We have a race between deleting an unused block group and balancing the
same block group that leads to an assertion failure/BUG(), producing the
following trace:

[181631.208236] BTRFS: assertion failed: 0, file: fs/btrfs/volumes.c, line: 2622
[181631.220591] ------------[ cut here ]------------
[181631.222959] kernel BUG at fs/btrfs/ctree.h:4062!
[181631.223932] invalid opcode: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[181631.224566] Modules linked in: btrfs dm_flakey dm_mod crc32c_generic xor raid6_pq nfsd auth_rpcgss oid_registry nfs_acl nfs lockd grace fscache sunrpc loop fuse acpi_cpufreq parpor$
[181631.224566] CPU: 8 PID: 17451 Comm: btrfs Tainted: G        W       4.1.0-rc5-btrfs-next-10+ #1
[181631.224566] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.8.1-0-g4adadbd-20150316_085822-nilsson.home.kraxel.org 04/01/2014
[181631.224566] task: ffff880127e09590 ti: ffff8800b5824000 task.ti: ffff8800b5824000
[181631.224566] RIP: 0010:[<ffffffffa03f19f6>]  [<ffffffffa03f19f6>] assfail.constprop.50+0x1e/0x20 [btrfs]
[181631.224566] RSP: 0018:ffff8800b5827ae8  EFLAGS: 00010246
[181631.224566] RAX: 0000000000000040 RBX: ffff8800109fc218 RCX: ffffffff81095dce
[181631.224566] RDX: 0000000000005124 RSI: ffffffff81464819 RDI: 00000000ffffffff
[181631.224566] RBP: ffff8800b5827ae8 R08: 0000000000000001 R09: 0000000000000000
[181631.224566] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800109fc200
[181631.224566] R13: ffff880020095000 R14: ffff8800b1a13f38 R15: ffff880020095000
[181631.224566] FS:  00007f70ca0b0c80(0000) GS:ffff88013ec00000(0000) knlGS:0000000000000000
[181631.224566] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[181631.224566] CR2: 00007f2872ab6e68 CR3: 00000000a717c000 CR4: 00000000000006e0
[181631.224566] Stack:
[181631.224566]  ffff8800b5827ba8 ffffffffa03f3916 ffff8800b5827b38 ffffffffa03d080e
[181631.224566]  ffffffffa03d1423 ffff880020095000 ffff88001233c000 0000000000000001
[181631.224566]  ffff880020095000 ffff8800b1a13f38 0000000a69c00000 0000000000000000
[181631.224566] Call Trace:
[181631.224566]  [<ffffffffa03f3916>] btrfs_remove_chunk+0xa4/0x6bb [btrfs]
[181631.224566]  [<ffffffffa03d080e>] ? join_transaction.isra.8+0xb9/0x3ba [btrfs]
[181631.224566]  [<ffffffffa03d1423>] ? wait_current_trans.isra.13+0x22/0xfc [btrfs]
[181631.224566]  [<ffffffffa03f3fbc>] btrfs_relocate_chunk.isra.29+0x8f/0xa7 [btrfs]
[181631.224566]  [<ffffffffa03f54df>] btrfs_balance+0xaa4/0xc52 [btrfs]
[181631.224566]  [<ffffffffa03fd388>] btrfs_ioctl_balance+0x23f/0x2b0 [btrfs]
[181631.224566]  [<ffffffff810872f9>] ? trace_hardirqs_on+0xd/0xf
[181631.224566]  [<ffffffffa04019a3>] btrfs_ioctl+0xfe2/0x2220 [btrfs]
[181631.224566]  [<ffffffff812603ed>] ? __this_cpu_preempt_check+0x13/0x15
[181631.224566]  [<ffffffff81084669>] ? arch_local_irq_save+0x9/0xc
[181631.224566]  [<ffffffff81138def>] ? handle_mm_fault+0x834/0xcd2
[181631.224566]  [<ffffffff81138def>] ? handle_mm_fault+0x834/0xcd2
[181631.224566]  [<ffffffff8103e48c>] ? __do_page_fault+0x211/0x424
[181631.224566]  [<ffffffff811755e6>] do_vfs_ioctl+0x3c6/0x479
(...)

The sequence of steps leading to this are:

           CPU 0                                         CPU 1

  btrfs_balance()
    btrfs_relocate_chunk()

      btrfs_relocate_block_group(bg X)
        btrfs_lookup_block_group(bg X)

                                               cleaner_kthread
                                                  locks fs_info->cleaner_mutex

                                                  btrfs_delete_unused_bgs()
                                                    finds bg X, which became
                                                    unused in the previous
                                                    transaction

                                                    checks bg X ->ro == 0,
                                                    so it proceeds
        sets bg X ->ro to 1
        (btrfs_set_block_group_ro(bg X))

        blocks on fs_info->cleaner_mutex
                                                    btrfs_remove_chunk(bg X)
                                                  unlocks fs_info->cleaner_mutex

        acquires fs_info->cleaner_mutex
        relocate_block_group()
          --> does nothing, no extents found in
              the extent tree from bg X
        unlocks fs_info->cleaner_mutex

      btrfs_relocate_block_group(bg X) returns

    btrfs_remove_chunk(bg X)
       extent map not found
          --> ASSERT(0)

Fix this by using a new mutex to make sure these 2 operations, block
group relocation and removal, are serialized.

This issue is reproducible by running fstests generic/038 (which stresses
chunk allocation and automatic removal of unused block groups) together
with the following balance loop:

    while true; do btrfs balance start -dusage=0 <mountpoint> ; done
Signed-off-by: NFilipe Manana <fdmanana@suse.com>
Signed-off-by: NChris Mason <clm@fb.com>

Although it is a rare case, we'd better free previous allocated
memory on error.
Signed-off-by: NZhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

It is introduced by:
 c404e0dc
 Btrfs: fix use-after-free in the finishing procedure of the device replace

But seems no relationship with that bug, this patch revirt these
code block for cleanup.
Signed-off-by: NZhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Currently, we can only set a limitation on a qgroup, but we
can not clear it.

This patch provide a choice to user to clear a limitation on
qgroup by passing a value of CLEAR_VALUE(-1) to kernel.
Reported-by: NTsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: NDongsheng Yang <yangds.fnst@cn.fujitsu.com>
Tested-by: NTsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

There is a cut and paste error so instead of freeing "head_ref", we free
"ref" twice.

Fixes: 3368d001 ('btrfs: qgroup: Record possible quota-related extent for qgroup.')
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NChris Mason <clm@fb.com>

Merge branch 'sysfs-fsdevices-4.2-part1' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux into anand

Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Tested-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NDavid Sterba <dsterba@suse.cz>

drivers/cpufreq/cpufreq.c is already using this function. And now btrfs
needs it as well. Export symbol kobject_move().
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Acked-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NDavid Sterba <dsterba@suse.cz>

This patch will add support to show the replacing target in sysfs
during the process of replacement.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.cz>

When btrfs on a device is overwritten with a new btrfs (mkfs),
the old btrfs instance in the kernel becomes stale. So with this
patch, if kernel finds device is overwritten then delete the stale
fsid/uuid.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>

Neil Horman pointed out a problem where if he did something like this

receive A
snap A B
change B
send -p A B

and then on another box do

recieve A
receive B

the receive B would fail because we use the UUID of A for the clone sources for
B.  This makes sense most of the time because normally you are sending from the
original sources, not a received source.  However when you use a recieved subvol
its UUID is going to be something completely different, so if you then try to
receive the diff on a different volume it won't find the UUID because the new A
will be something else.  The only constant is the received uuid.  So instead
check to see if we have received_uuid set on the root, and if so use that as the
clone source, as btrfs receive looks for matches either in received_uuid or
uuid.  Thanks,
Reported-by: NNeil Horman <nhorman@redhat.com>
Signed-off-by: NJosef Bacik <jbacik@fb.com>
Reviewed-by: NHugo Mills <hugo@carfax.org.uk>
Signed-off-by: NChris Mason <clm@fb.com>

@log_root_tree should not be referenced after kfree.
Signed-off-by: NLiu Bo <bo.li.liu@oracle.com>
Reviewed-by: NDavid Sterba <dsterba@suse.cz>
Reported-by: NJulia Lawall <julia.lawall@lip6.fr>
Signed-off-by: NChris Mason <clm@fb.com>

btrfs will report no_space when we run following write and delete
file loop:
 # FILE_SIZE_M=[ 75% of fs space ]
 # DEV=[ some dev ]
 # MNT=[ some dir ]
 #
 # mkfs.btrfs -f "$DEV"
 # mount -o nodatacow "$DEV" "$MNT"
 # for ((i = 0; i < 100; i++)); do dd if=/dev/zero of="$MNT"/file0 bs=1M count="$FILE_SIZE_M"; rm -f "$MNT"/file0; done
 #

Reason:
 iput() and evict() is run after write pages to block device, if
 write pages work is not finished before next write, the "rm"ed space
 is not freed, and caused above bug.

Fix:
 We can add "-o flushoncommit" mount option to avoid above bug, but
 it have performance problem. Actually, we can to wait for on-the-fly
 writes only when no-space happened, it is which this patch do.
Signed-off-by: NZhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

qgroup.

Make snapshot accounting work with new extent-oriented mechanism by
skipping given root in new/old_roots in create_pending_snapshot().
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

This is used by later qgroup fix patches for snapshot.

As current snapshot accounting is done by btrfs_qgroup_inherit(), but
new extent oriented quota mechanism will account extent from
btrfs_copy_root() and other snapshot things, causing wrong result.

So add this ability to handle snapshot accounting.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

This function will delete unode with given (val,aux) pair.
And with this patch, seqnum for debug usage doesn't have any meaning
now, so remove them.

This is used by later patches to skip snapshot root.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Goodbye, the old mechanisim.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Since the self test transaction don't have delayed_ref_roots, so use
find_all_roots() and export btrfs_qgroup_account_extent() to simulate it
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Switch from old ref_node based qgroup to extent based qgroup mechanism
for normal operations.

The new mechanism should hugely reduce the overhead of btrfs quota
system, and further more, the codes and logic should be more clean and
easier to maintain.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Switch rescan to use the new new extent oriented mechanism.

As rescan is also based on extent, new mechanism is just a perfect match
for rescan.

With re-designed internal functions, rescan is quite easy, just call
btrfs_find_all_roots() and then btrfs_qgroup_account_one_extent().
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

btrfs_qgroup_account_extents().

The new btrfs_qgroup_account_extents() function should be called in
btrfs_commit_transaction() and it will update all the qgroup according
to delayed_ref_root->dirty_extent_root.

The new function can handle both normal operation during
commit_transaction() or in rescan in a unified method with clearer
logic.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

btrfs_find_all_roots().

Allow btrfs_find_all_roots() to skip all delayed_ref_head lock and tree
lock to do tree search.

This is important for later qgroup implement which will call
find_all_roots() after fs trees are committed.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Add function btrfs_qgroup_prepare_account_extents() to get old_roots
which are needed for qgroup.

We do it in commit_transaction() and before switch_roots(), and only
search commit_root, so it gives a quite accurate view for previous
transaction.

With old_roots from previous transaction, we can use it to do accurate
account with current transaction.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Add hook in add_delayed_ref_head() to record quota-related extent record
into delayed_ref_root->dirty_extent_record rb-tree for later qgroup
accounting.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Add function qgroup_update_counters(), which will update related
qgroups' rfer/excl according to old/new_roots.

This is one of the two core functions for the new qgroup implement.

This is based on btrfs_adjust_coutners() but with clearer logic and
comment.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

This function is used to update refcnt for qgroups.
And is one of the two core functions used in the new qgroup implement.

This is based on the old update_old/new_refcnt, but provides a unified
logic and behavior.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

btrfs: extent-tree: Use ref_node to replace unneeded parameters in __inc_extent_ref() and __free_extent()

__btrfs_inc_extent_ref() and __btrfs_free_extent() have already had too
many parameters, but three of them can be extracted from
btrfs_delayed_ref_node struct.

So use btrfs_delayed_ref_node struct as a single parameter to replace
the bytenr/num_byte/no_quota parameters.

The real objective of this patch is to allow btrfs_qgroup_record_ref()
get the delayed_ref_node in incoming qgroup patches.

Other functions calling btrfs_qgroup_record_ref() are not affected since
the rest will only add/sub exclusive extents, where node is not used.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Use inline functions to do such things, to improve readability.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Acked-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NChris Mason <clm@fb.com>

Cleanup the rb_tree merge/insert/update functions, since now we use list
instead of rb_tree now.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

This patch replace the rbtree used in ref_head to list.
This has the following advantage:
1) Easier merge logic.
With the new list implement, we only need to care merging the tail
ref_node with the new ref_node.
And this can be done quite easy at insert time, no need to do a
indicated merge at run_delayed_refs().
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

Old __merge_refs() in backref.c will even merge refs whose root_id are
different, which makes qgroup gives wrong result.

Fix it by checking ref_for_same_block() before any mode specific works.
Signed-off-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NChris Mason <clm@fb.com>

lockdep report following warning in test:
 [25176.843958] =================================
 [25176.844519] [ INFO: inconsistent lock state ]
 [25176.845047] 4.1.0-rc3 #22 Tainted: G        W
 [25176.845591] ---------------------------------
 [25176.846153] inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
 [25176.846713] fsstress/26661 [HC0[0]:SC1[1]:HE1:SE0] takes:
 [25176.847246]  (&wr_ctx->wr_lock){+.?...}, at: [<ffffffffa04cdc6d>] scrub_free_ctx+0x2d/0xf0 [btrfs]
 [25176.847838] {SOFTIRQ-ON-W} state was registered at:
 [25176.848396]   [<ffffffff810bf460>] __lock_acquire+0x6a0/0xe10
 [25176.848955]   [<ffffffff810bfd1e>] lock_acquire+0xce/0x2c0
 [25176.849491]   [<ffffffff816489af>] mutex_lock_nested+0x7f/0x410
 [25176.850029]   [<ffffffffa04d04ff>] scrub_stripe+0x4df/0x1080 [btrfs]
 [25176.850575]   [<ffffffffa04d11b1>] scrub_chunk.isra.19+0x111/0x130 [btrfs]
 [25176.851110]   [<ffffffffa04d144c>] scrub_enumerate_chunks+0x27c/0x510 [btrfs]
 [25176.851660]   [<ffffffffa04d3b87>] btrfs_scrub_dev+0x1c7/0x6c0 [btrfs]
 [25176.852189]   [<ffffffffa04e918e>] btrfs_dev_replace_start+0x36e/0x450 [btrfs]
 [25176.852771]   [<ffffffffa04a98e0>] btrfs_ioctl+0x1e10/0x2d20 [btrfs]
 [25176.853315]   [<ffffffff8121c5b8>] do_vfs_ioctl+0x318/0x570
 [25176.853868]   [<ffffffff8121c851>] SyS_ioctl+0x41/0x80
 [25176.854406]   [<ffffffff8164da17>] system_call_fastpath+0x12/0x6f
 [25176.854935] irq event stamp: 51506
 [25176.855511] hardirqs last  enabled at (51506): [<ffffffff810d4ce5>] vprintk_emit+0x225/0x5e0
 [25176.856059] hardirqs last disabled at (51505): [<ffffffff810d4b77>] vprintk_emit+0xb7/0x5e0
 [25176.856642] softirqs last  enabled at (50886): [<ffffffff81067a23>] __do_softirq+0x363/0x640
 [25176.857184] softirqs last disabled at (50949): [<ffffffff8106804d>] irq_exit+0x10d/0x120
 [25176.857746]
 other info that might help us debug this:
 [25176.858845]  Possible unsafe locking scenario:
 [25176.859981]        CPU0
 [25176.860537]        ----
 [25176.861059]   lock(&wr_ctx->wr_lock);
 [25176.861705]   <Interrupt>
 [25176.862272]     lock(&wr_ctx->wr_lock);
 [25176.862881]
  *** DEADLOCK ***

Reason:
 Above warning is caused by:
 Interrupt
 -> bio_endio()
 -> ...
 -> scrub_put_ctx()
 -> scrub_free_ctx() *1
 -> ...
 -> mutex_lock(&wr_ctx->wr_lock);

 scrub_put_ctx() is allowed to be called in end_bio interrupt, but
 in code design, it will never call scrub_free_ctx(sctx) in interrupe
 context(above *1), because btrfs_scrub_dev() get one additional
 reference of sctx->refs, which makes scrub_free_ctx() only called
 withine btrfs_scrub_dev().

 Now the code runs out of our wish, because free sequence in
 scrub_pending_bio_dec() have a gap.

 Current code:
 -----------------------------------+-----------------------------------
 scrub_pending_bio_dec()            |  btrfs_scrub_dev
 -----------------------------------+-----------------------------------
 atomic_dec(&sctx->bios_in_flight); |
 wake_up(&sctx->list_wait);         |
                                    | scrub_put_ctx()
                                    | -> atomic_dec_and_test(&sctx->refs)
 scrub_put_ctx(sctx);               |
 -> atomic_dec_and_test(&sctx->refs)|
 -> scrub_free_ctx()                |
 -----------------------------------+-----------------------------------

 We expected:
 -----------------------------------+-----------------------------------
 scrub_pending_bio_dec()            |  btrfs_scrub_dev
 -----------------------------------+-----------------------------------
 atomic_dec(&sctx->bios_in_flight); |
 wake_up(&sctx->list_wait);         |
 scrub_put_ctx(sctx);               |
 -> atomic_dec_and_test(&sctx->refs)|
                                    | scrub_put_ctx()
                                    | -> atomic_dec_and_test(&sctx->refs)
                                    | -> scrub_free_ctx()
 -----------------------------------+-----------------------------------

Fix:
 Move scrub_pending_bio_dec() to a workqueue, to avoid this function run
 in interrupt context.
 Tested by check tracelog in debug.

Changelog v1->v2:
 Use workqueue instead of adjust function call sequence in v1,
 because v1 will introduce a bug pointed out by:
 Filipe David Manana <fdmanana@gmail.com>
Reported-by: NQu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: NZhao Lei <zhaolei@cn.fujitsu.com>
Reviewed-by: NFilipe Manana <fdmanana@suse.com>
Signed-off-by: NChris Mason <clm@fb.com>

The extent-same code rejects requests with an unaligned length. This
poses a problem when we want to dedupe the tail extent of files as we
skip cloning the portion between i_size and the extent boundary.

If we don't clone the entire extent, it won't be deleted. So the
combination of these behaviors winds up giving us worst-case dedupe on
many files.

We can fix this by allowing a length that extents to i_size and
internally aligining those to the end of the block. This is what
btrfs_ioctl_clone() so we can just copy that check over.
Signed-off-by: NMark Fasheh <mfasheh@suse.de>
Signed-off-by: NChris Mason <clm@fb.com>

max_to_defrag represents the number of pages to defrag rather than the last
page of the file range to be defragged.

Consider a file having 10 4k blocks (i.e. blocks in the range [0 - 9]). If the
defrag ioctl was invoked for the block range [3 - 6], then max_to_defrag
should actually have the value 4. Instead in the current code we end up
setting it to 6.

Now, this does not (yet) cause an issue since the first part of the while loop
condition in btrfs_defrag_file() (i.e. "i <= last_index") causes the control
to flow out of the while loop before any buggy behavior is actually caused. So
the patch just makes sure that max_to_defrag ends up having the right value
rather than fixing a bug. I did run the xfstests suite to make sure that the
code does not regress.

Changelog: v1->v2:
Provide a much descriptive commit message.
Signed-off-by: NChandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: NChris Mason <clm@fb.com>

Read-ahead is done for the pages in the range [ra_index, ra_index + cluster -
1]. So the next read-ahead should be starting from the page at index 'ra_index
+ cluster' (unless we deemed that the extent at 'ra_index + cluster' as
non-defraggable) rather than from the page at index 'ra_index +
max_cluster'. This patch fixes this. I did run the xfstests suite to make sure
that the code does not regress.
Signed-off-by: NChandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: NChris Mason <clm@fb.com>

When allocating a new chunk or removing one we need to update num_devs
device items and insert or remove a chunk item in the chunk tree, so
in the worst case the space needed in the chunk space_info is:

  btrfs_calc_trunc_metadata_size(chunk_root, num_devs) +
     btrfs_calc_trans_metadata_size(chunk_root, 1)

That is, in the worst case we need to cow num_devs paths and cow 1 other
path that can result in splitting every node and leaf, and each path
consisting of BTRFS_MAX_LEVEL - 1 nodes and 1 leaf. We were requiring
some additional chunk_root->nodesize * BTRFS_MAX_LEVEL * num_devs bytes,
which were unnecessary since updating the existing device items does
not result in splitting the nodes and leaf since after updating them
they remain with the same size.
Signed-off-by: NFilipe Manana <fdmanana@suse.com>
Signed-off-by: NChris Mason <clm@fb.com>

We don't need to attach ordered extents that have completed to the current
transaction. Doing so only makes us hold memory for longer than necessary
and delaying the iput of the inode until the transaction is committed (for
each created ordered extent we do an igrab and then schedule an asynchronous
iput when the ordered extent's reference count drops to 0), preventing the
inode from being evictable until the transaction commits.
Signed-off-by: NFilipe Manana <fdmanana@suse.com>
Signed-off-by: NChris Mason <clm@fb.com>

Commit 3a8b36f3 ("Btrfs: fix data loss in the fast fsync path") added
a performance regression for that causes an unnecessary sync of the log
trees (fs/subvol and root log trees) when 2 consecutive fsyncs are done
against a file, without no writes or any metadata updates to the inode in
between them and if a transaction is committed before the second fsync is
called.

Huang Ying reported this to lkml (https://lkml.org/lkml/2015/3/18/99)
after a test sysbench test that measured a -62% decrease of file io
requests per second for that tests' workload.

The test is:

  echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
  echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
  echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
  echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
  mkfs -t btrfs /dev/sda2
  mount -t btrfs /dev/sda2 /fs/sda2
  cd /fs/sda2
  for ((i = 0; i < 1024; i++)); do fallocate -l 67108864 testfile.$i; done
  sysbench --test=fileio --max-requests=0 --num-threads=4 --max-time=600 \
    --file-test-mode=rndwr --file-total-size=68719476736 --file-io-mode=sync \
    --file-num=1024 run

A test on kvm guest, running a debug kernel gave me the following results:

Without 3a8b36f3:             16.01 reqs/sec
With 3a8b36f3:                 3.39 reqs/sec
With 3a8b36f3 and this patch: 16.04 reqs/sec
Reported-by: NHuang Ying <ying.huang@intel.com>
Tested-by: NHuang, Ying <ying.huang@intel.com>
Signed-off-by: NFilipe Manana <fdmanana@suse.com>
Signed-off-by: NChris Mason <clm@fb.com>

Merge branch 'send_fixes_4.2' of git://git.kernel.org/pub/scm/linux/kernel/git/fdmanana/linux into for-linus-4.2

Zygo Blaxell and other users have reported occasional hangs while an
inode is being evicted, leading to traces like the following:

[ 5281.972322] INFO: task rm:20488 blocked for more than 120 seconds.
[ 5281.973836]       Not tainted 4.0.0-rc5-btrfs-next-9+ #2
[ 5281.974818] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 5281.976364] rm              D ffff8800724cfc38     0 20488   7747 0x00000000
[ 5281.977506]  ffff8800724cfc38 ffff8800724cfc38 ffff880065da5c50 0000000000000001
[ 5281.978461]  ffff8800724cffd8 ffff8801540a5f50 0000000000000008 ffff8801540a5f78
[ 5281.979541]  ffff8801540a5f50 ffff8800724cfc58 ffffffff8143107e 0000000000000123
[ 5281.981396] Call Trace:
[ 5281.982066]  [<ffffffff8143107e>] schedule+0x74/0x83
[ 5281.983341]  [<ffffffffa03b33cf>] wait_on_state+0xac/0xcd [btrfs]
[ 5281.985127]  [<ffffffff81075cd6>] ? signal_pending_state+0x31/0x31
[ 5281.986715]  [<ffffffffa03b4b71>] wait_extent_bit.constprop.32+0x7c/0xde [btrfs]
[ 5281.988680]  [<ffffffffa03b540b>] lock_extent_bits+0x5d/0x88 [btrfs]
[ 5281.990200]  [<ffffffffa03a621d>] btrfs_evict_inode+0x24e/0x5be [btrfs]
[ 5281.991781]  [<ffffffff8116964d>] evict+0xa0/0x148
[ 5281.992735]  [<ffffffff8116a43d>] iput+0x18f/0x1e5
[ 5281.993796]  [<ffffffff81160d4a>] do_unlinkat+0x15b/0x1fa
[ 5281.994806]  [<ffffffff81435b54>] ? ret_from_sys_call+0x1d/0x58
[ 5281.996120]  [<ffffffff8107d314>] ? trace_hardirqs_on_caller+0x18f/0x1ab
[ 5281.997562]  [<ffffffff8123960b>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[ 5281.998815]  [<ffffffff81161a16>] SyS_unlinkat+0x29/0x2b
[ 5281.999920]  [<ffffffff81435b32>] system_call_fastpath+0x12/0x17
[ 5282.001299] 1 lock held by rm/20488:
[ 5282.002066]  #0:  (sb_writers#12){.+.+.+}, at: [<ffffffff8116dd81>] mnt_want_write+0x24/0x4b

This happens when we have readahead, which calls readpages(), happening
right before the inode eviction handler is invoked. So the reason is
essentially:

1) readpages() is called while a reference on the inode is held, so
   eviction can not be triggered before readpages() returns. It also
   locks one or more ranges in the inode's io_tree (which is done at
   extent_io.c:__do_contiguous_readpages());

2) readpages() submits several read bios, all with an end io callback
   that runs extent_io.c:end_bio_extent_readpage() and that is executed
   by other task when a bio finishes, corresponding to a work queue
   (fs_info->end_io_workers) worker kthread. This callback unlocks
   the ranges in the inode's io_tree that were previously locked in
   step 1;

3) readpages() returns, the reference on the inode is dropped;

4) One or more of the read bios previously submitted are still not
   complete (their end io callback was not yet invoked or has not
   yet finished execution);

5) Inode eviction is triggered (through an unlink call for example).
   The inode reference count was not incremented before submitting
   the read bios, therefore this is possible;

6) The eviction handler starts executing and enters the loop that
   iterates over all extent states in the inode's io_tree;

7) The loop picks one extent state record and uses its ->start and
   ->end fields, after releasing the inode's io_tree spinlock, to
   call lock_extent_bits() and clear_extent_bit(). The call to lock
   the range [state->start, state->end] blocks because the whole
   range or a part of it was locked by the previous call to
   readpages() and the corresponding end io callback, which unlocks
   the range was not yet executed;

8) The end io callback for the read bio is executed and unlocks the
   range [state->start, state->end] (or a superset of that range).
   And at clear_extent_bit() the extent_state record state is used
   as a second argument to split_state(), which sets state->start to
   a larger value;

9) The task executing the eviction handler is woken up by the task
   executing the bio's end io callback (through clear_state_bit) and
   the eviction handler locks the range
   [old value for state->start, state->end]. Shortly after, when
   calling clear_extent_bit(), it unlocks the range
   [new value for state->start, state->end], so it ends up unlocking
   only part of the range that it locked, leaving an extent state
   record in the io_tree that represents the unlocked subrange;

10) The eviction handler loop, in its next iteration, gets the
    extent_state record for the subrange that it did not unlock in the
    previous step and then tries to lock it, resulting in an hang.

So fix this by not using the ->start and ->end fields of an existing
extent_state record. This is a simple solution, and an alternative
could be to bump the inode's reference count before submitting each
read bio and having it dropped in the bio's end io callback. But that
would be a more invasive/complex change and would not protect against
other possible places that are not holding a reference on the inode
as well. Something to consider in the future.

Many thanks to Zygo Blaxell for reporting, in the mailing list, the
issue, a set of scripts to trigger it and testing this fix.
Reported-by: NZygo Blaxell <ce3g8jdj@umail.furryterror.org>
Tested-by: NZygo Blaxell <ce3g8jdj@umail.furryterror.org>
Signed-off-by: NFilipe Manana <fdmanana@suse.com>
Signed-off-by: NChris Mason <clm@fb.com>