When working with a user who had a broken file system I noticed that we were
reading a bad copy of a block when the other copy was perfectly fine. This is
because we don't keep track of the parent generation for tree blocks, so we just
read whichever copy we damned well please with no regards for which is best.
This fixes this problem by recording the parent generation of the tree block so
we can be sure to read the most correct copy before we check it, which will give
us a better chance of fixing really broken filesystems. Thanks,
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
When using gcc 4.8.2, -Wmaybe-uninitialized will report root_item may be
used uninitialized.
Since root_item_valid variant is used to determine the root_item valid,
it's a false alert and to avoid the warning, just init it on allocation.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
The code was incorrectly adding the file extent items' data offset to the logical
disk address of the extent (bytenr) when the extent is compressed. The offset is
relative to the uncompressed data and not to what we store on disk (compressed).
Also it attempted to copy ram_bytes to destination, which is incorrect when the
data offset field is non-zero, it must use num_bytes instead.
A test case for xfstests follows.
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
get_fs_info() provides the info of the specific
device/devid, however when we delete the missing disk
the super-block on the disk isn't cleared, and since
btrfs-progs makes its decision by reading the disk super
block, so it doesn't know about the kernel previous action,
And now when we tried to probe kernel for the devid it fails.
reproducer:
$ mkfs.btrfs -d raid1 -m raid1 /dev/sde /dev/sdf
$ modprobe -r btrfs && modprobe btrfs
$ mount -o degraded /dev/sde /btrfs
$ btrfs dev add /dev/sdd /btrfs
$ btrfs dev del missing /btrfs
$ btrfs scrub start -B /dev/sdf
btrfs: utils.c:1741: get_fs_info: Assertion `!(ndevs == 0)' failed.
Aborted (core dumped)
Signed-off-by: Anand Jain <Anand.Jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
btrfs-progs picks the latest_dev based on first probed
greatest trans-id. However below test case proofs that
approach is wrong.
$ mkfs.btrfs -d raid1 -m raid1 /dev/sde /dev/sdf
$ modprobe -r btrfs && modprobe btrfs
$ mount -o degraded /dev/sde /btrfs
$ touch /btrfs/testfile && btrfs fi sync /btrfs
The above steps will make /dev/sdf not part of the btrfs.
and as below when you use /dev/sdf the btrfs dev stat
and dev scrub picks up wrong disk
$ btrfs dev stat /dev/sdf
[/dev/sde].write_io_errs 0
[/dev/sde].read_io_errs 0
[/dev/sde].flush_io_errs 0
[/dev/sde].corruption_errs 0
[/dev/sde].generation_errs 0
$ btrfs scrub start -B /dev/sdf
scrub done for 2e99c881-6abd-4f8a-8290-e2f8d0acc575
scrub started at Mon Feb 24 14:45:06 2014 and finished after 0 seconds
total bytes scrubbed: 256.00KiB with 0 errors
Signed-off-by: Anand Jain <Anand.Jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
as of now, when we replace a disk, it is added to the
dev list with devid 0. And we fail to obtain details
of devid 0 because we don't query devid 0 at all.
reproducer:
btrfs rep start /dev/sdb /dev/sdf /btrfs
btrfs fi show
Label: none uuid: f8fb9819-16c8-47b7-b62f-0ff90f8c56cd
Total devices 3 FS bytes used 1.94GiB
devid 1 size 1.10GiB used 1.10GiB path /dev/sdb
devid 2 size 1.10GiB used 1.08GiB path /dev/sdc
devid 0 size 0.00 used 0.00 path
this patch will make it proper by querying devid 0.
btrfs repl start /dev/sdb /dev/sdf /btrfs
btrfs fi show /btrfs
Label: none uuid: f8fb9819-16c8-47b7-b62f-0ff90f8c56cd
Total devices 3 FS bytes used 1.94GiB
devid 0 size 1.10GiB used 1.10GiB path /dev/sdf
devid 1 size 1.10GiB used 1.10GiB path /dev/sdb
devid 2 size 1.10GiB used 1.08GiB path /dev/sdc
Its fine to query devid 0 when there is no replace
activity as well, because we just skip the error ENODEV
btrfs fi show /btrfs
Label: none uuid: f8fb9819-16c8-47b7-b62f-0ff90f8c56cd
Total devices 2 FS bytes used 1.94GiB
devid 1 size 1.10GiB used 1.10GiB path /dev/sdf
devid 2 size 1.10GiB used 1.08GiB path /dev/sdc
Signed-off-by: Anand Jain <Anand.Jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
We found btrfsck will output backrefs mismatch while the filesystem
is defenitely ok.
The problem is that check_block() don't return right value,which
makes btrfsck won't walk all tree blocks thus we don't get a consistent
filesystem, we will fail to check extent refs etc.
Reported-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Steps to reproduce:
# mkfs.btrfs -f /dev/sda9
# btrfs check /dev/sda9 --init-extent-tree --init-csum-tree
# btrfs check /dev/sda9
During reinitting extent tree, we will pin all metadata blocks to
avoid overwritting existing metadata space. However, those space will
be unpinned after committing transaction.
If we try to reinit csum tree after reiniting extent tree, we may
overwrite existing space. Fix this problem by making reinit extent tree
and csum tree in same transaction.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Fix corporate name for copyright.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Since commit c652e4ef changes default metadata blocksize, update
corresponding options in man page.
Signed-off-by: Rakesh Pandit <rakesh@tuxera.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
During restoring of image (-r using btrfs-image) we zero out RAID
profile in chunk type but forget to save BTRFS_BLOCK_GROUP_DUP if
present. This results in some false messages being printed by btrfsck.
$ ./mkfs.btrfs /dev/sdb2 -f
$ ./btrfs-image /dev/sdb2 btrfs_image_output
$ ./btrfs-image -r btrfs_image_output disk-image
$ ./btrfsck disk-image
Checking filesystem on disk-image
UUID: e644be2d-7701-4bd4-8804-7487f560d2a7
checking extents
Chunk[256, 228, 20971520]: length(8388608), offset(20971520), type(2) mismatch with block group[20971520, 192, 8388608]: offset(8388608), objectid(20971520), flags(34)
Chunk[256, 228, 29360128]: length(1073741824), offset(29360128), type(4) mismatch with block group[29360128, 192, 1073741824]: offset(1073741824), objectid(29360128), flags(36)
Block group[20971520, 8388608] (flags = 34) didn't find the relative chunk.
Block group[29360128, 1073741824] (flags = 36) didn't find the relative chunk.
Even though ./btrfsck on /dev/sdb2 seemed fine. This is due to type
mismatch above and type mismatch occured because we zero'ed out
BTRFS_BLOCK_GROUP_DUP while handling chunk trees.
Signed-off-by: Rakesh Pandit <rakesh@tuxera.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Allow the use of get_device_info() for different units.
Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
When exec btrfsck as non-root user on a disk, btrfsck will always
warn that "No such file or directory", despite that a directory
(e.g. /dev/vboxusb)actually exists. We just have no permission.
In this case, return the -errno set by the opendir call in
btrfs_scan_one_dir rather than blindly return -ENOENT.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
For btrfs-convert, btrfstune, btrfs rescue, they report "device busy"
when given a device that does not actually exist e.g.
# btrfstune -x abcdefg (this device does not exist)
$ ...device busy...
We deal with this case by add "ret < 0" error check when
judging the return value of check_mounted.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Switch to new helper arg_strtou64(), also check if user assign
a valid super copy.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
switch to arg_strtou64 plus some cleanups to remove unnecessary
codes.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
There are many places that need parse string to u64 for btrfs commands,
in fact, we do such things *too casually*, using atoi/atol/atoll..is not
right at all, and even we don't check whether it is a valid string.
Let's do everything more gracefully, we introduce a new helper
arg_strtou64() which will do all the necessary checks.If we fail to
parse string to u64, we will output message and exit directly, this is
something like what usage() is doing. It is ok to not return erro to
it's caller, because this function should be called when parsing arg
(just like usage!)
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Add close_ctree()s before the "returns" on errors after open_ctree()
Also merge the err returns into the "goto + single return" pattern.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Sometimes it is useful to see what btrfs restore is going to do
before provisioning enough external storage to restore onto.
Add a dry-run option so we can see what files and paths are found
by restore, without actually restoring any data.
Signed-off-by: Justin Maggard <jmaggard10@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
The "ret" will be soon used to hold the return value of another function,
assign -1 to it before is nonsense.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
The usage() in help.c calls exit(1), so the break behind is nonsense
and should be removed.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
To be consistent with the other cmds, replace the warning msg
with usage() when send/receive are used without any args.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
The function call that set the ret parameter evaluated in this
BUG_ON was removed in a previous commit:
11be10f71e
Btrfs-progs: make fsck fix certain file extent inconsistencies
Signed-off-by: Mitch Harder <mitch.harder@sabayonlinux.org>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Fix a problem that does not use the result of realpath(), which caused
check_arg_type() can't handle mount point which ends with a final '/'.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Cc: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
The return value in process_one_leaf could be over-written while
looping over the items in the leaf.
This patch will preserve a non-zero return value to the calling
function if a non-zero return value is encountered in the loop.
The return value of one (1) is consistent with non-zero values
that could be returned while processing the leaf.
The only caller of this function (walk_down_tree) would ignore
the return value anyway. But this patch will correct the
behaviour in case future changes intend to utilize the return
value.
Signed-off-by: Mitch Harder <mitch.harder@sabayonlinux.org>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Don't output normal messages into stderr, this make xfstests
filter output easier.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Move find_mount_root to utils.[ch] for general use.
Signed-off-by: Qu Wenruo <quwenruo@cn.fuijitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Change the definition of BUG() to use assert instead of abort to
provide information about the location of the issue.
Signed-off-by: Mitch Harder <mitch.harder@sabayonlinux.org>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
In btrfs/003 of xfstest, it will check whether btrfs fi show can find
missing devices.
But before the patch, btrfs-progs will not check whether device missing
if given a mounted btrfs mountpoint/block device.
This patch fixes the bug and will pass btrfs/003.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Cc: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
A BTRFS_IOC_CLONE_RANGE request with a src_length value of zero has the
effect of cloning all data from src_offset through to end-of-file.
Document this behaviour in the header file for those who (like me)
incorrectly assume that no data is cloned in such a case.
Signed-off-by: David Disseldorp <ddiss@suse.de>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Commit "Btrfs-progs: make send/receive compatible with older kernels"
adds code that will become deprecated, let's clearly mark it in the
sources.
CC: Stefan Behrens <sbehrens@giantdisaster.de>
CC: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
btrfs filesystem show <not-found-label> should return non zero
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
A new test case when disk is unmounted and if the non mapper
disk path is given as the argument to the btrfs filesystem show <arg>
we still need this to work but lblkid will pull only mapper disks,
it won't match. So this will normalize the input to find btrfs
by fsid and pass it to the search.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
With this property, one can enable compression for individual files
without the need to mount the filesystem with the compress or
compress-force options, and specify the compression algorithm.
When applied against a directory, files created under that directory
will inherit the compression property.
This requires the corresponding kernel patch, which adds the support
for setting and getting properties and implements the compression
property.
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
So that we can get the label of a mounted filesystem.
Before this change:
$ btrfs prop get /mnt/btrfs label
ERROR: object is not compatible with property
$ btrfs prop get /dev/sdb3 label
ERROR: dev /dev/sdb3 is mounted, use mount point
ERROR: failed to set/get property for object.
After this change:
$ btrfs prop get /mnt/btrfs label
label=foobar
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Several fixes:
1) The function check_is_root() returns 0 if the object is root;
2) Don't treat any error from get fsid ioctl as meaning the target
is root. Only -ENOTTY means it's a root (parent directory is
not a btrfs fs) and a -ENOTDIR means our target object is not a
directory, therefore it can be the root;
3) Fix the comparison of the target and target's parent fs ids. If
they are different, it means the target is a mount point in a
btrfs fs, therefore it's a root, otherwise it isn't.
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
"btrfs filesystem property" is a generic interface to set/get
properties on filesystem objects (inodes/subvolumes/filesystems
/devs).
This patch adds the generic framework for properties and also
implements two properties. The first is the read-only property
for subvolumes and the second is the label property for devices.
Signed-off-by: Alexander Block <ablock84@googlemail.com>
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Some users complaint that with latest btrfs-progs, they will
fail to use send/receive. The problem is new tool will try
to use uuid tree while it dosen't work on older kernel.
Now we first check if we support uuid tree, if not we fall into
normal search as previous way.i copy most of codes from Alexander
Block's previous codes and did some adjustments to make it work.
Signed-off-by: Alexander Block <ablock84@googlemail.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Steps to reproduce:
# mkfs.btrfs -f /dev/sda8
# mount /dev/sda8 /mnt
# btrfs sub create /mnt/a
# touch /mnt/b
# btrfs sub create /mnt/c
# btrfs sub delete /mnt/*
Above steps will trigger following abortion:
ERROR: 'b' is not a subvolume
*** Error in `btrfs': double free or corruption (out): 0x0000000002116060 ***
======= Backtrace: =========
/lib64/libc.so.6[0x3fa467cef8]
/lib64/libc.so.6(closedir+0xd)[0x3fa46b846d]
btrfs[0x43e608]
btrfs[0x40622f]
btrfs[0x403d19]
btrfs[0x4062c6]
btrfs[0x403f68]
We try to fix it by resetting @fd && @dirstream before trying next
subvolume deletion.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Currently, as of 8cae1840af when running
btrfs-convert I get a bus error.
The problem is that struct btrfs_key has __attribute__ ((__packed__))
so it is not aligned. Then, a pointer to it's objectid field is taken,
cast to a void*, then eventually cast back to a u64* and
dereferenced. The problem is that the dereferenced u64* is not
necessarily aligned (ie, not necessarily a valid u64*), resulting in
undefined behavior.
This patch adds a local u64 variable which would of course be properly
aligned and then uses a pointer to that.
I did not modify the call from btrfs_fs_roots_compare_roots as that
uses struct btrfs_root which is a regular struct and would thus have
it's members correctly aligned to begin with.
After patching this I realized Liu Bo had already written a similar
patch, but I think mine is cleaner, so I'm sending it anyway.
Signed-off-by: Ivan Jager <aij+@mrph.org>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Remove the extraneous `to' from `Can't access to X'.
Signed-off-by: Mitchel Humpherys <mitch.special@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
this patch will make btrfsck operations to open disk in exclusive mode,
so that mount will fail when btrfsck is running
Signed-off-by: Anand Jain <Anand.Jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
The following steps could trigger btrfs segfault:
mkfs -t btrfs -m raid5 -d raid5 /dev/loop{0..3}
losetup -d /dev/loop2
btrfs check /dev/loop0
The reason is that read_tree_block() returns NULL and
add_root_to_pending() dereferences it without checking it first.
Also replace a BUG_ON with proper error checking.
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
for now the manual sync up of new ioctls introduced in the btrfs
kernel. For which there wasn't any btrfs-progs patch.
however we might have better idea for the long run.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
This adds the flag to ctree.h, adds the feature option to mkfs to turn it on and
fixes fsck so it doesn't complain about missing hole extents in files when this
flag is set.
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Skip non-regular files to avoid ioctl errors while defragmenting.
They are silently ignored in recursive mode but reported as errors when
used as command-line arguments.
Signed-off-by: Pascal VITOUX <vitoux.pascal@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Steps to reproduce:
# mkfs.btrfs -f /dev/sda8
# mount /dev/sda8 /mnt
# mkdir /mnt/subvolumes
# btrfs sub create /mnt/subvolumes/subv1
# btrfs sub create /mnt/subvolumes/subv1/subv1.1
# btrfs sub list -o /mnt/subvolumes/subv1 <----we did not list anything
The problem is that we don't set @top_id right, fix it.
Reported-by: Alex <alex@bpmit.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Josef Bacik