A user had a corrupt fs where one of his file extents pointed to a completely
bogus disk bytenr. This patch allows us to corrupt a file system in a similar
way in order to test btrfsck. Thanks,
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
If you set an file extent item's disk_bytenr to something completely wrong we
won't be able to fix this if it is the only one who has a ref on the original
disk bytenr. Our extent records know exactly who is supposed to point at them,
so if we have an extent record that has no backrefs we can go and try to lookup
the backrefs ourselves. If these backrefs do not point to an extent record that
was actually found then we can be pretty sure this extent record is valid and
the backref is bogus. Then the verify_backref code can do its thing and reset
the backref to point to the right extent record and we can all carry on. This
fixes a user reported corruption. Thanks,
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
A user reported a problem where he was unable to rmdir an empty directory. This
is because his isize was wrong. This patch will fix this sort of corruption and
allow him to rmdir his directory. Thanks
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
A user reported a problem with his fs where he had a bogus isize on his
directory. In order to make sure my patch for fsck fixes this properly I needed
to be able to corrupt an inode like this, which is what this patch is for.
Eventually I want to extend this to corrupt everything so we can integrate tests
into btrfs-progs to run btrfsck against to make sure we don't regress on fixing
things with btrfsck. Thanks,
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Raid5 and raid6 at least need three and foure devices respectively,
fix it.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
I found that mkfs.btrfs aborts when assigned multi volumes contain
a small volume:
# parted /dev/sdf p
Model: LSI MegaRAID SAS RMB (scsi)
Disk /dev/sdf: 72.8GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos
Number Start End Size Type File system Flags
1 32.3kB 72.4GB 72.4GB primary
2 72.4GB 72.8GB 461MB primary
# ./mkfs.btrfs -f /dev/sdf1 /dev/sdf2
:
SMALL VOLUME: forcing mixed metadata/data groups
adding device /dev/sdf2 id 2
mkfs.btrfs: volumes.c:852: btrfs_alloc_chunk: Assertion `!(ret)' failed.
Aborted (core dumped)
This failure of btrfs_alloc_chunk was caused by following steps:
1) since there is only small space in the small device, mkfs was
going to allocate a chunk from free space as much as available.
So mkfs called btrfs_alloc_chunk with
size = device->total_bytes - device->used_bytes.
2) (According to the comment in source code, to avoid overwriting
superblock,) btrfs_alloc_chunk starts taking chunks at an offset
of 1MB. It means that the layout of a disk will be like:
[[1MB at beginning for sb][allocated chunks]* ... free space ... ]
and you can see that the available free space for allocation is:
avail = device->total_bytes - device->used_bytes - 1MB.
3) Therefore there is only free space 1MB less than requested. damn.
>From further investigations I also found that this issue is easily
reproduced by using -A, --alloc-start option:
# truncate --size=1G testfile
# ./mkfs.btrfs -A900M -f testfile
:
mkfs.btrfs: volumes.c:852: btrfs_alloc_chunk: Assertion `!(ret)' failed.
Aborted (core dumped)
In this case there is only 100MB for allocation but btrfs_alloc_chunk
was going to allocate more than the 100MB.
The root cause of both of above troubles is a same simple bug:
btrfs_chunk_alloc does not calculate available bytes properly even
though it researches how many devices have enough room to have a
chunk to be allocated.
So this patch introduces new function btrfs_device_avail_bytes()
which returns available bytes for allocation in specified device.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The previous patch works fine if the size of specified volume to mkfs
is less than 4MB. However usually btrfs requires more than 4MB to work,
and the minimum preferred size is depending on the raid setting etc.
This patch let mkfs print error message if it cannot allocate one of
chunks should be there at first.
[before]
# truncate --size=4500K testfile
# ./mkfs.btrfs -f testfile
:
SMALL VOLUME: forcing mixed metadata/data groups
mkfs.btrfs: mkfs.c:84: make_root_dir: Assertion `!(ret)' failed.
Aborted (core dumped)
[After]
# truncate --size=4500K testfile
# ./mkfs.btrfs -f testfile
:
SMALL VOLUME: forcing mixed metadata/data groups
no space to alloc data/metadata chunk
failed to setup the root directory
TBD is calculate minimum size for setting and put it in the error
message to let user know how large amount of volume is required.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Eric pointed out that mkfs abort if specified volume is too small:
# truncate --size=2m testfile
# ./mkfs.btrfs testfile
:
SMALL VOLUME: forcing mixed metadata/data groups
mkfs.btrfs: volumes.c:852: btrfs_alloc_chunk: Assertion `!(ret)' failed.
Aborted (core dumped)
As the first step to fix problems around there, let mkfs to report
error if the size of target volume is less than the size of the first
system block group, BTRFS_MKFS_SYSTEM_GROUP_SIZE (= 4MB).
Reported-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
As a result of a successful call to btrfs_read_sys_array(), the 'ret'
variable is already set to 0. Hence the function would return 0 even
if the call to read_tree_block() fails.
Signed-off-by: chandan <chandan@linux.vnet.ibm.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The strdup()s not freed are reported as memory leaks by valgrind.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Originally the local pending_list is not guaranteed to be freed upon
fails, it should be emptyed and the elements should be freed.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The seen cache_tree in run_next_block freed.
Originally, this "missing" causes memory leaks, reported by valgrind.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The local probe variable in is_ssd() freed upon unsuccessful return;
The local dir_head list in make_image() freed upon unsuccessful return.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The variable "buf" passed into find_collision() as parameter "name"
should be freed on unsuccessful returns.
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
There will be four kinds of return value for command "scrub start":
0: scrub dosen't find errors and return success.
1: usage or syntax errors.
3: scrub finds errors and correct all of them.
4: scrub finds errors and some of them are not correctable.
Three kinds of return values for scrub cancel/resume:
0: cancel successfully.
1: usage or syntax errors.
2: cancel a not started or finished scrub.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
There are 3 kinds of return values in replace cancel:
0: cancel successfully.
1: usage or syntal errors
2: cancel a not started or finished replacing operations.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
If btrfs send return failure, we return 1，otherwise 0 will be returned.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The patch also fixes some coding styles problems.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
if usage or syntax error happens, we return 1.
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
mkfs.c: In function ‘is_ssd’:
mkfs.c:1168:26: warning: ignoring return value of ‘blkid_devno_to_wholedisk’,
declared with attribute warn_unused_result [-Wunused-result]
blkid_devno_to_wholedisk(devno, wholedisk, sizeof(wholedisk), NULL);
Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
This fix the regression introduced by 830427d
that it no more creates the FS if disk is small
and if no mixed option is provided.
This patch will bring it to the original design
which will force mixed profile when disk is small
and go ahead to create the FS.
Which also means that before we open the device
for the write we should also check if disk is small.
v2: fixes the checkpatch.pl warnings
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
This patch provides fix for the following bug,
When mkfs.btrfs fails the disks shouldn't be written.
------------
btrfs fi show /dev/sdb
Label: none uuid: 60fb76f4-3b4d-4632-a7da-6a44dea5573d
Total devices 1 FS bytes used 24.00KiB
devid 1 size 2.00GiB used 20.00MiB path /dev/sdb
mkfs.btrfs -dsingle -mraid1 /dev/sdb -f
::
unable to create FS with metadata profile 16 (have 1 devices)
btrfs fi show /dev/sdb
Label: none uuid: 2da2179d-ecb1-4a4e-a44d-e7613a08c18d
Total devices 1 FS bytes used 24.00KiB
devid 1 size 2.00GiB used 20.00MiB path /dev/sdb
-------------
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
When there are files that have parts shared with snapshots, the
restore command was incorrectly restoring them, as it was not
taking into account the offset and number of bytes fields from
the file extent item. Besides leaving the recovered file corrupt,
it was also inneficient as it read and wrote more data than needed
(with each extent copy overwriting portions of the one previously
written).
The following steps show how to reproduce this corruption issue:
$ mkfs.btrfs -f /dev/sdb3
$ mount /dev/sdb3 /mnt/btrfs
$ perl -e '$d = "\x41" . ("\x00" x (1024*1024+349)); open($f,">","/mnt/btrfs/foobar"); print $f $d; close($f);'
$ du -b /mnt/btrfs/foobar
1048926 /mnt/btrfs/foobar
$ md5sum /mnt/btrfs/foobar
f9f778f3a7410c40e4ed104a3a63c3c4 /mnt/btrfs/foobar
$ btrfs subvolume snapshot /mnt/btrfs /mnt/btrfs/my_snap
$ perl -e 'open($f, "+<", "/mnt/btrfs/foobar"); seek($f, 4096, 0); print $f "\xff"; close($f);'
$ md5sum /mnt/btrfs/foobar
b983fcefd4622a03a78936484c40272b /mnt/btrfs/foobar
$ umount /mnt/btrfs
$ btrfs restore /dev/sdb3 /tmp/copy
$ du -b /tmp/copy/foobar
1048926 /tmp/copy/foobar
$ md5sum /tmp/copy/foobar
88db338cbc1c44dfabae083f1ce642d5 /tmp/copy/foobar
$ od -t x1 -j 8192 -N 4 /tmp/copy/foobar
0020000 41 00 00 00
0020004
$ mount /dev/sdb3 /mnt/btrfs
$ od -t x1 -j 8192 -N 4 /mnt/btrfs/foobar
0020000 00 00 00 00
0020004
$ md5sum /mnt/btrfs/foobar
b983fcefd4622a03a78936484c40272b /mnt/btrfs/foobar
Tested this change with zlib, lzo compression and file sizes larger
than 1GiB, and found no regression or other corruption issues (so far
at least).
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Preparatory patch to move cmd & test files into their
own subdirs.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Before this change, passing -O skinny-metadata to mkfs.btrfs would
only set the skinny metadata incompat flag in the super block after
the filesystem was created. This change makes mkfs.btrfs directly
create a filesystem with only skinny extents for metadata.
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
This is a recent flag added to the restore command that allows
to restore xattrs. It was missing in the man page.
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
There were a few problems that were breaking sparse checking:
- We were defining CHECK_ENDIAN late in the environment, after
linux/fs.h has been included which defines __force and __bitwise in
confusing ways that conflict with ours. Define it up with __CHECKER__
so that linux/fs.h and our copy are acting on the same input.
- We had manually set a few of gcc's internal defines to give to sparse.
It's easier to just ask gcc for all the defines it sets and hand those
to sparse.
- We weren't passing the same *FLAGS to sparse as we were to CC.
- glibc has so many errors with FORTIFY turned on that sparse gives up
and doesn't show us any errors from our code. It's a questionable
hack to always turn on FORTIFY ourselves, so we'll just not do that
when building with sparse.
And add a nice '[SP]' quiet output line for sparse checks.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
These were mostly in option structs but there were a few gross string
pointer arguments given as 0.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
raid6.c is built without access to the prototypes of functions it
exports.
warning: symbol 'raid6_gen_syndrome' was not declared. Should it be static?
They could be changed and get out of sync of the exported prototypes
without errors. So we add disk-io.h, and its dependency ctree.h, so
that it has a chance to check that its exported prototypes are correct.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
sparse can freak out when <linux/fs.h> is included because it redefines
approximately a gazillion symbols already found in <sys/mount.h>:
/usr/include/linux/fs.h:203:9: warning: preprocessor token MS_RDONLY redefined
/usr/include/sys/mount.h:37:9: this was the original definition
Happily, we don't actually need to include the low-level <linux/fs.h>
for anything. One assumes it was just carried over from kernel space.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
The _una_ struct's entire job is to pass an argument to le*_to_cpu. So
it's a little embarassing that it uses a native cpu types and generates
endian warnings.
ctree.h:1616:1: warning: incorrect type in assignment (different base types)
ctree.h:1616:1: expected unsigned long long [unsigned] [usertype] x
ctree.h:1616:1: got restricted __le64 [usertype] <noident>
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
qgroup.c:82:23: warning: memcpy with byte count of 0
qgroup.c:83:23: warning: memcpy with byte count of 0
The inheritance wasn't copying qgroups[] because a confused sizeof()
gave 0 byte memcpy()s. It's been like this for the year since it was
merged, so I guess this isn't a very important thing to do :).
Signed-off-by: Zach Brown <zab@redhat.com>
Reviewed-by: Arne Jansen <sensille@gmx.net>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Storing fixed-endian values in native cpu types defeats the purpose of
using sparse endian types to find endian conversion bugs.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Extents rebuilt from backrefs can have their objectid mangled. The code
tried to build a disk_key by hand and got the swabbing backwards.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
A disk_key was set by hand instead of using the endian helpers.
I *think* the second one is just a typo. The chunk's num_stripes was
already initialized from the record, but it's le16. So we'll set the
item's size based on the record's native num_stripes.
Signed-off-by: Zach Brown <zab@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
