On Wed, May 11, 2016 at 10:14:42AM +0800, Qu Wenruo wrote:
>
>
> Filipe Manana wrote on 2016/05/10 11:01 +0100:
> >On Tue, May 10, 2016 at 9:39 AM, Qu Wenruo <quwenruo@xxxxxxxxxxxxxx> wrote:
> >>For a completely deduped file, which means all its file extent are
> >>pointing to one bytenr, if calling fiemap on it, btrfs will cause soft
> >>hang up or just takes years long.
> >>
> >>This bug can be reproduced even without any in-band or out-of-band
> >>dedupe, normal clone_file_range() call can create such situation.
> >>
> >>This test case will detect it.
> >
> >Why isn't this a generic test?
> >There's nothing btrfs specific anymore...
> >
> >Thanks.
>
> I'm OK to move it to generic, just as original planned.
Thank you!
> BTW, does other fs support reflink file range?
As Christoph said, future-XFS and NFS.
> I found a lot xfs test cases using reflink, but I still can't reflink a file
> range inside the same inode
> ---
> $ xfs_io -c "reflink test.file 0 128k 128k" test.file
> XFS_IOC_CLONE_RANGE: Operation not supported
<shrug> It should work...
...and currently works for me (4.6-rc7) on both btrfs and xfs:
# rm -rf a ; dd if=/dev/zero of=a bs=131072 count=1 ; xfs_io -c 'reflink a 0 128k 128k' a ; filefrag -v a ; grep $PWD /proc/mounts
1+0 records in
1+0 records out
131072 bytes (131 kB, 128 KiB) copied, 0.000539818 s, 243 MB/s
linked 131072/131072 bytes at offset 131072
128 KiB, 1 ops; 0.0000 sec (120.077 MiB/sec and 960.6148 ops/sec)
Filesystem type is: 9123683e
File size of a is 262144 (64 blocks of 4096 bytes)
ext: logical_offset: physical_offset: length: expected: flags:
0: 0.. 31: 3088.. 3119: 32:
1: 32.. 63: 3088.. 3119: 32: 3120: last,eof
a: 2 extents found
/dev/sda /mnt btrfs rw,relatime,space_cache,subvolid=5,subvol=/ 0 0
# cd /opt
# rm -rf a ; dd if=/dev/zero of=a bs=131072 count=1 ; xfs_io -c 'reflink a 0 128k 128k' a ; filefrag -v a ; grep $PWD /proc/mounts
1+0 records in
1+0 records out
131072 bytes (131 kB, 128 KiB) copied, 0.00237377 s, 55.2 MB/s
linked 131072/131072 bytes at offset 131072
128 KiB, 1 ops; 0.0000 sec (87.047 MiB/sec and 696.3788 ops/sec)
Filesystem type is: 58465342
File size of a is 262144 (64 blocks of 4096 bytes)
ext: logical_offset: physical_offset: length: expected: flags:
0: 0.. 31: 24.. 55: 32: shared
1: 32.. 63: 24.. 55: 32: 56: last,shared,eof
a: 2 extents found
/dev/sdb /opt xfs rw,relatime,attr2,inode64,noquota 0 0
That said, I haven't checked with latest xfsprogs master.
--D
> ---
>
> >
> >>
> >>Reported-by: Tsutomu Itoh <t-itoh@xxxxxxxxxxxxxx>
> >>Signed-off-by: Qu Wenruo <quwenruo@xxxxxxxxxxxxxx>
> >>---
> >> tests/btrfs/028 | 78 +++++++++++++++++++++++++++++++++++++++++++++++++++++
> >> tests/btrfs/028.out | 3 +++
> >> tests/btrfs/group | 1 +
> >> 3 files changed, 82 insertions(+)
> >> create mode 100755 tests/btrfs/028
> >> create mode 100644 tests/btrfs/028.out
> >>
> >>diff --git a/tests/btrfs/028 b/tests/btrfs/028
> >>new file mode 100755
> >>index 0000000..62bcc9d
> >>--- /dev/null
> >>+++ b/tests/btrfs/028
> >>@@ -0,0 +1,78 @@
> >>+#! /bin/bash
> >>+# FS QA Test 028
> >>+#
> >>+# Test fiemap ioctl on heavily deduped file.
> >>+#
> >>+# This test will cause btrfs to soft hang up or takes years long to finish
> >
> >Haven't tried it, but I doubt it will take years...
> >Are you sure that the soft lookup, which is what makes the test fail
> >due to the dmesg warning, is triggered on very fast machines as well?
> >I.e. this may not be reliable on better hardware.
>
> On a fast test server too, using the same test case, but your concern is
> valid.
>
> The reporter initially triggered the bug on a even faster server with
> similar file layout with 100% possibility, but with nr set to 8192.
>
> I reduced the nr from 8192 (which is always reproducible) to 4096 to save
> some time creating file, but considering the scale of loops, considering the
> loop scale (at least n^3), the halved nr seems to hugely reduce the time.
>
> The know loop scale is n^3 ~ n^4:
> 1. Loop all file extents (* 4096)
> 2. Loop all backrefs of one extent (* 4096)
> 3. Loop each backref in __merge_refs(list_for_each_entry_safe_continue) (*
> 4096)
> 4. Loop to the list end in "while(eie & eie->next) {eie=eie->next}" (*4096)
>
> What about change nr to (8192 * $LOAD_FACTOR)?
>
> Thanks,
> Qu
>
>
> Thanks,
> Qu
>
> >
> >
> >>+#
> >>+#-----------------------------------------------------------------------
> >>+# Copyright (c) 2016 Fujitsu. All Rights Reserved.
> >>+#
> >>+# This program is free software; you can redistribute it and/or
> >>+# modify it under the terms of the GNU General Public License as
> >>+# published by the Free Software Foundation.
> >>+#
> >>+# This program is distributed in the hope that it would be useful,
> >>+# but WITHOUT ANY WARRANTY; without even the implied warranty of
> >>+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> >>+# GNU General Public License for more details.
> >>+#
> >>+# You should have received a copy of the GNU General Public License
> >>+# along with this program; if not, write the Free Software Foundation,
> >>+# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
> >>+#-----------------------------------------------------------------------
> >>+#
> >>+
> >>+seq=`basename $0`
> >>+seqres=$RESULT_DIR/$seq
> >>+echo "QA output created by $seq"
> >>+
> >>+here=`pwd`
> >>+tmp=/tmp/$$
> >>+status=1 # failure is the default!
> >>+trap "_cleanup; exit \$status" 0 1 2 3 15
> >>+
> >>+_cleanup()
> >>+{
> >>+ cd /
> >>+ rm -f $tmp.*
> >>+}
> >>+
> >>+# get standard environment, filters and checks
> >>+. ./common/rc
> >>+. ./common/filter
> >>+. ./common/reflink
> >>+
> >>+# remove previous $seqres.full before test
> >>+rm -f $seqres.full
> >>+
> >>+# real QA test starts here
> >>+
> >>+# Modify as appropriate.
> >>+_supported_fs btrfs
> >>+_supported_os Linux
> >>+_require_scratch_reflink
> >>+
> >>+blocksize=$(( 128 * 1024 ))
> >>+nr=4096
> >>+file="$SCRATCH_MNT/tmp"
> >>+
> >>+_scratch_mkfs
> >>+_scratch_mount
> >>+
> >>+# write the initial block for later reflink
> >>+$XFS_IO_PROG -f -c "pwrite 0 $blocksize" -c "fsync" $file | _filter_xfs_io
> >>+
> >>+# use reflink to create the rest of the file, whose all extents are all
> >>+# pointing to the first extent
> >>+for i in $(seq 1 $nr); do
> >>+ $XFS_IO_PROG -c "reflink $file 0 $(( $i * $blocksize )) $blocksize" \
> >>+ $SCRATCH_MNT/tmp > /dev/null || _fail "reflink failed"
> >>+done
> >>+
> >>+# then call fiemap on that file, which shouldn't hang the fs by all means
> >>+$XFS_IO_PROG -c "fiemap" $file >> $seqres.full
> >>+
> >>+# success, all done
> >>+status=0
> >>+exit
> >>diff --git a/tests/btrfs/028.out b/tests/btrfs/028.out
> >>new file mode 100644
> >>index 0000000..2b5a9a5
> >>--- /dev/null
> >>+++ b/tests/btrfs/028.out
> >>@@ -0,0 +1,3 @@
> >>+QA output created by 028
> >>+wrote 131072/131072 bytes at offset 0
> >>+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> >>diff --git a/tests/btrfs/group b/tests/btrfs/group
> >>index da0e27f..8f6f877 100644
> >>--- a/tests/btrfs/group
> >>+++ b/tests/btrfs/group
> >>@@ -30,6 +30,7 @@
> >> 025 auto quick send clone
> >> 026 auto quick compress prealloc
> >> 027 auto replace
> >>+028 auto clone
> >> 029 auto quick clone
> >> 030 auto quick send
> >> 031 auto quick subvol clone
> >>--
> >>2.5.5
> >>
> >>
> >>
> >>--
> >>To unsubscribe from this list: send the line "unsubscribe fstests" in
> >>the body of a message to majordomo@xxxxxxxxxxxxxxx
> >>More majordomo info at http://vger.kernel.org/majordomo-info.html
> >
> >
> >
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe fstests" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
