If sectorsize is not BTRFS_SUPER_INFO_SIZE (4k), the superblock checksum
is wrong and mkfs fails. This has been reported on ppc64 where we pick
sectorisize from page size (64k). This has been broken since ages
(2008) and discovered by the recently added superblock checks.
Reported-by: Dinar Valeev <dvaleev@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The loop_info64::lo_file_name might not be null terminated. Avoid strlen
and trim the length to whatever size of the loop_info buffer.
Signed-off-by: David Sterba <dsterba@suse.com>
When using partitions on a loop device, the device's name can be
e.g. /dev/loop0p1 or similar, and no relevant entry exists in the /sys
filesystem, so the current resolve_loop_device function fails.
Instead of using string functions to extract the device name and reading
this file, this patch uses the loop device API through ioctl to get the
correct backing file.
Signed-off-by: Florian Margaine <florian@platform.sh>
[ changed checks of error values from open and ioctl ]
Signed-off-by: David Sterba <dsterba@suse.com>
When do following command in a vm, whose disks are created by
qemu-img create -f raw 11 2.6G:
# mkfs.btrfs -f /dev/vdd /dev/vde /dev/vdf
# btrfs-show-super /dev/vdd /dev/vde /dev/vdf | grep dev_item.total_bytes
dev_item.total_bytes 2791727104
dev_item.total_bytes 2791729152
dev_item.total_bytes 2791729152
We can see that the first device's size is little smaller.
And it fails xfstests btrfs/011.
Reason:
First device's size is rounded down to sectorsize in make_btrfs(),
but other devices are not.
Fix:
Round down remain devices' size in btrfs_add_to_fsid().
Reported-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are two total_bytes in btrfs_add_to_fsid(), local variable
of total_bytes means fs_total_bytes, and device->total_bytes means
device's total_bytes.
And device's total_bytes in argument is named block_count in current
code.
This patch rename:
total_bytes -> fs_total_bytes
block_count -> device_total_bytes
To make code more readable.
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When creating small Btrfs filesystem instances (i.e. filesystem size <= 1GiB),
mkfs.btrfs fails if both sectorsize and nodesize are specified on the command
line and sectorsize != nodesize, since mixed block groups involves both data
and metadata blocks sharing the same block group. This is an incorrect behavior
when '-M' option isn't specified on the command line.
This commit makes optional the creation of mixed block groups i.e. Mixed block
groups are created only when -M option is specified on the command line.
Since we now allow small filesystem instances with sectorsize != nodesize to
be created, we can end up in the following situation,
[root@localhost ~]# mkfs.btrfs -f -n 65536 /dev/loop0
btrfs-progs v3.19-rc2-405-g976307c
See http://btrfs.wiki.kernel.org for more information.
Performing full device TRIM (512.00MiB) ...
Label: (null)
UUID: 49fab72e-0c8b-466b-a3ca-d1bfe56475f0
Node size: 65536
Sector size: 4096
Filesystem size: 512.00MiB
Block group profiles:
Data: single 8.00MiB
Metadata: DUP 40.00MiB
System: DUP 12.00MiB
SSD detected: no
Incompat features: extref, skinny-metadata
Number of devices: 1
Devices:
ID SIZE PATH
1 512.00MiB /dev/loop0
[root@localhost ~]# mount /dev/loop0 /mnt/
mount: mount /dev/loop0 on /mnt failed: No space left on device
The ENOSPC occurs during the creation of the UUID tree. This is because of
things like large metadata block size, DUP mode used for metadata and global
reservation consuming space. Also, large nodesize does not make sense on small
filesystems, hence this should not be an issue.
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Use btrfs_open_dir() in open_path_or_dev_mnt() to make the function
return error when target is neither block device nor btrfs mount point.
Also add "verbose" argument to let function output common error
message instead of putting duplicated lines in caller.
Before patch:
# ./btrfs device stats /mnt/tmp1
ERROR: getting dev info for devstats failed: Inappropriate ioctl for device
# ./btrfs replace start /dev/vdd /dev/vde /mnt/tmp1
ERROR: ioctl(DEV_REPLACE_STATUS) failed on "/mnt/tmp1": Inappropriate ioctl for device
After patch:
# ./btrfs device stats /mnt/tmp1
ERROR: not a btrfs filesystem: /mnt/tmp1
# ./btrfs replace start /dev/vdd /dev/vde /mnt/tmp1
ERROR: not a btrfs filesystem: /mnt/tmp1
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
it was highlighted to me is_block_device(), returns
1 if the file is a block device,
< 0 in case of an error (eg: file not found)
0 otherwise
This patch makes proper return checks at all the places
where is_block_device() is used. Thanks to Goffredo.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Suggested-by: Goffredo Baroncelli <kreijack@inwind.it>
Signed-off-by: David Sterba <dsterba@suse.com>
Check nodesize against features, not only sectorsize.
In fact, one of the btrfs-convert and mkfs differs in the nodesize
check.
This patch also provides the basis for later btrfs-convert fix.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Old kernels before 3.9 do not provide ioctl BTRFS_IOC_GET_FSLABEL.
So we need to provide a fail safe logic for btrfs-progs running
on those kernel.
In this patch when get_label_mounted() fails on the old kernel
it will fail back to the old method and uses get_label_unmounted(),
where it will read from the disk directly.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Sparse reports:
warning: non-ANSI function declaration of function
and we're using func(void) elsewhere.
Signed-off-by: David Sterba <dsterba@suse.com>
Failure during superblock wiping is not always a hard error as we're
going to overwrite it anyway but it may catch some errors earlier. The
error message is not very descriptive though, because we don't get back
much information from blkid.
Signed-off-by: David Sterba <dsterba@suse.com>
We are using separate code for parsing unit mode in current code,
better to use common function.
This patch introduces a common function to specify units as arguments
and a common help message, to make every tool in btrfs having same unit
argument.
The benefits are:
1: Unify current tool's arguments for unit
2: Make tools in future easy to implement such argument
3: Changes (enhancement) in common function have effect on all
relative tools
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In btrfs_add_to_fsid(), strdup() allocates memory to device->name, but
the return value is not checked.
So add the return value check and error handling code.
And clean-up error handling code for ENOMEM case.
Signed-off-by: Byongho Lee <bhlee.kernel@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This patch introduce open_btrfs_dir() to open a dir in btrfs
filesystem.
It can be used for several tools in btrfs-progs.
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
[renamed from open_btrfs_dir, adjusted error messages]
Signed-off-by: David Sterba <dsterba@suse.com>
At some places we do not clear the whole ioctl structure and could
pass garbage to kernel. Zero the ioctl vol_args and use a helper for
copying the path.
Signed-off-by: David Sterba <dsterba@suse.cz>
The path bufferes should be PATH_MAX but BTRFS_PATH_NAME_MAX is shorter
due to embedding in 4k aligned structures.
The only reason to use BTRFS_PATH_NAME_MAX is for the respective
structures btrfs_ioctl_vol_args::name.
Signed-off-by: David Sterba <dsterba@suse.cz>
Anand reports that the static buffers used for pertty size strings cause
a stack overflow on SPARC. Zach proposed to change the printf format to
wrap the number and the suffix into a macro. This would require to
change all callsites of pretty_size* and is not very convienient to
write.
This patch replaces the per-call-site static buffers with a limited
number for slots that would be used on each invokation of pretty_size
and wrap around. The number of array slots shall be 10 for now, in
current codebase there are no more than 2 calls to pretty_size in a
single argument list.
Reported-by: Anand Jain <Anand.Jain@oracle.com>
CC: Zach Brown <zab@zabbo.net>
Signed-off-by: David Sterba <dsterba@suse.cz>
The error string buffer passed as an argument is of a fixed size, though
we could print up to PATH_MAX + something bytes. Print the error message
directly.
Signed-off-by: David Sterba <dsterba@suse.cz>
The error string buffer passed as an argument is of a fixed size, though
we could print up to PATH_MAX + something bytes. Print the error message
directly.
Signed-off-by: David Sterba <dsterba@suse.cz>
We should zero the dirstream in case we close it ourselves, so the
caller does not do it again. Most current callers do not do that and
exit immediatelly.
Signed-off-by: David Sterba <dsterba@suse.cz>
The variable argv0_buf was duplicated and the changes done in utils.c
were not propagated to help.c. So if an unknown commandline token was
found, the error message did not contain the known part:
$ btrfs scrub test
: unknown token 'test'
instead of
$ btrfs scrub test
btrfs scrub: uknown token 'test'
Signed-off-by: David Sterba <dsterba@suse.cz>
The function make_btrfs() has as argument the fsid of the filesystem.
If this fsid is empty or null make_btrfs() generates a new fsid. However
If the buffer is valid (but the string is empty) the generated fsid is
copied back to the caller.
Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
Reviewed-by: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
When creating a new btrfs_device, copy the path to track it. This path
is then used by mkfs.btrfs to list all devices.
Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
Reviewed-by: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
Currently BTRFS allows you to make bad choices of data and
metadata levels. For example -d raid1 -m raid0 means you can
only use half your total disk space, but will lose everything
if 1 disk fails. It should give a warning in these cases.
When making a filesystem, check that metadata mode is at least
as redundant as the data mode. For example give warning when:
-d raid1 -m raid0
Signed-off-by: Sam Tygier <samtygier@yahoo.co.uk>
[make the check more visible in mkfs output]
Signed-off-by: David Sterba <dsterba@suse.cz>
After Patch:
remove BTRFS_SCAN_PROC scan method
There isn't any consumer for btrfs_scan_block_devices() so delete it.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
mkfs and convert will not support the same features, -O will print only
the list according to the given mask.
Signed-off-by: David Sterba <dsterba@suse.cz>
Now parse_qgroupid() can resolve subvolume path into qgroupid.
This is quite handy for handling level 0 qgroupid, and user don't need
to resolve rootid by hand now.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[constify string in __is_subvol]
Signed-off-by: David Sterba <dsterba@suse.cz>
check_node_or_leaf_size in utils.c now prints 'nodesize (or leafsize)'
instead of 'leafsize (or nodesize)' in the error messages, in order to
be less confusing for the user, as leafsize in mkfs is deprecated.
'ERROR: ' is also prepended to be consistent with other error messages.
Signed-off-by: Sebastian Thorarensen <sebth@naju.se>
Signed-off-by: David Sterba <dsterba@suse.cz>
Move the constant DEFAULT_MKFS_LEAF_SIZE to utils.h and rename it to
BTRFS_MKFS_DEFAULT_NODE_SIZE for consistency. Move the function
check_leaf_or_node_size to utils.c and rename it to
btrfs_check_node_or_leaf_size.
Signed-off-by: Sebastian Thorarensen <sebth@naju.se>
[added btrfs_ prefix]
Signed-off-by: David Sterba <dsterba@suse.cz>
glibc 2.10+ (5+ years old) enables all the desired features:
_XOPEN_SOURCE 700, __XOPEN2K8, POSIX_C_SOURCE, DEFAULT_SOURCE; with a
single _GNU_SOURCE define in the makefile alone. For portability to
other libc implementations (e.g. dietlibc) _XOPEN_SOURCE=700 is also
defined.
This also resolves Debian bug report filed by Michael Tautschnig -
"Inconsistent use of _XOPEN_SOURCE results in conflicting
declarations". Whilst I was not able to reproduce the results, the
reported fact is that _XOPEN_SOURCE set to 500 in one set of files
(e.g. cmds-filesystem.c) generates/defines different struct stat from
other files (cmds-replace.c).
This patch thus cleans up all feature defines, and sets them at a
consistent level.
Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=747969
Signed-off-by: Dimitri John Ledkov <dimitri.j.ledkov@intel.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
The newly introduced search_chunk_tree_for_fs_info() won't count devid 0
in fi_arg->num_devices, which will cause buffer overflow since later
get_device_info() will fill di_args with devid.
This can be trigger by fstests/btrfs/069 and any operations needs to
iterate over all the devices like 'fi show' or 'dev stat' while
replacing.
The fix is do an extra probe specifically for devid 0 after
search_chunk_tree_for_fs_info() and change num_devices if needed.
Reported-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
The commit 1bad43fbe0 ("btrfs-progs: refine btrfs-debug-tree error
prompt when a mount point given")
add judgement on btrfs-debug-tree to restrict only block device to be
executed on, but the command can also be used on regular file, so add
regular file support for the judgement.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.cz>
David Sterba