Return device pointer (with the IS_ERR semantics) from
btrfs_scan_one_device so we don't have to return in through pointer.

And since btrfs_fs_devices can be obtained from btrfs_device, return that.
Signed-off-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
[ fixed conflics after recent changes to btrfs_scan_one_device ]
Signed-off-by: NDavid Sterba <dsterba@suse.com>

fs_devices is always passed to btrfs_scan_one_device which overrides it.
In the call stack below fs_devices is passed to btrfs_scan_one_device
from btrfs_mount_root.  In btrfs_mount_root the output fs_devices of
this call stack is not used.

btrfs_mount_root
  btrfs_parse_early_options
    btrfs_scan_one_device

So, it is not necessary to pass fs_devices from btrfs_mount_root, using
a local variable in btrfs_parse_early_options is enough.
Signed-off-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NAnand Jain <Anand.Jain@oracle.com>
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Technically this extends the critical section covered by uuid_mutex to:

- parse early mount options -- here we can call device scan on paths
  that can be passed as 'device=/dev/...'

- scan the device passed to mount

- open the devices related to the fs_devices -- this increases
  fs_devices::opened

The race can happen when mount calls one of the scans and there's
another one called eg. by mkfs or 'btrfs dev scan':

Mount                                  Scan
-----                                  ----
scan_one_device (dev1, fsid1)
                                       scan_one_device (dev2, fsid1)
				           add the device
					   free stale devices
					       fsid1 fs_devices::opened == 0
					           find fsid1:dev1
					           free fsid1:dev1
					           if it's the last one,
					            free fs_devices of fsid1
						    too

open_devices (dev1, fsid1)
   dev1 not found

When fixed, the uuid mutex will make sure that mount will increase
fs_devices::opened and this will not be touched by the racing scan
ioctl.

Reported-and-tested-by: syzbot+909a5177749d7990ffa4@syzkaller.appspotmail.com
Reported-and-tested-by: syzbot+ceb2606025ec1cc3479c@syzkaller.appspotmail.com
Reviewed-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

In preparation to take a big lock, move resource initialization before
the critical section. It's not obvious from the diff, the desired order
is:

- initialize mount security options
- allocate temporary fs_info
- allocate superblock buffers
Reviewed-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Prepartory work to fix race between mount and device scan.

btrfs_parse_early_options calls the device scan from mount and we'll
need to let mount completely manage the critical section.
Reviewed-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Prepartory work to fix race between mount and device scan.

The callers will have to manage the critical section, eg. mount wants to
scan and then call btrfs_open_devices without the ioctl scan walking in
and modifying the fs devices in the meantime.
Reviewed-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Prepartory work to fix race between mount and device scan.

The callers will have to manage the critical section, eg. mount wants to
scan and then call btrfs_open_devices without the ioctl scan walking in
and modifying the fs devices in the meantime.
Reviewed-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

btrfs_free_stale_devices() finds a stale (not opened) device matching
path in the fs_uuid list. We are already under uuid_mutex so when we
check for each fs_devices, hold the device_list_mutex too.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Over the years we named %fs_devices and %devices to represent the
struct btrfs_fs_devices and the struct btrfs_device. So follow the same
scheme here too. No functional changes.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Make sure the device_list_lock is held the whole time:

* when the device is being looked up
* new device is initialized and put to the list
* the list counters are updated (fs_devices::opened, fs_devices::total_devices)
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
[ update changelog ]
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

btrfs_free_stale_devices() looks for device path reused for another
filesystem, and deletes the older fs_devices::device entry.

In preparation to handle locking in device_list_add, move
btrfs_free_stale_devices outside as these two functions serve a
different purpose.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Since commit 88c14590 ("btrfs: use RCU in btrfs_show_devname for
device list traversal") btrfs_show_devname no longer takes
device_list_mutex. As such the deadlock that 0ccd0528 ("btrfs: fix a
possible umount deadlock") aimed to fix no longer exists, we can free
the devices immediatelly and remove the code that does the pending work.
Signed-off-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NAnand Jain <anand.jain@oracle.com>
[ update changelog ]
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

This function is not used since the alloc_start parameter has been
obsoleted in commit 0d0c71b3 ("btrfs: obsolete and remove
mount option alloc_start").
Signed-off-by: NQu Wenruo <wqu@suse.com>
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Since parameter flags is no more used since commit d7407606 ("btrfs:
split parse_early_options() in two"), remove it.
Signed-off-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

In case of deleting the seed device the %cur_devices (seed) and the
%fs_devices (parent) are different. Now, as the parent
fs_devices::total_devices also maintains the total number of devices
including the seed device, so decrement its in-memory value for the
successful seed delete. We are already updating its corresponding
on-disk btrfs_super_block::number_devices value.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Commit 5d23515b ("btrfs: Move qgroup rescan on quota enable to
btrfs_quota_enable") not only resulted in an easier to follow code but
it also introduced a subtle bug. It changed the timing when the initial
transaction rescan was happening:

- before the commit: it would happen after transaction commit had occured
- after the commit: it might happen before the transaction was committed

This results in failure to correctly rescan the quota since there could
be data which is still not committed on disk.

This patch aims to fix this by moving the transaction creation/commit
inside btrfs_quota_enable, which allows to schedule the quota commit
after the transaction has been committed.

Fixes: 5d23515b ("btrfs: Move qgroup rescan on quota enable to btrfs_quota_enable")
Reported-by: NMisono Tomohiro <misono.tomohiro@jp.fujitsu.com>
Link: https://marc.info/?l=linux-btrfs&m=152999289017582Signed-off-by: NNikolay Borisov <nborisov@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Add fall-back code to catch failure of full_stripe_write. Proper error
handling from inside run_plug would need more code restructuring as it's
called at arbitrary points by io scheduler.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

There's only one call site of the unlocked helper so it can be folded
into the caller.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Signed-off-by: NDavid Sterba <dsterba@suse.com>

Signed-off-by: NDavid Sterba <dsterba@suse.com>

Signed-off-by: NDavid Sterba <dsterba@suse.com>

Add helper that schedules a given function to run on the rmw workqueue.
This will replace several standalone helpers.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

The helper is trivial and marked as deprecated.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

The loops iterating eb pages use unsigned long, that's an overkill as
we know that there are at most 16 pages (64k / 4k), and 4 by default
(with nodesize 16k).
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Almost all callers pass the start and len as 2 arguments but this is not
necessary, all the information is provided by the eb. By reordering the
calls to num_extent_pages, we don't need the local variables with
start/len.
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Remove includes if none of the interfaces and exports is used in the
given source file.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Use the helper that's possibly optimized for full page copies.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Functions that get btrfs inode can simply reach the fs_info by
dereferencing the root and this looks a bit more straightforward
compared to the btrfs_sb(...) indirection.

If the transaction handle is available and not NULL it's used instead.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

There are several places when the btrfs inode is converted to the
generic inode, back to btrfs and then passed to btrfs_ino. We can remove
the extra back and forth conversions.
Signed-off-by: NDavid Sterba <dsterba@suse.com>

io_ctl_set_generation() assumes that the generation number shares
the same page with inline CRCs. Let's make sure this is always true.
Signed-off-by: NZhihui Zhang <zzhsuny@gmail.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

There is only usage of the declared devices variable, instead use its
value directly.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

There are many instances of the %fs_info->fs_devices pointer
dereferences, use a temporary variable instead.
Signed-off-by: NAnand Jain <anand.jain@oracle.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Invalid reloc tree can cause kernel NULL pointer dereference when btrfs
does some cleanup of the reloc roots.

It turns out that fs_info::reloc_ctl can be NULL in
btrfs_recover_relocation() as we allocate relocation control after all
reloc roots have been verified.
So when we hit: note, we haven't called set_reloc_control() thus
fs_info::reloc_ctl is still NULL.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=199833Reported-by: NXu Wen <wen.xu@gatech.edu>
Signed-off-by: NQu Wenruo <wqu@suse.com>
Tested-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

A crafted image has empty root tree block, which will later cause NULL
pointer dereference.

The following trees should never be empty:
1) Tree root
   Must contain at least root items for extent tree, device tree and fs
   tree

2) Chunk tree
   Or we can't even bootstrap as it contains the mapping.

3) Fs tree
   At least inode item for top level inode (.).

4) Device tree
   Dev extents for chunks

5) Extent tree
   Must have corresponding extent for each chunk.

If any of them is empty, we are sure the fs is corrupted and no need to
mount it.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=199847Reported-by: NXu Wen <wen.xu@gatech.edu>
Signed-off-by: NQu Wenruo <wqu@suse.com>
Tested-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

A crafted image with invalid block group items could make free space cache
code to cause panic.

We could detect such invalid block group item by checking:
1) Item size
   Known fixed value.
2) Block group size (key.offset)
   We have an upper limit on block group item (10G)
3) Chunk objectid
   Known fixed value.
4) Type
   Only 4 valid type values, DATA, METADATA, SYSTEM and DATA|METADATA.
   No more than 1 bit set for profile type.
5) Used space
   No more than the block group size.

This should allow btrfs to detect and refuse to mount the crafted image.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=199849Reported-by: NXu Wen <wen.xu@gatech.edu>
Signed-off-by: NQu Wenruo <wqu@suse.com>
Reviewed-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Tested-by: NGu Jinxiang <gujx@cn.fujitsu.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

The v0 extent type checks are the right case for the unlikely
annotations as we don't expect to ever see them, so let's give the
compiler some hint.
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

Following the removal of the v0 handling code let's be courteous and
print an error message when such extents are handled. In the cases
where we have a transaction just abort it, otherwise just call
btrfs_handle_fs_error. Both cases result in the FS being re-mounted RO.

In case the error handling would be too intrusive, leave the BUG_ON in
place, like extent_data_ref_count, other proper handling would catch
that earlier.
Signed-off-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

The v0 compat code was introduced in commit 5d4f98a2
("Btrfs: Mixed back reference  (FORWARD ROLLING FORMAT CHANGE)") 9
years ago, which was merged in 2.6.31. This means that the code is
there to support filesystems which are _VERY_ old and if you are using
btrfs on such an old kernel, you have much bigger problems. This coupled
with the fact that no one is likely testing/maintining this code likely
means it has bugs lurking. All things considered I think 43 kernel
releases later it's high time this remnant of the past got removed.

This patch removes all code wrapped in #ifdefs but leaves the BUG_ONs in case
we have a v0 with no support intact as a sort of safety-net.
Signed-off-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>

It's only coding style fix not functinal change.  When if/else has only
one statement then the braces are not needed.
Signed-off-by: NChengguang Xu <cgxu519@gmx.com>
Reviewed-by: NNikolay Borisov <nborisov@suse.com>
Reviewed-by: NDavid Sterba <dsterba@suse.com>
Signed-off-by: NDavid Sterba <dsterba@suse.com>