We get a use after free if we had a transaction to cleanup since there could be
delayed inodes which refer to their respective fs_root.  Thanks
Reported-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>
Signed-off-by: NChris Mason <chris.mason@fusionio.com>

Btrfs has been pointer tagging bi_private and using bi_bdev
to store the stripe index and mirror number of failed IOs.

As bios bubble back up through the call chain, we use these
to decide if and how to retry our IOs.  They are also used
to count IO failures on a per device basis.

Recently a bio tracepoint was added lead to crashes because
we were abusing bi_bdev.

This commit adds a btrfs bioset, and creates explicit fields
for the mirror number and stripe index.  The plan is to
extend this structure for all of the fields currently in
struct btrfs_bio, which will mean one less kmalloc in
our IO path.
Signed-off-by: NChris Mason <chris.mason@fusionio.com>
Reported-by: NTejun Heo <tj@kernel.org>

If we fail to load the chunk tree we'll call free_root_pointers, except we may
not have assigned the roots for the dev_root/extent_root/csum_root yet, so we
could NULL pointer deref at this point.  Just add checks to make sure these
roots are set to keep us from panicing.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

btrfs_invalidate_inodes() may sleep, so we should not invoke it in the
spin lock context. Fix it.
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We have checked if ->node is NULL or not, so it is unnecessary to
use BUG_ON() to check again. Remove it.
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

I'm sorry, theres no excuse for this sort of work.  We need to use
root->leafsize since eb may be NULL.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Quota tree has been missing from lockdep annotations, though no warning
has been seen in the wild.

There's currently one entry that does not belong there,
BTRFS_ORPHAN_OBJECTID.  No such tree exists, it's probably a copy &
paste mistake, the id is defined among tree ids.
Signed-off-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We've added new checks to make sure the super block crc is correct
during mount.  A fresh filesystem from an older mkfs won't have the
crc set.  This adds a warning when it finds a newly created filesystem
but doesn't fail the mount.
Signed-off-by: NChris Mason <chris.mason@fusionio.com>

The superblock checksum is not verified upon mount. <awkward silence>

Add that check and also reorder existing checks to a more logical
order.

Current mkfs.btrfs does not calculate the correct checksum of
super_block and thus a freshly created filesytem will fail to mount when
this patch is applied.

First transaction commit calculates correct superblock checksum and
saves it to disk.

Reproducer:
$ mfks.btrfs /dev/sda
$ mount /dev/sda /mnt
$ btrfs scrub start /mnt
$ sleep 5
$ btrfs scrub status /mnt
... super:2 ...
Signed-off-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>
Signed-off-by: NChris Mason <chris.mason@fusionio.com>

It's unused since 0b32f4bb.
Signed-off-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Big patch, but all it does is add statics to functions which
are in fact static, then remove the associated dead-code fallout.

removed functions:

btrfs_iref_to_path()
__btrfs_lookup_delayed_deletion_item()
__btrfs_search_delayed_insertion_item()
__btrfs_search_delayed_deletion_item()
find_eb_for_page()
btrfs_find_block_group()
range_straddles_pages()
extent_range_uptodate()
btrfs_file_extent_length()
btrfs_scrub_cancel_devid()
btrfs_start_transaction_lflush()

btrfs_print_tree() is left because it is used for debugging.
btrfs_start_transaction_lflush() and btrfs_reada_detach() are
left for symmetry.

ulist.c functions are left, another patch will take care of those.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

If you try to mount -o loop a restored file system it will panic if the file
ends up being smaller than the original disk.  This is because we go to try and
get a block for a super that may be past the EOF which makes __getblk return
NULL for a buffer head when we aren't expecting it to.  Fix this by dealing with
this case and just jacking up the errors count.  With this patch we no longer
panic when mounting a restored file system loopback.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

If qgroup tracking is out of sync, a rescan operation can be started. It
iterates the complete extent tree and recalculates all qgroup tracking data.
This is an expensive operation and should not be used unless required.

A filesystem under rescan can still be umounted. The rescan continues on the
next mount.  Status information is provided with a separate ioctl while a
rescan operation is in progress.
Signed-off-by: NJan Schmidt <list.btrfs@jan-o-sch.net>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Sequence numbers for delayed refs have been introduced in the first version
of the qgroup patch set. To solve the problem of find_all_roots on a busy
file system, the tree mod log was introduced. The sequence numbers for that
were simply shared between those two users.

However, at one point in qgroup's quota accounting, there's a statement
accessing the previous sequence number, that's still just doing (seq - 1)
just as it would have to in the very first version.

To satisfy that requirement, this patch makes the sequence number counter 64
bit and splits it into a major part (used for qgroup sequence number
counting) and a minor part (incremented for each tree modification in the
log). This enables us to go exactly one major step backwards, as required
for qgroups, while still incrementing the sequence counter for tree mod log
insertions to keep track of their order. Keeping them in a single variable
means there's no need to change all the code dealing with comparisons of two
sequence numbers.

The sequence number is reset to 0 on commit (not new in this patch), which
ensures we won't overflow the two 32 bit counters.

Without this fix, the qgroup tracking can occasionally go wrong and WARN_ONs
from the tree mod log code may happen.
Signed-off-by: NJan Schmidt <list.btrfs@jan-o-sch.net>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

It is a rare exception that a new tree is created, like the qgroups
tree. So far these new trees have an all-zero UUID in their root
items. All trees that mkfs.btrfs has created get an UUID during the
first mount when btrfs_read_root_item() rewrites the root_item to
the v2 structure style. These UUID are never used so far, but
anyway, since it is better to have it uniform for all trees, this
commit adds some lines that generate and write an UUID for newly
created trees.
Signed-off-by: NStefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Signed-off-by: NStefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

I have a broken file system that when it aborts leaves all sorts of accounting
things wrong and gives you lots of WARN_ON()'s other than the abort.  This is
because we're not cleaning up various parts of the file system when we abort.
The first chunks are specific to mount failures, we weren't cleaning up the
block group cached inodes and we weren't cleaning up any transactions that had
been aborted, which leaves a bunch of things laying around.

The second half of this are related to the cleanup parts.  First we don't need
to release space for the dirty pages from the trans_block_rsv, that's all
handled by the trans handles so this is just plain wrong.  The other thing is we
need to pin down extents that were set ->must_insert_reserved for delayed refs.
This isn't so much for the pinning but more for the cleaning up the
cache->reserved counter since we are no longer going to use those reserved
bytes.  With this patch I no longer see a bunch of WARN_ON()'s when I try to
mount this broken file system, just the initial one from the abort.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We can just look up the extent_buffers for the range and free stuff that way.
This makes the cleanup a bit cleaner and we can make sure to evict the
extent_buffers pretty quickly by marking them as stale.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We can run the tree logging recovery or the orphan cleanup on mount, so we'll
end up looking up a random fs tree in the meantime.  So we need to clean this up
so we don't leave extent buffers hanging around on the cache.  With this patch
we no longer leak extent buffers on failure to mount.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We kept leaking extent buffers when mounting a broken file system and it turns
out it's because not everybody uses read_tree_block properly.  You need to check
and make sure the extent_buffer is uptodate before you use it.  This patch fixes
everybody who calls read_tree_block directly to make sure they check that it is
uptodate and free it and return an error if it is not.  With this we no longer
leak EB's when things go horribly wrong.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

With a users corrupted fs I was getting weird behavior and panics and it turns
out it was because one of his tree blocks had a bogus header level.  So add this
to the sanity checks in the endio handler for tree blocks.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Martin Steigerwald reported a BUG_ON() where we were given a bogus bytenr to
map.  Turns out he is using > PAGESIZE leafsizes.  The readahead stuff is called
every time we do a completion, but we may not have finished reading in all the
pages, so the bytenr we read off the node could be completely bogus.  Fix this
by only calling the readahead hook once all pages have been read in.  Thanks,
Reported-by: NMartin Steigerwald <Martin@lichtvoll.de>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Dave reported a BUG_ON() that happened in end_page_writeback() after an abort.
This happened because we unconditionally call end_page_writeback() in the endio
case, which is right.  However when we abort the transaction we will call
end_page_writeback() on any writeback pages we find, which is wrong.  We need to
lock the page and wait on page writeback to complete if it is.  There is nothing
unsafe about this since we are discarding the transaction anyway.  Thanks,
Reported-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

The following case will make the incompat/compat flag of the super block
be recovered.
 Task1					|Task2
 flags = btrfs_super_incompat_flags();	|
					|flags = btrfs_super_incompat_flags();
 flags |= new_flag1;			|
					|flags |= new_flag2;
 btrfs_set_super_incompat_flags(flags);	|
					|btrfs_set_super_incompat_flags(flags);
the new_flag1 is recovered.

In order to avoid this problem, we introduce a lock named super_lock into
the btrfs_fs_info structure. If we want to update incompat/compat flags
of the super block, we must hold it.
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

The original code has one spin_lock 'qgroup_lock' to protect quota
configurations in memory. If we want to add a BTRFS_QGROUP_INFO_KEY,
it will be added to Btree firstly, and then update configurations in
memory,however, a race condition may happen between these operations.
For example:
	->add_qgroup_info_item()
		->add_qgroup_rb()

For the above case, del_qgroup_info_item() may happen just before
add_qgroup_rb().

What's worse, when we want to add a qgroup relation:
	->add_qgroup_relation_item()
		->add_qgroup_relations()

We don't have any checks whether 'src' and 'dst' exist before
add_qgroup_relation_item(), a race condition can also happen for
the above case.

To avoid race condition and have all the necessary checks, we introduce
a mutex lock 'qgroup_ioctl_lock', and we make all the user change operations
protected by the mutex lock.
Signed-off-by: NWang Shilong <wangsl-fnst@cn.fujitsu.com>
Reviewed-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

A user sent me a btrfs-image of a file system that was panicing on mount during
the log recovery.  I had originally thought these problems were from a bug in
the free space cache code, but that was just a symptom of the problem.  The
problem is if your application does something like this

[prealloc][prealloc][prealloc]

the internal extent maps will merge those all together into one extent map, even
though on disk they are 3 separate extents.  So if you go to write into one of
these ranges the extent map will be right since we use the physical extent when
doing the write, but when we log the extents they will use the wrong sizes for
the remainder prealloc space.  If this doesn't happen to trip up the free space
cache (which it won't in a lot of cases) then you will get bogus entries in your
extent tree which will screw stuff up later.  The data and such will still work,
but everything else is broken.  This patch fixes this by not allowing extents
that are on the modified list to be merged.  This has the side effect that we
are no longer adding everything to the modified list all the time, which means
we now have to call btrfs_drop_extents every time we log an extent into the
tree.  So this allows me to drop all this speciality code I was using to get
around calling btrfs_drop_extents.  With this patch the testcase I've created no
longer creates a bogus file system after replaying the log.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Each time pick one dead root from the list and let the caller know if
it's needed to continue. This should improve responsiveness during
umount and balance which at some point waits for cleaning all currently
queued dead roots.

A new dead root is added to the end of the list, so the snapshots
disappear in the order of deletion.

The snapshot cleaning work is now done only from the cleaner thread and the
others wake it if needed.
Signed-off-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Share the exactly same code of stopping workers.
Signed-off-by: NLiu Bo <bo.li.liu@oracle.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We currently store the first key of the tree block inside the reference for the
tree block in the extent tree.  This takes up quite a bit of space.  Make a new
key type for metadata which holds the level as the offset and completely removes
storing the btrfs_tree_block_info inside the extent ref.  This reduces the size
from 51 bytes to 33 bytes per extent reference for each tree block.  In practice
this results in a 30-35% decrease in the size of our extent tree, which means we
COW less and can keep more of the extent tree in memory which makes our heavy
metadata operations go much faster.  This is not an automatic format change, you
must enable it at mkfs time or with btrfstune.  This patch deals with having
metadata stored as either the old format or the new format so it is easy to
convert.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

free_root_pointers() has been introduced to cleanup all of tree roots,
so just use it instead.
Signed-off-by: NLiu Bo <bo.li.liu@oracle.com>
Reviewed-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Argument 'root' is no more used in btrfs_csum_data().
Signed-off-by: NLiu Bo <bo.li.liu@oracle.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

We should free leaf and root before returning from the error
handling code.
Signed-off-by: NTsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>
Signed-off-by: NChris Mason <chris.mason@fusionio.com>

Now that we use bit operation to check fs_state, update
btrfs_free_fs_root()'s checker, otherwise we get back to
memory leak case.
Signed-off-by: NLiu Bo <bo.li.liu@oracle.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>
Signed-off-by: NChris Mason <chris.mason@fusionio.com>

There are several bugs at error path of create_snapshot() when the
transaction commitment failed.
- access the freed transaction handler. At the end of the
  transaction commitment, the transaction handler was freed, so we
  should not access it after the transaction commitment.
- we were not aware of the error which happened during the snapshot
  creation if we submitted a async transaction commitment.
- pending snapshot access vs pending snapshot free. when something
  wrong happened after we submitted a async transaction commitment,
  the transaction committer would cleanup the pending snapshots and
  free them. But the snapshot creators were not aware of it, they
  would access the freed pending snapshots.

This patch fixes the above problems by:
- remove the dangerous code that accessed the freed handler
- assign ->error if the error happens during the snapshot creation
- the transaction committer doesn't free the pending snapshots,
  just assigns the error number and evicts them before we unblock
  the transaction.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

The stripe hash table is large, starting with allocation order 4 and can go as
high as order 7 in case lock debugging is turned on and structure padding
happens.

Observed mount failure:

mount: page allocation failure: order:7, mode:0x200050
Pid: 8234, comm: mount Tainted: G        W    3.8.0-default+ #267
Call Trace:
 [<ffffffff81114353>] warn_alloc_failed+0xf3/0x140
 [<ffffffff811171d2>] ? __alloc_pages_direct_compact+0x92/0x250
 [<ffffffff81117ac3>] __alloc_pages_nodemask+0x733/0x9d0
 [<ffffffff81152878>] ? cache_alloc_refill+0x3f8/0x840
 [<ffffffff811528bc>] cache_alloc_refill+0x43c/0x840
 [<ffffffff811302eb>] ? is_kernel_percpu_address+0x4b/0x90
 [<ffffffffa00a00ac>] ? btrfs_alloc_stripe_hash_table+0x5c/0x130 [btrfs]
 [<ffffffff811531d7>] kmem_cache_alloc_trace+0x247/0x270
 [<ffffffffa00a00ac>] btrfs_alloc_stripe_hash_table+0x5c/0x130 [btrfs]
 [<ffffffffa003133f>] open_ctree+0xb2f/0x1f90 [btrfs]
 [<ffffffff81397289>] ? string+0x49/0xe0
 [<ffffffff813987b3>] ? vsnprintf+0x443/0x5d0
 [<ffffffffa0007cb6>] btrfs_mount+0x526/0x600 [btrfs]
 [<ffffffff8115127c>] ? cache_alloc_debugcheck_after+0x4c/0x200
 [<ffffffff81162b90>] mount_fs+0x20/0xe0
 [<ffffffff8117db26>] vfs_kern_mount+0x76/0x120
 [<ffffffff811801b6>] do_mount+0x386/0x980
 [<ffffffff8112a5cb>] ? strndup_user+0x5b/0x80
 [<ffffffff81180840>] sys_mount+0x90/0xe0
 [<ffffffff81962e99>] system_call_fastpath+0x16/0x1b
Signed-off-by: NDavid Sterba <dsterba@suse.cz>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

When we abort a transaction while fsyncing, we'll skip freeing log roots
part of committing a transaction, which leads to memory leak.

This adds a 'free log roots' in putting super when no more users hold
references on log roots, so it's safe and clean.
Signed-off-by: NLiu Bo <bo.li.liu@oracle.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

super.magic is an le64 but it's treated as an unterminated string when
compared against BTRFS_MAGIC which is defined as a string.  Instead
define BTRFS_MAGIC as a normal hex value and use endian helpers to
compare it to the super's magic.

I tested this by mounting an fs made before the change and made sure
that it didn't introduce sparse errors.  This matches a similar cleanup
that is pending in btrfs-progs.  David Sterba pointed out that we should
fix the kernel side as well :).
Signed-off-by: NZach Brown <zab@redhat.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

Miao made the ordered operations stuff run async, which introduced a
deadlock where we could get somebody (sync) racing in and committing the
transaction while a commit was already happening.  The new committer would
try and flush ordered operations which would hang waiting for the commit to
finish because it is done asynchronously and no longer inherits the callers
trans handle.  To fix this we need to make the ordered operations list a per
transaction list.  We can get new inodes added to the ordered operation list
by truncating them and then having another process writing to them, so this
makes it so that anybody trying to add an ordered operation _must_ start a
transaction in order to add itself to the list, which will keep new inodes
from getting added to the ordered operations list after we start committing.
This should fix the deadlock and also keeps us from doing a lot more work
than we need to during commit.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

open_ctree() need read the metadata to initialize the global information
of btrfs. But it may fail after it submit some bio, and then it will jump
to the error path. Unfortunately, it doesn't check if there are some bios
in flight, and just stop all the worker threads. As a result, when the
submitted bios end, they can not find any worker thread which can deal with
subsequent work, then oops happen.

kernel BUG at fs/btrfs/async-thread.c:605!

Fix this problem by invoking invalidate_inode_pages2() before we stop the
worker threads. This function will wait until the bio end because it need
lock the pages which are going to be invalidated, and if a page is under
disk read IO, it must be locked. invalidate_inode_pages2() need wait until
end bio handler to unlocked it.
Reported-and-Tested-by: NTsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>

When we abort we've been just free'ing up all the ordered extents and
hoping for the best.  This results in lots of warnings from various places,
warnings from btrfs_destroy_inode() because it's ENOSPC accounting isn't
fixed.  It will also screw up lots of pages who have been set private but
never get cleared because the ordered extents are never allowed to be
submitted.  This patch fixes those warnings.  Thanks,
Signed-off-by: NJosef Bacik <jbacik@fusionio.com>