We need to build outside of the topdir so we can use the "btrfs/" prefix
for includes and not accidentally include other files.
Make magic is simple:
- build dependencies inside TOPDIR
- build inside temporary directory, link back to TOPDIR
- library-test.o not built anymore obviously
Signed-off-by: David Sterba <dsterba@suse.com>
The library-test is supposed to compile and link as an external
program, so we should update the way the headers are included.
Signed-off-by: David Sterba <dsterba@suse.com>
The patch "btrfs-progs: Introduce kernel sizes to cleanup large
intermediate number" (a2203246ae) was taken from kernel but not
properly ported so the build breaks because the header linux/sizes.h is
not exported.
The build tests of library do not cover the case when the macro
BTRFS_FLAT_INCLUDES is not defined (ie. an external build).
Reported-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Build under musl libc fails because of missing PATH_MAX and XATTR_NAME_MAX
macro declarations. Add the required headers.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Rework rollback to a more easy to understand way.
New convert behavior makes us to have a more flex chunk layout, which
only data chunk containing old fs data will be at the same physical
location, while new chunks (data/meta/sys) can be mapped anywhere else.
This behavior makes old rollback behavior can't handle it.
As old behavior assumes all data/meta is mapped in a large chunk, which is
mapped 1:1 on disk.
So rework rollback to handle new convert behavior, enhance the check by
only checking all file extents of convert image, only to check if these
file extents and therir chunks are mapped 1:1.
This new rollback check behavior can handle both new and old convert
behavior, as the new behavior is a superset of old behavior.
Further more, introduce a simple rollback mechanisim:
1) Read reserved data (offset = file offset) from convert image
2) Write reserved data into disk (offset = physical offset)
Since old fs image is a valid fs, and we only need to rollback
superblocks (btrfs reserved ranges), then we just read out data in
reserved range, and write it back.
Due to the fact that all other file extents of converted image is mapped
1:1 on disk, we put the missing piece back, then the fs is as good as
old one.
Then what we do in btrfs is just another dream.
With this new rollback mechanisim, we can open btrfs read-only, so we
won't cause any damage to current btrfs, until the final piece (0~1M,
containing 1st super block) is put back.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ port to v4.10 ]
Signed-off-by: David Sterba <dsterba@suse.com>
Introduce a function, check_convert_image() to check if that image is
rollback-able.
This means all file extents except one of the btrfs reserved ranges, must
be mapped 1:1 on disk.
1:1 mapped file extents must match the following conditions:
1) Their file_offset(key.offset) matches its disk_bytenr
2) The corresponding chunk must be mapped 1:1 on disk
That's to say, it's a SINGLE chunk, and chunk logical matches with
stripe physical.
Above 2 conditions ensured that file extent lies the exactly the same
position as in the old filesystem.
For data in reserved ranges of btrfs, they are relocated to new places,
and in that case, we use btrfs_read_file() to read out the content for
later rollback use.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Introduce a new function, read_reserved_ranges(), to allow later
rollback to use these data to do rollback.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Introduce a new function, btrfs_read_file(), to read out data of a file
inode.
This function will iterate through EXTENT_DATA items and handle
inline/prealloc/hole file extents.
Compression is not supported yet.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since we have reserved ranges array now, we can use them to skip all
these open codes.
And add some comment and asciidoc art for related part.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ port to v4.10 ]
Signed-off-by: David Sterba <dsterba@suse.com>
Introduce a new strucutre, simple_range, to present one contingous
range.
Also, use such structure to define btrfs_reserved_ranges(), which
convert and rollback will use.
Suggested-by: David Sterba <dsterba@suse.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ split hunks to new file structure ]
Signed-off-by: David Sterba <dsterba@suse.com>
Convert is now a little complex due to that fact we need to separate
metadata and data chunks for different profiles.
Add a comment with ascii art explaining the whole design and point
out the really complex part, so any newcomers interested in convert can
get a quick overview of it before digging into the hard to read code.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ wording and formatting adjustments ]
Signed-off-by: David Sterba <dsterba@suse.com>
clone needs to resolve the paths of the involved subvolumes in the target
fs from their UUIDs. When doing so it might need to strip the prefix
that is mounted as the root of the fs from those paths.
It didn't do so correctly when processing the source of "clone" commands
This is a regression test for
btrfs-progs: receive: handle root subvol path in clone
Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
[ copied the fstests version, will be updated later ]
Signed-off-by: David Sterba <dsterba@suse.com>
testcase:
# ro subvol /src/parent
# rw subvol /src/foo
clone /src/parent/file /src/foo/file
subvol snapshot -r /src/foo /src/foo.snap
# generates a "clone parent/file -> foo.snap/file" send command
send -p /src/parent /src/foo.snap
# target fs:
# dest/
# |--- parent/...
# mounted with -o subvol=dest, such that "parent" is at <target>/parent
receive <target>
result:
ERROR: cannot open dest/parent/file: No such file or directory
expected:
"dest/" get's stripped from the clone source path to get the actual
path in the target fs, if reachable from the mount point/chroot.
This is exactly what process_snapshot does, which gets called on
_every_ incremental receive and I'm quite certain is correct in
doing so
Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Two fixes:
1)
Check that the parent subvol actually is reachable via our root path.
The previous code wouldn't catch
parent subvol: foo/bar
root path: bar (i.e. mounted with -o subvol=bar)
where the parent isn't reachable from the root path.
(but the original "strstr(parent, root_path) == NULL" check still doesn't hold)
Also check for the slash after "root_path", i.e. throw an error on
parent subvol: foobar
root path: foo
2)
If the parent subvol is the one that is mounted we obviously can't
receive into it, as it has to be read-only by definition.
We'd get a rather cryptic:
At subvol /tmp/test/dest.snap
At snapshot dest.snap
ERROR: creating snapshot / -> dest.snap failed: Invalid cross-device link
(not sure what it says if "/" isn't even a btrfs)
But with this we get
At subvol /tmp/test/dest.snap
At snapshot dest.snap
ERROR: creating snapshot . -> dest.snap failed: Read-only file system
which is both more helpful and more correct.
Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
If the final fsync() on the Btrfs device fails, we just swallow the
error and don't alert the user in any way. This was uncovered by xfstest
generic/405, which checks that mkfs fails when it encounters EIO.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Drop the dependency on xfs_io as it's not a standard tool, though it
provides convenience. We use a simple write here so dd can manage.
Signed-off-by: David Sterba <dsterba@suse.com>
If the tests are started from non-root user, the fallocate and xfs_io
fail. Use the root helper as a workaround, we'd should fix the perms
instead.
Signed-off-by: David Sterba <dsterba@suse.com>
In lowmem mode, 'walk_down_tree_v2' returns negative values wheather
the error is fatal or not. It causes the loop where 'walk_down_tree_v2'
is to break even the error is tolerated and then subsequent nodes process
will be skipped.
Fix it by redefining meanings of values 'walk_down_tree_v2' returns.
Do a similar fix for 'process_one_leaf_v2'.
Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Current common.local doesn't handle lowmem mode well.
It passes "--mode=lowmem" alone with "--repair", making it unable to
check lowmem mode.
It's caused by the following bugs:
1) Wrong variable in test/common.local
We should check TEST_ARGS_CHECK, not TEST_CHECK, which is not defined
so we never return 1.
2) Wrong parameter passed to _cmd_spec() in test/common
This prevents us from grepping the correct parameters.
Fix it.
Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Lowmem mode exposed several false alerts, all related to file extents
check.
1) Partly written prealloc extent
Cause lowmem mode to report missing csum or prealloc extent should
not have csum
2) Compressed inline extent
Cause lowmem mode to find mismatch on inline len and item len.
While no error message is output but exit silently.
Reported-by: Chris Murphy <chris@colorremedies.com>
Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fsck-tests/013-extent-tree-rebuild uses "--init-extent-tree", which
implies "--repair".
But the test script doesn't specify "--repair" for lowmem mode test to
detect it.
Add it so lowmem mode test can be happy with it.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add a minimal image which can reproduce the block group used space
false alert for lowmem mode fsck.
Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The test case fsck-tests/015-check-bad-memory-access can't be repair by
btrfs check, and it's a fortunate bug makes original mode to forget the
error code from extent tree, making original mode pass it.
So fuzz-tests is more suitable for it.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
If first inode item is missing, lowmem check will detect it but does not
output any error message.
Add error message for it.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Old lowmem check doesn't check if the inline extent is compressed and
always checks extent numbytes against inline item size.
And when it finds the extent numbytes mismatch with inline item size it
doesn't output any error message, just return error silently, making it
quite hard to debug.
Fix it by only checking extent numbytes against inline item size when
the extent is not compressed, and output error message.
Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
If one extent item has no inline ref, btrfs lowmem mode check can give
false alert without outputting any error message.
The problem is lowmem mode always assumes that extent item has inline
refs, and when it encounters such case it flags the extent item has
wrong size, but doesn't output the error message.
Although we already have such image submitted, at the commit time due to
another bug in cmds-check return value, it doesn't detect it until that
bug is fixed.
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Btrfs lowmem check can report false csum error like:
ERROR: root 5 EXTENT_DATA[257 0] datasum missing
ERROR: root 5 EXTENT_DATA[257 4096] prealloc shouldn't have datasum
This is because lowmem check code always compare the found csum size
with the whole extent which data extents points to.
Normally it's OK, but when prealloc extent is written, or reflink is
done, data extent can points to part of a larger extent, making the csum
check wrong.
To fix it, the csum check part is modified to handle plain and
compressed extents in different ways:
1) Plain extent
Only search csums for the range it refers to.
So the search range is from (disk_bytenr + extent_offset) and search
length is (extent_num_bytes)
2) Compressed extent
Search the whole extent.
Search range is from (disk_bytner) and search length is
(disk_num_bytes)
Reported-by: Chris Murphy <chris@colorremedies.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Although we output error like "errors found in extent allocation tree or
chunk allocation", but we lacks such output for other trees, but leaving
the final "found error is %d" to catch the last return value(and
sometime it's cleared)
This patch adds extra error message for top level error path, and modify
the last "found error is %d" to "error(s) found" or "no error found".
Cc: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since btrfs_search_slot() can point to the slot which is beyond the
leaves' capacity, we should pay extra attention when doing afterward
search.
While for lowmem check, several places uses afterward search:
1) Block group item used space check
2) Device item used space check
3) Data extent backref check.
In the following case for block group item check, btrfs lowmem mode
check will skip the block group and report false alert:
leaf 29405184 items 37 free space 1273 generation 11 owner 2
...
item 36 key (77594624 EXTENT_ITEM 2097152)
extent refs 1 gen 8 flags DATA
extent data backref root 5 objectid 265 offset 0 count 1
leaf 29409280 items 43 free space 670 generation 11 owner 2
item 0 key (96468992 EXTENT_ITEM 2097152)
extent refs 1 gen 8 flags DATA
extent data backref root 5 objectid 274 offset 0 count 1
item 1 key (96468992 BLOCK_GROUP_ITEM 33554432)
block group used 2265088 chunk_objectid 256 flags DATA
When checking block group item, we will search key (96468992 0 0) to
start from the first item in the block group.
While search_slot() will point to leaf 29405184, slot 37 which is beyond
leaf capacity.
And when reading key from slot 37, uninitialized data can be read out
and cause us to exit block group item check, leading to false alert.
Fix it by checking path.slot[0] before reading out the key.
Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Reported-by: Chris Murphy <chris@colorremedies.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The symbol __error was undefined and the rule did not use the
dependencies from the rule definiton, like the rest.
Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba
Baruch Siach
Qu Wenruo
Benedikt Morbach
Omar Sandoval
Su Yue
Lu Fengqi
Nicholas D Steeves
Joey Sabey
Matthias Krüger
Erik Logtenberg