| Message ID | bebfc5c98e0fb878f55c1312d8750ce7b7f82e07.1311776403.git.sbehrens@giantdisaster.de (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Thu, Jul 28, 2011 at 10:28:00AM +0200, Stefan Behrens wrote: > Added btrfs to the list of supported filesystems for test 015, and > increased free space reporting tolerance to 10% for btrfs. > Replaced the call to _scratch_mkfs_xfs with the XFS specific size > parameter by the generic one for sized filesystem creation which is > _scratch_mkfs_sized. ACK for the _scratch_mkfs_sized changed, but I'm really curious why we would allow so much more tolerance for btrfs. And again, why can't these be marked generic? -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 7/28/2011 10:47 AM, Christoph Hellwig wrote: > On Thu, Jul 28, 2011 at 10:28:00AM +0200, Stefan Behrens wrote: >> Added btrfs to the list of supported filesystems for test 015, and >> increased free space reporting tolerance to 10% for btrfs. >> Replaced the call to _scratch_mkfs_xfs with the XFS specific size >> parameter by the generic one for sized filesystem creation which is >> _scratch_mkfs_sized. > > ACK for the _scratch_mkfs_sized changed, but I'm really curious why > we would allow so much more tolerance for btrfs. > > And again, why can't these be marked generic? To add a 10% tolerance for btrfs was a bad idea. Since the output of df(1) is not yet reliable on btrfs volumes while data is not flushed to disk, the better implementation would be to either let this test fail, or to force a flush to disk before taking the output of df(1). The latter is what I have implemented now. This is a demo of what currently happens: # mkfs.btrfs -b 2048M --mixed $SCRATCH_DEV WARNING! - Btrfs v0.19-104-gd0fa1a8-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using Created a data/metadata chunk of size 8388608 fs created label (null) on /dev/sdu1 nodesize 4096 leafsize 4096 sectorsize 4096 size 2.00GB Btrfs v0.19-104-gd0fa1a8-dirty # mount $SCRATCH_DEV $SCRATCH_MNT # df -k $SCRATCH_MNT Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdu1 2097152 28 2093040 1% /mnt3 # dd if=/dev/zero of=$SCRATCH_MNT/foo bs=1M dd: writing `/mnt3/foo': No space left on device 1881+0 records in 1880+0 records out 1971322880 bytes (2.0 GB) copied, 7.61947 s, 259 MB/s # df -k $SCRATCH_MNT Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdu1 2097152 756 2092304 1% /mnt3 # sync # df -k $SCRATCH_MNT Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdu1 2097152 1928420 164640 93% /mnt3 # rm -f $SCRATCH_MNT/foo # df -k $SCRATCH_MNT Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdu1 2097152 1928420 164640 93% /mnt3 # sync # df -k $SCRATCH_MNT Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdu1 2097152 676 2092384 1% /mnt3 -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Jul 28, 2011 at 07:54:45PM +0200, Stefan Behrens wrote: > To add a 10% tolerance for btrfs was a bad idea. > Since the output of df(1) is not yet reliable on btrfs volumes while > data is not flushed to disk, the better implementation would be to > either let this test fail, or to force a flush to disk before taking > the output of df(1). > The latter is what I have implemented now. I don't think it's correct either. dellalloc blocks should be included in the statfs output, else it it is pretty pointless. Can you send the patch to make it generic without that adjustment for now. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/015 b/015 index 8f2be7c..7020e67 100755 --- a/015 +++ b/015 @@ -48,13 +48,13 @@ _free() } # real QA test starts here -_supported_fs xfs +_supported_fs xfs btrfs _supported_os IRIX Linux _require_scratch _require_nobigloopfs -_scratch_mkfs_xfs -d size=50m >/dev/null || _fail "mkfs failed" +_scratch_mkfs_sized `expr 50 \* 1024 \* 1024` >/dev/null || _fail "mkfs failed" _scratch_mount || _fail "mount failed" out=$SCRATCH_MNT/fillup.$$ rm -f $seq.full @@ -120,7 +120,12 @@ fi echo "free space after delete $free2" >> $seq.full echo -n " !!! " -_within_tolerance "free space" $free2 $free0 1% -v +if [ $FSTYP = btrfs ] +then + _within_tolerance "free space" $free2 $free0 10% -v +else + _within_tolerance "free space" $free2 $free0 1% -v +fi status=0 exit
Added btrfs to the list of supported filesystems for test 015, and increased free space reporting tolerance to 10% for btrfs. Replaced the call to _scratch_mkfs_xfs with the XFS specific size parameter by the generic one for sized filesystem creation which is _scratch_mkfs_sized. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> --- 015 | 11 ++++++++--- 1 files changed, 8 insertions(+), 3 deletions(-)