Ancient commit f4c697e6406d ("btrfs: return EINVAL if start >
total_bytes in fitrim ioctl") introduced a regression where btrfs
may fail to trim any free space in existing block groups.
It's caused by confusion with btrfs_super_block->total_bytes and
btrfs logical address space.
Unlike physical address, any aligned bytenr in range [0, U64_MAX) is
valid in btrfs logical address space, and it's chunk mapping
mechanism of btrfs to handle the logical<->physical mapping.
The test case will craft a btrfs with the following features:
0) Single data/meta profile
Make trimmed bytes reporting and chunk allocation more predictable.
1) All chunks start beyond super_block->total_bytes (1G)
By relocating these blocks several times.
2) Unallocated space is less than 50% of the whole fs
3) Fragmented data chunks
Data chunks will be full of fragments, 50% of data chunks will be
free space.
So in theory fstrim should be able to trim over 50% space of the fs.
(after fix, 64% of the fs can be trimmed)
While the regression makes btrfs only able to trim unallocated
space, which is less than 50% of the total space.
(without fix, it's only 31%)
Fixed by patch named "btrfs: Ensure btrfs_trim_fs can trim the whole
fs".
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
The check for empty index in this test was wrongly copied from test
overlay/034. In test overlay/034 lower file starts as a hardlink, so
nlink accounting is done from the first copy up and index can be cleaned
on last upper hardlink unlink. In this test, lower starts as
non-hardlink, so first copy up does not perform nlink accounting and
therefore, union nlink count does not drop to 0 at the end of the test
and the index is not expected to be cleaned.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Test nlink accounting of overlay hardlinks with offline modifications.
nlink of overlay inode should account for the union of lower and upper
hardlinks. Orphan index inodes with union nlink 0 should be cleaned on
mount.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
- file A and B are hardlinked in lower
- modify A to trigger copy up and index lower
- unlink A and mount cycle
- check that B still contains the modified data
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Use TIME_FACTOR to increase the amount of time we spend testing log
shutdown recovery.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Test that an incremental send operation works if a file that has
multiple hard links has some of its hard links renamed in the send
snapshot, with one of them getting the same path that some other
inode had in the send snapshot.
At the moment this test fails on btrfs and a fix is provived by a
linux kernel patch titled:
"Btrfs: incremental send, fix wrong unlink path after renaming file"
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
I got the following message when running generic/465 in ext4
data=journal mode
---------------------------------------------------------------
QA output created by 465
non-aio dio test
encounter an error: block 0 offset 4096, content 62
encounter an error: block 0 offset 122880, content 62
encounter an error: block 0 offset 274432, content 62
encounter an error: block 0 offset 86016, content 62
aio-dio test
encounter an error: block 0 offset 28672, content 62
encounter an error: block 0 offset 12288, content 62
encounter an error: block 2 offset 16384, content 62
encounter an error: block 1 offset 565248, content 62
---------------------------------------------------------------
In ext4 data=journal mode, direct read will fall back to buffer
read, and buffer read doesn't take inode lock, so it doesn't need to
wait for the writer to finish first and sees the intermediate inode
size and returns data less than 1M.
We can just check the actual read data instead of the whole read
buffer.
Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
util-linux commit 6dede2f2f7c5 ("libmount: support MS_RDONLY on
write-protected devices") changed the error message on read-only
block device, and in the failure case printed one line message
instead of two (for details please see comments in common/filter),
and this change broke generic/050 and overlay/035.
Fix it by adding more filter rules to _filter_ro_mount and updating
associated .out files to unify the output from both old and new
util-linux versions.
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
util-linux v2.30 changed error message of a busy mount and caused
overlay/036 to fail. e.g.
- mount: <device> is already mounted or <mountpoint> busy
+ mount: <mountpoint>: <device> already mounted or mount point busy.
Filter the mount output by a newly introduced _filter_busy_mount
into a unified format.
mount: device already mounted or mount point busy
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
util-linux commit ea848180dd34 ("libmount: add
mnt_context_get_excode()") since v2.30 changed the error message on
EUCLEAN and ESTALE again (and maybe other errno too):
- mount: <device> on <mountpoint> failed: Structure needs cleaning
+ mount: <mountpoint>: mount(2) system call failed: Structure needs cleaning.
and it causes xfs/005, overlay/037 to fail (and probably xfs/333 too,
but it's always _notrun for now).
And what's more, the mentioned tests would also fail when testing
with util-linux prior to v2.21, no one complained just because the
tests are usually _notrun on such old distributions that ship
util-linux < v2.21.
So let's filter out the changing parts and keep the error message
simple.
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Add a link to Ted's post with more details about the bug
and the fix.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This commit implements similar changes made to overlay/038 test i.e. in
addition to checking consistent values of st_ino/d_ino for "." and ".."
entries, check also the values for subdir inside each tested directory
setup.
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This commit implements similar changes made to overlay/041 test i.e.
- Remove duplicate merged dir tests
- Remove duplicate pure upper dir inside merged dir tests
- Sort out comments and error messages
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This uses the new fio results perf helpers to run a rand write
buffered workload on the scratch device.
[eguan: add Makefile]
Signed-off-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This patch does the nuts and bolts of grabbing fio results and
storing them in a database in order to check against for future
runs. This works by storing the results in resuts/fio-results.db as
a sqlite database. The src/perf directory has all the supporting
python code for parsing the fio json results, storing it in the
database, and loading previous results from the database to compare
with the current results.
This also adds a PERF_CONFIGNAME option that must be set for this to
work. Since we all have various ways we run fstests it doesn't make
sense to compare different configurations with each other (unless
specifically desired). The PERF_CONFIGNAME will allow us to
separate out results for different test run configurations to make
sure we're comparing results correctly.
Currently we only check against the last perf result. In the future
I will flesh this out to compare against the average of N number of
runs to be a little more complete, and hopefully that will allow us
to also watch latencies as well.
[eguan: add required Makefile updates]
Signed-off-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Mount non-samefs dirs with OVERLAY_MOUNT_OPTIONS, so user can opt-in
for constant inode numbers on non-samefs setup (i.e. -o xino) when
it is implemented.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
In addition to checking consistent values of st_ino/d_ino for "."
and ".." entries, check also the values for subdir inside each
tested directory setup.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
- Remove duplicate merged dir tests
- Remove duplicate pure upper dir inside merged dir tests
- Sort out comments and error messages
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This commit adds a test to check hardlink breakage when the
overlayfs instance is composed of two different underlying
filesystem instances.
For example,
$ mount -t xfs /dev/loop0 /mnt/test
$ mount -t xfs /dev/loop1 /mnt/scratch
$ mkdir /mnt/scratch/upper
$ mkdir /mnt/scratch/work
$ mount -t overlay overlay -o lowerdir=/mnt/test \
-o upperdir=/mnt/scratch/upper \
-o workdir=/mnt/scratch/work /mnt/merge
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This commit adds a test to verify consistent st_ino feature when
the overlayfs instance is composed of two different underlying
filesystem instances.
For example,
$ mount -t xfs /dev/loop0 /mnt/test
$ mount -t xfs /dev/loop1 /mnt/scratch
$ mkdir /mnt/scratch/upper
$ mkdir /mnt/scratch/work
$ mount -t overlay overlay -o lowerdir=/mnt/test \
-o upperdir=/mnt/scratch/upper \
-o workdir=/mnt/scratch/work /mnt/merge
The goal of this test is to verify that overlayfs returns consistent
st_ino for the following scenarios,
- Copy-up of lowerdir files
- Rename files and drop dentry/inode cache
- Remount the overlayfs instance
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Constant d_ino for hardlinks is possible only when "index" config
feature/mount option is enabled.
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Constant st_ino/d_ino for hardlinks requires "index" config
feature/mount option to be enabled. overlay/018 already tests
constant st_ino feature for hardlinks. A future commit will add
constant d_ino test to overlay/018 test.
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
From mmap(2) manpage, "a file is mapped in multiples of the page
size. For a file that is not a multiple of the page size, the
remaining memory is zeroed when mapped", this test is to test this
behavior on truncate down.
This is inspired by an XFS bug that truncate down fails to zero page
cache beyond new EOF and causes stale data written to disk
unexpectedly and a subsequent mmap sees non-zeros post EOF.
Patch "xfs: truncate pagecache before writeback in
xfs_setattr_size()" fixed the bug on XFS.
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Qu Wenruo
Amir Goldstein
Darrick J. Wong
Filipe Manana
Xiao Yang
Eryu Guan
Chandan Rajendra
Josef Bacik