c0000000
  [ 3876.799579] R10: ffff8dd57fd788c8 R11: 000000007511b030 R12: ffff8dd781ddc000
  [ 3876.799599] R13: ffff8dd9e6240578 R14: ffff8dd6896f7a88 R15: ffff8dd688cf90b8
  [ 3876.799620] FS:  00007f23ddd97700(0000) GS:ffff8dda20200000(0000) knlGS:0000000000000000
  [ 3876.799643] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  [ 3876.799660] CR2: 00007f23d4024000 CR3: 0000000710bb0005 CR4: 00000000003606f0
  [ 3876.799682] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  [ 3876.799703] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  [ 3876.799723] Call Trace:
  [ 3876.799735]  ? do_raw_spin_unlock+0x49/0xc0
  [ 3876.799749]  ? _raw_spin_unlock+0x24/0x30
  [ 3876.799779]  resolve_indirect_refs+0x1eb/0xc80 [btrfs]
  [ 3876.799810]  find_parent_nodes+0x38d/0x1180 [btrfs]
  [ 3876.799841]  btrfs_check_shared+0x11a/0x1d0 [btrfs]
  [ 3876.799870]  ? extent_fiemap+0x598/0x6e0 [btrfs]
  [ 3876.799895]  extent_fiemap+0x598/0x6e0 [btrfs]
  [ 3876.799913]  do_vfs_ioctl+0x45a/0x700
  [ 3876.799926]  ksys_ioctl+0x70/0x80
  [ 3876.799938]  ? trace_hardirqs_off_thunk+0x1a/0x20
  [ 3876.799953]  __x64_sys_ioctl+0x16/0x20
  [ 3876.799965]  do_syscall_64+0x62/0x220
  [ 3876.799977]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
  [ 3876.799993] RIP: 0033:0x7f23e0013dd7
  (...)
  [ 3876.800056] RSP: 002b:00007f23ddd96ca8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
  [ 3876.800078] RAX: ffffffffffffffda RBX: 00007f23d80210f8 RCX: 00007f23e0013dd7
  [ 3876.800099] RDX: 00007f23d80210f8 RSI: 00000000c020660b RDI: 0000000000000003
  [ 3876.800626] RBP: 000055fa2a2a2440 R08: 0000000000000000 R09: 00007f23ddd96d7c
  [ 3876.801143] R10: 00007f23d8022000 R11: 0000000000000246 R12: 00007f23ddd96d80
  [ 3876.801662] R13: 00007f23ddd96d78 R14: 00007f23d80210f0 R15: 00007f23ddd96d80
  (...)
  [ 3876.805107] ---[ end trace e53161e179ef04f9 ]---

Fix that by saving the root's header owner field into a local variable
before freeing the root's extent buffer, and then use that local variable
when needed.

Fixes: 30b0463a9394d9 ("Btrfs: fix accessing the root pointer in tree mod log functions")
CC: stable@vger.kernel.org # 3.10+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: replace: BTRFS_DEV_REPLACE_ITEM_STATE_x defines should go
Anand Jain [Thu, 8 Aug 2019 04:32:44 +0000 (12:32 +0800)]
btrfs: replace: BTRFS_DEV_REPLACE_ITEM_STATE_x defines should go

The BTRFS_DEV_REPLACE_ITEM_STATE_x defines, as shown in [1], are
unused in both kernel and btrfs-progs (except for one instance of
BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED in kernel).

[1]
btrfs.h:#define BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED        2
btrfs.h:#define BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED        3
btrfs.h:#define BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED       4

Further these define-values are different form its counterpart
BTRFS_IOCTL_DEV_REPLACE_STATE_x series as shown in [2].

[2]
btrfs_tree.h:#define BTRFS_DEV_REPLACE_ITEM_STATE_SUSPENDED   2
btrfs_tree.h:#define BTRFS_DEV_REPLACE_ITEM_STATE_FINISHED    3
btrfs_tree.h:#define BTRFS_DEV_REPLACE_ITEM_STATE_CANCELED    4

So this patch deletes the BTRFS_DEV_REPLACE_ITEM_STATE_x altogether, and
one instance of BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED is replaced
with BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED in the kernel.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: introduce an evict flushing state
Josef Bacik [Thu, 1 Aug 2019 22:19:37 +0000 (18:19 -0400)]
btrfs: introduce an evict flushing state

We have this weird space flushing loop inside inode.c for evict where
we'll do the normal LIMIT flush, and then commit the transaction and
hope we get our space.  This is super janky, and in fact there's really
nothing stopping us from using FLUSH_ALL except that we run delayed
iputs, which means we could deadlock.  So introduce a new flush state
for eviction that does the normal priority flushing with all of the
states that are safe for eviction.

The nice side-effect of this is that we'll try harder for evictions.
Previously if (for example generic/269) you had a bunch of other
operations happening on the fs you could race with those reservations
when committing the transaction, and eventually miss getting a
reservation for the evict.  With this code we'll have our ticket in
place through the transaction commit, so any pinned bytes will go to our
pending evictions first.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: refactor priority_reclaim_metadata_space
Josef Bacik [Thu, 1 Aug 2019 22:19:36 +0000 (18:19 -0400)]
btrfs: refactor priority_reclaim_metadata_space

With the eviction flushing stuff we'll want to allow for different
states, but still work basically the same way that
priority_reclaim_metadata_space works currently.  Refactor this to take
the flushing states and size as an argument so we can use the same logic
for limit flushing and eviction flushing.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor out the ticket flush handling
Josef Bacik [Thu, 1 Aug 2019 22:19:35 +0000 (18:19 -0400)]
btrfs: factor out the ticket flush handling

We're going to make this logic a little more complicated for evict, so
factor the ticket flushing/waiting code out of __reserve_metadata_bytes.
This has no functional change.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: unify error handling for ticket flushing
Josef Bacik [Thu, 1 Aug 2019 22:19:34 +0000 (18:19 -0400)]
btrfs: unify error handling for ticket flushing

Currently we handle the cleanup of errored out tickets in both the
priority flush path and the normal flushing path.  This is the same code
in both places, so just refactor so we don't duplicate the cleanup work.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: add a flush step for delayed iputs
Josef Bacik [Thu, 1 Aug 2019 22:19:33 +0000 (18:19 -0400)]
btrfs: add a flush step for delayed iputs

Delayed iputs could very well free up enough space without needing to
commit the transaction, so make this step it's own step.  This will
allow us to skip the step for evictions in a later patch.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: unexport the temporary exported functions
Josef Bacik [Thu, 20 Jun 2019 19:38:07 +0000 (15:38 -0400)]
btrfs: unexport the temporary exported functions

These were renamed and exported to facilitate logical migration of
different code chunks into block-group.c.  Now that all the users are in
one file go ahead and rename them back, move the code around, and make
them static.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group cleanup code
Josef Bacik [Thu, 20 Jun 2019 19:38:06 +0000 (15:38 -0400)]
btrfs: migrate the block group cleanup code

This can now be easily migrated as well.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ refresh on top of sysfs cleanups ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the alloc_profile helpers
Josef Bacik [Thu, 20 Jun 2019 19:38:05 +0000 (15:38 -0400)]
btrfs: migrate the alloc_profile helpers

These feel more at home in block-group.c.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ refresh, adjust btrfs_get_alloc_profile exports ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the chunk allocation code
Josef Bacik [Thu, 20 Jun 2019 19:38:04 +0000 (15:38 -0400)]
btrfs: migrate the chunk allocation code

This feels more at home in block-group.c than in extent-tree.c.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>i
[ refresh ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group space accounting helpers
Josef Bacik [Thu, 20 Jun 2019 19:38:02 +0000 (15:38 -0400)]
btrfs: migrate the block group space accounting helpers

We can now easily migrate this code as well.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: export block group accounting helpers
Josef Bacik [Thu, 20 Jun 2019 19:38:01 +0000 (15:38 -0400)]
btrfs: export block group accounting helpers

Want to move these functions into block-group.c, so export them.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the dirty bg writeout code
Josef Bacik [Thu, 20 Jun 2019 19:38:00 +0000 (15:38 -0400)]
btrfs: migrate the dirty bg writeout code

This can be easily migrated over now.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update comments ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate inc/dec_block_group_ro code
Josef Bacik [Thu, 20 Jun 2019 19:37:59 +0000 (15:37 -0400)]
btrfs: migrate inc/dec_block_group_ro code

This can easily be moved now.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ refresh ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: temporarily export btrfs_get_restripe_target
Josef Bacik [Thu, 20 Jun 2019 19:37:58 +0000 (15:37 -0400)]
btrfs: temporarily export btrfs_get_restripe_target

This gets used by a few different logical chunks of the block group
code, export it while we move things around.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group read/creation code
Josef Bacik [Thu, 20 Jun 2019 19:37:57 +0000 (15:37 -0400)]
btrfs: migrate the block group read/creation code

All of the prep work has been done so we can now cleanly move this chunk
over.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ refresh, add btrfs_get_alloc_profile export, comment updates ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group removal code
Josef Bacik [Thu, 20 Jun 2019 19:37:55 +0000 (15:37 -0400)]
btrfs: migrate the block group removal code

This is the removal code and the unused bgs code.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ refresh, move clear_incompat_bg_bits ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: temporarily export inc_block_group_ro
Josef Bacik [Thu, 20 Jun 2019 19:37:54 +0000 (15:37 -0400)]
btrfs: temporarily export inc_block_group_ro

This is used in a few logical parts of the block group code, temporarily
export it so we can move things in pieces.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group caching code
Josef Bacik [Tue, 6 Aug 2019 14:43:19 +0000 (16:43 +0200)]
btrfs: migrate the block group caching code

We can now just copy it over to block-group.c.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: sysfs: move helper macros to sysfs.c
David Sterba [Fri, 2 Aug 2019 10:52:48 +0000 (12:52 +0200)]
btrfs: sysfs: move helper macros to sysfs.c

None of the macros is used outside of sysfs.c.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: sysfs: move type conversion helpers to sysfs.c
David Sterba [Fri, 2 Aug 2019 11:07:38 +0000 (13:07 +0200)]
btrfs: sysfs: move type conversion helpers to sysfs.c

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: cleanup kobject.h includes
David Sterba [Thu, 1 Aug 2019 17:46:20 +0000 (19:46 +0200)]
btrfs: cleanup kobject.h includes

The kobject should be pulled in via sysfs.h and that needs to include it
because it needs various definitions like kobj_attribute or kobject.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor out sysfs code for updating sprout fsid
David Sterba [Thu, 1 Aug 2019 16:50:16 +0000 (18:50 +0200)]
btrfs: factor out sysfs code for updating sprout fsid

Wrap the fsid renaming code and move it to sysfs.c.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor out sysfs code for deleting block group and space infos
David Sterba [Thu, 1 Aug 2019 16:50:16 +0000 (18:50 +0200)]
btrfs: factor out sysfs code for deleting block group and space infos

The helpers to create block group and space info directories already
live in sysfs.c, move the deletion part there too.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor out sysfs code for sending device uevent
David Sterba [Thu, 1 Aug 2019 16:50:16 +0000 (18:50 +0200)]
btrfs: factor out sysfs code for sending device uevent

The device uevent belongs to the sysfs API.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: sysfs: replace direct access to feature set names with a helper
David Sterba [Thu, 1 Aug 2019 17:07:55 +0000 (19:07 +0200)]
btrfs: sysfs: replace direct access to feature set names with a helper

In order to unexport the feature type array, add a helper for the
enum-to-string conversion.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: sysfs: unexport space_info_ktype
David Sterba [Thu, 1 Aug 2019 15:55:55 +0000 (17:55 +0200)]
btrfs: sysfs: unexport space_info_ktype

The last non-sysfs usage of space_info_ktype has been moved to a private
helper in previous patch so the variable can be made static.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor out sysfs code for creating space infos
David Sterba [Thu, 1 Aug 2019 16:50:16 +0000 (18:50 +0200)]
btrfs: factor out sysfs code for creating space infos

Move creation of data/metadata/system space info directories to sysfs.c.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: sysfs: unexport btrfs_raid_ktype
David Sterba [Thu, 1 Aug 2019 15:55:55 +0000 (17:55 +0200)]
btrfs: sysfs: unexport btrfs_raid_ktype

The last non-sysfs usage of btrfs_raid_ktype has been moved to a private
helper in previous patch so the variable can be made static.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor sysfs code out of link_block_group
David Sterba [Thu, 1 Aug 2019 15:49:55 +0000 (17:49 +0200)]
btrfs: factor sysfs code out of link_block_group

The part of link_block_group that just creates the sysfs object is
independent and can be factored out to a helper.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move sysfs declarations out of ctree.h
David Sterba [Thu, 1 Aug 2019 15:34:41 +0000 (17:34 +0200)]
btrfs: move sysfs declarations out of ctree.h

As the header for sysfs code already exists, use it to clean up ctree.h.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: opencode reset of all device stats
Anand Jain [Wed, 7 Aug 2019 08:21:20 +0000 (16:21 +0800)]
btrfs: opencode reset of all device stats

__btrfs_reset_dev_stats() is a small helper function to reset devices stat
values, and is used only once, instead just open code it.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: reset device stat using btrfs_dev_stat_set
Anand Jain [Wed, 7 Aug 2019 08:21:19 +0000 (16:21 +0800)]
btrfs: reset device stat using btrfs_dev_stat_set

btrfs_dev_stat_reset() is an overdo in terms of wrapping. So this patch
open codes btrfs_dev_stat_reset().

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: qgroup: Try our best to delete qgroup relations
Qu Wenruo [Tue, 6 Aug 2019 14:05:07 +0000 (22:05 +0800)]
btrfs: qgroup: Try our best to delete qgroup relations

When we try to delete qgroups, we're pretty cautious, we make sure both
qgroups exist and there is a relationship between them, then try to
delete the relation.

This behavior is OK, but the problem is we need to two relation items,
and if we failed the first item deletion, we error out, leaving the
other relation item in qgroup tree.

Sometimes the error from del_qgroup_relation_item() could just be
-ENOENT, thus we can ignore that error and continue without any problem.

Further more, such cautious behavior makes qgroup relation deletion
impossible for orphan relation items.

This patch will enhance __del_qgroup_relation():
- If both qgroups and their relation items exist
  Go the regular deletion routine and update their accounting if needed.

- If any qgroup or relation item doesn't exist
  Then we still try to delete the orphan items anyway, but don't trigger
  the accounting update.

By this, we try our best to remove relation items, and can handle orphan
relation items properly, while still keep the existing behavior for good
qgroup tree.

Reported-by: Andrei Borzenkov <arvidjaar@gmail.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: clarify btrfs_ioctl_get_dev_stats padding
Hans van Kranenburg [Sat, 3 Aug 2019 21:36:34 +0000 (23:36 +0200)]
btrfs: clarify btrfs_ioctl_get_dev_stats padding

In commit c11d2c236cc26 ("Btrfs: add ioctl to get and reset the device
stats") the get_dev_stats ioctl was added.

Shortly thereafter, in commit b27f7c0c150f7 ("btrfs: join DEV_STATS
ioctls to one") , the flags field was added.  However, the calculation
for unused padding space was not updated, which also invalidated the
comment.

Clarify what happened to reduce confusion and wasted time for anyone
implementing this.

Signed-off-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agoBtrfs: make test_find_first_clear_extent_bit fail on incorrect results
Filipe Manana [Mon, 5 Aug 2019 09:57:41 +0000 (10:57 +0100)]
Btrfs: make test_find_first_clear_extent_bit fail on incorrect results

If any call to find_first_clear_extent_bit() returns an unexpected result,
the test should fail and not just print an error message, otherwise it
makes detection of regressions much harder to notice.

Fixes: 1eaebb341d2b41 ("btrfs: Don't trim returned range based on input value in find_first_clear_extent_bit")
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agoBtrfs: fix memory leaks in the test test_find_first_clear_extent_bit
Filipe Manana [Sat, 3 Aug 2019 08:53:16 +0000 (09:53 +0100)]
Btrfs: fix memory leaks in the test test_find_first_clear_extent_bit

The test creates an extent io tree and sets several ranges with the
CHUNK_ALLOCATED and CHUNK_TRIMMED bits, resulting in the allocation of
several extent state structures. However the test never clears those
ranges, resulting in memory leaks of the extent state structures.

This is detected when CONFIG_BTRFS_DEBUG is set once we remove the
btrfs module (rmmod btrfs):

[57399.787918] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1
[57399.790155] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1
[57399.791941] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1
[57399.793753] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1
[57399.795188] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1
[57399.796453] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1
[57399.797765] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1
[57399.799049] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1
[57399.800142] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1
[57399.801126] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1
[57399.802106] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1
[57399.803119] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1
[57399.804153] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1
[57399.805196] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1
[57399.806191] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1

The start and end offsets reported correspond exactly to the ranges
used by the test.

So fix that by clearing all the ranges when the test finishes.

Fixes: 1eaebb341d2b41 ("btrfs: Don't trim returned range based on input value in find_first_clear_extent_bit")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: delete debugfs code
David Sterba [Thu, 13 Jun 2019 15:27:36 +0000 (17:27 +0200)]
btrfs: delete debugfs code

Replaced by the sysfs exports that provide a more fine grained interface
for filesystem debugging.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: sysfs: add debugging exports
David Sterba [Thu, 13 Jun 2019 15:23:02 +0000 (17:23 +0200)]
btrfs: sysfs: add debugging exports

Add 'debug' directories to global sysfs and per-filesystem. This will
replace the debugfs directory. The sysfs location is simpler and builds
on top of the existing file hierarchy so there will hopefully be no more
questions about the sample debugfs file.

The directory is called 'debug' and only present under
CONFIG_BTRFS_DEBUG so this will not affect productions builds.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: make caching_thread use btrfs_find_next_key
Josef Bacik [Thu, 20 Jun 2019 19:37:52 +0000 (15:37 -0400)]
btrfs: make caching_thread use btrfs_find_next_key

extent-tree.c has a find_next_key that just walks up the path to find
the next key, but it is used for both the caching stuff and the snapshot
delete stuff.  The snapshot deletion stuff is special so it can't really
use btrfs_find_next_key, but the caching thread stuff can.  We just need
to fix btrfs_find_next_key to deal with ->skip_locking and then it works
exactly the same as the private find_next_key helper.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: temporarily export fragment_free_space
Josef Bacik [Thu, 20 Jun 2019 19:37:51 +0000 (15:37 -0400)]
btrfs: temporarily export fragment_free_space

This is used in caching and reading block groups, so export it while we
move these chunks independently.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: export the caching control helpers
Josef Bacik [Thu, 20 Jun 2019 19:37:50 +0000 (15:37 -0400)]
btrfs: export the caching control helpers

Man a lot of people use this stuff.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: export the excluded extents helpers
Josef Bacik [Thu, 20 Jun 2019 19:37:49 +0000 (15:37 -0400)]
btrfs: export the excluded extents helpers

We'll need this to move the caching stuff around.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: export the block group caching helpers
Josef Bacik [Thu, 20 Jun 2019 19:37:48 +0000 (15:37 -0400)]
btrfs: export the block group caching helpers

This will make it so we can move them easily.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ coding style updates ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate nocow and reservation helpers
Josef Bacik [Thu, 20 Jun 2019 19:37:47 +0000 (15:37 -0400)]
btrfs: migrate nocow and reservation helpers

These are relatively straightforward as well.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group ref counting stuff
Josef Bacik [Thu, 20 Jun 2019 19:37:46 +0000 (15:37 -0400)]
btrfs: migrate the block group ref counting stuff

Another easy set to move over to block-group.c.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: migrate the block group lookup code
Josef Bacik [Thu, 20 Jun 2019 19:37:45 +0000 (15:37 -0400)]
btrfs: migrate the block group lookup code

Move these bits first as they are the easiest to move.  Export two of
the helpers so they can be moved all at once.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ minor style updates ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move basic block_group definitions to their own header
Josef Bacik [Thu, 20 Jun 2019 19:37:44 +0000 (15:37 -0400)]
btrfs: move basic block_group definitions to their own header

This is prep work for moving all of the block group cache code into its
own file.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ minor comment updates ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: move btrfs_add_free_space out of a header file
Josef Bacik [Thu, 20 Jun 2019 19:37:43 +0000 (15:37 -0400)]
btrfs: move btrfs_add_free_space out of a header file

This is prep work for moving block_group_cache around.  Having this in
the header file makes the header file include need to be in a certain
order, which is awkward, so just move it into free-space-cache.c and
then we can re-arrange later.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tree-log: use symbolic name for first replay stage
David Sterba [Thu, 1 Aug 2019 12:50:35 +0000 (14:50 +0200)]
btrfs: tree-log: use symbolic name for first replay stage

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: async-thread: convert defines to enums
David Sterba [Thu, 1 Aug 2019 12:50:33 +0000 (14:50 +0200)]
btrfs: async-thread: convert defines to enums

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tree-log: convert defines to enums
David Sterba [Thu, 1 Aug 2019 12:50:30 +0000 (14:50 +0200)]
btrfs: tree-log: convert defines to enums

Used only for in-memory state tracking.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: remove unused key type set/get helpers
David Sterba [Thu, 1 Aug 2019 17:53:04 +0000 (19:53 +0200)]
btrfs: remove unused key type set/get helpers

The switch to open coded set/get has happend long time ago in
962a298f3511 ("btrfs: kill the key type accessor helpers"), remove the
stray helpers.

Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: remove unused btrfs_device::flush_bio_sent
David Sterba [Thu, 1 Aug 2019 17:53:02 +0000 (19:53 +0200)]
btrfs: remove unused btrfs_device::flush_bio_sent

The status of flush bio is tracked as a status bit, changed in commit
1c3063b6dbfa ("btrfs: cleanup device states define
BTRFS_DEV_STATE_FLUSH_SENT"), the flush_bio_sent was forgotten.

Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agoBtrfs: remove unnecessary condition in btrfs_clone() to avoid too much nesting
Filipe Manana [Tue, 2 Jul 2019 14:23:07 +0000 (15:23 +0100)]
Btrfs: remove unnecessary condition in btrfs_clone() to avoid too much nesting

The bulk of the work done when cloning extents, at ioctl.c:btrfs_clone(),
is done inside an if statement that checks if the found key has the type
BTRFS_EXTENT_DATA_KEY. That if statement is redundant however, because
btrfs_search_slot() always leaves us in a leaf slot that points to a key
that is always greater then or equals to the search key, and our search
key here has that type, and because we bail out before that if statement
if the key at the given leaf slot is greater then BTRFS_EXTENT_DATA_KEY.

Therefore just remove that if statement, not only because it is useless
but mostly because it increases the nesting/indentation level in this
function which is quite big and makes things a bit awkward whenever I need
to fix something that requires changing btrfs_clone() (and it has been
like that for many years already).

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Refactor btrfs_calc_avail_data_space
Nikolay Borisov [Wed, 3 Jul 2019 12:32:59 +0000 (15:32 +0300)]
btrfs: Refactor btrfs_calc_avail_data_space

Simplify the code by removing variables that don't bring any real value
as well as simplifying the checks when buidling the candidate list of
devices. No functional changes.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Remove unnecessary check from join_running_log_trans
Nikolay Borisov [Thu, 23 May 2019 11:51:26 +0000 (14:51 +0300)]
btrfs: Remove unnecessary check from join_running_log_trans

join_running_log_trans checks btrfs_root::log_root outside of
btrfs_root::log_mutex to avoid contention on the mutex. Turns out this
check is not necessary because the two callers of join_running_log_trans
(both of which deal with removing entries from the tree-log during
unlink) explicitly check whether the respective inode has been logged in
the current transaction.

If it hasn't then it won't have any items in the tree-log and call path
will return before calling join_running_log_trans. If the check passes,
however, then it's guaranteed that btrfs_root::log_root is set because
the inode is logged.

Those guarantees allows us to remove the speculative as well as the
implicity and tricky memory barrier.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>