On Wed, May 11, 2016 at 10:01:18PM -0700, Darrick J. Wong wrote:
> On Thu, May 12, 2016 at 12:51:35PM +0800, Qu Wenruo wrote:
> > For fully dedupe file, which means all its file exntents are pointing to
> > the same bytenr, btrfs can cause soft lockup when calling fiemap ioctl
> > on that file, like the following output:
> > ------
> > CPU: 1 PID: 7500 Comm: xfs_io Not tainted 4.5.0-rc6+ #2
> > Hardware name: XXXXXXXXXXXXXXXXXXXXXXXXXXX
> > task: ffff880027681b40 ti: ffff8800276e0000 task.ti: ffff8800276e0000
> > RIP: 0010:[<ffffffffa02583e4>] [<ffffffffa02583e4>]
> > __merge_refs+0x34/0x120 [btrfs]
> > RSP: 0018:ffff8800276e3c08 EFLAGS: 00000202
> > RAX: ffff8800269cc330 RBX: ffff8800269cdb18 RCX: 0000000000000007
> > RDX: 00000000000061b0 RSI: ffff8800269cc4c8 RDI: ffff8800276e3c88
> > RBP: ffff8800276e3c20 R08: 0000000000000000 R09: 0000000000000001
> > R10: 0000000000000000 R11: 0000000000000000 R12: ffff880026ea3cb0
> > R13: ffff8800276e3c88 R14: ffff880027132a50 R15: ffff880027430000
> > FS: 00007f10201df700(0000) GS:ffff88003fa00000(0000)
> > knlGS:0000000000000000
> > CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > CR2: 00007f10201ec000 CR3: 0000000027603000 CR4: 00000000000406e0
> > Stack:
> > 0000000000000000 0000000000000000 0000000000000000 ffff8800276e3ce8
> > ffffffffa0259f38 0000000000000005 ffff8800274c6870 ffff8800274c7d88
> > 0000000000c10000 0000000000000000 0000000000000001 0000000027431190
> > Call Trace:
> > [<ffffffffa0259f38>] find_parent_nodes+0x448/0x740 [btrfs]
> > [<ffffffffa025a4f2>] btrfs_check_shared+0x102/0x1b0 [btrfs]
> > [<ffffffff811fdcad>] ? __might_fault+0x4d/0xa0
> > [<ffffffffa021899c>] extent_fiemap+0x2ac/0x550 [btrfs]
> > [<ffffffff811ce156>] ? __filemap_fdatawait_range+0x96/0x160
> > [<ffffffffa01f8ee0>] ? btrfs_get_extent+0xb30/0xb30 [btrfs]
> > [<ffffffffa01f5da5>] btrfs_fiemap+0x45/0x50 [btrfs]
> > [<ffffffff81246bb8>] do_vfs_ioctl+0x498/0x670
> > [<ffffffff81246e09>] SyS_ioctl+0x79/0x90
> > [<ffffffff8184e997>] entry_SYSCALL_64_fastpath+0x12/0x6f
> > Code: 41 55 41 54 53 4c 8b 27 4c 39 e7 0f 84 e9 00 00 00 49 89 fd 49 8b
> > 34 24 49 39 f5 48 8b 1e 75 17 e9 d5 00 00 00 49 39 dd 48 8b 03 <48> 89
> > de 0f 84 b9 00 00 00 48 89 c3 8b 46 2c 41 39 44 24 2c 75
> > ------
> >
> > Also btrfs will return wrong flag for all these extents, they should
> > have SHARED(0x2000) flag, while btrfs still consider them as exclusive
> > extents.
> >
> > On the other hand, with unmerged xfs reflink patches, xfs can handle it
> > without problem, and for patched btrfs, it can also handle it well.
> >
> > This test case will create a large fully deduped file to check if the fs
> > can handle the fiemap ioctl and return correct SHARED flag for any fs
> > which support reflink.
> >
> > Reported-by: Tsutomu Itoh <t-itoh@xxxxxxxxxxxxxx>
> > Signed-off-by: Qu Wenruo <quwenruo@xxxxxxxxxxxxxx>
> > ---
> > common/punch | 17 +++++++++++
> > tests/generic/352 | 82 +++++++++++++++++++++++++++++++++++++++++++++++++++
> > tests/generic/352.out | 5 ++++
> > tests/generic/group | 1 +
> > 4 files changed, 105 insertions(+)
> > create mode 100755 tests/generic/352
> > create mode 100644 tests/generic/352.out
> >
> > diff --git a/common/punch b/common/punch
> > index 43f04c2..44c6e1c 100644
> > --- a/common/punch
> > +++ b/common/punch
> > @@ -218,6 +218,23 @@ _filter_fiemap()
> > _coalesce_extents
> > }
> >
> > +_filter_fiemap_flags()
> > +{
> > + $AWK_PROG '
> > + $3 ~ /hole/ {
> > + print $1, $2, $3;
> > + next;
> > + }
> > + $5 ~ /0x[[:xdigit:]]*8[[:xdigit:]][[:xdigit:]]/ {
> > + print $1, $2, "unwritten";
> > + next;
> > + }
> > + $5 ~ /0x[[:xdigit:]]+/ {
> > + print $1, $2, $5;
> > + }' |
> > + _coalesce_extents
> > +}
> > +
> > # Filters fiemap output to only print the
> > # file offset column and whether or not
> > # it is an extent or a hole
> > diff --git a/tests/generic/352 b/tests/generic/352
> > new file mode 100755
> > index 0000000..83f24ec
> > --- /dev/null
> > +++ b/tests/generic/352
> > @@ -0,0 +1,82 @@
> > +#! /bin/bash
> > +# FS QA Test 352
> > +#
> > +# Test fiemap ioctl on heavily deduped file
> > +#
> > +# This test case will check if extent backref search goes
> > +# without problem and return correct SHARED flag.
> > +# Which btrfs will soft lock up and return wrong shared flag.
> > +#
> > +#-----------------------------------------------------------------------
> > +# 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
> > +. ./common/punch
> > +
> > +# remove previous $seqres.full before test
> > +rm -f $seqres.full
> > +
> > +# real QA test starts here
> > +
> > +# Modify as appropriate.
> > +_supported_fs generic
> > +_supported_os Linux
> > +_require_scratch_reflink
> > +
> > +_scratch_mkfs > /dev/null 2>&1
> > +_scratch_mount
> > +
> > +nr=$((8192 * $LOAD_FACTOR))
> > +blocksize=$((128 * 1024))
> > +file="$SCRATCH_MNT/tmp"
> > +
> > +# write the initial block for later reflink
> > +$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" \
> > + $file > /dev/null
>
> FYI, there's a _reflink_range helper for this now.
>
> (There's also a _pwrite_byte helper.)
>
> > +done
> > +
> > +# then call fiemap on that file to test both the shared flag and if
> > +# reserved extent mapping search will cause soft lockup
> > +$XFS_IO_PROG -c "fiemap -v" $file | _filter_fiemap_flags
>
> I think this test needs _require_fiemap or else it'll fail on things
> that don't support fiemap, like NFS.
>
> PS: How well does btrfs handle generic/17[56]?
Well, running filefrag -v while it's running will indeed cause soft lockup
problems. Ulp.
<reboots vm, tries again>
So left alone, 175 finishes in ~90s. But then it runs btrfsck.
I also see that btrfsck seems to be doing this over and over:
pread64(4, "...", 16384, 347668480) = 16384 <0.000047>
pread64(4, "...", 16384, 151470080) = 16384 <0.000081>
pread64(4, "...", 16384, 337936384) = 16384 <0.000054>
readahead(4, 347783168, 16384) = 0 <0.000064>
readahead(4, 347815936, 16384) = 0 <0.000058>
readahead(4, 347832320, 16384) = 0 <0.000045>
readahead(4, 347848704, 16384) = 0 <0.000045>
readahead(4, 347881472, 16384) = 0 <0.000044>
readahead(4, 347897856, 16384) = 0 <0.000043>
readahead(4, 347914240, 16384) = 0 <0.000044>
readahead(4, 347930624, 16384) = 0 <0.000047>
readahead(4, 347947008, 16384) = 0 <0.000044>
readahead(4, 347963392, 16384) = 0 <0.000044>
readahead(4, 347979776, 16384) = 0 <0.000043>
readahead(4, 348012544, 16384) = 0 <0.000047>
readahead(4, 348028928, 16384) = 0 <0.000045>
readahead(4, 348045312, 16384) = 0 <0.000048>
readahead(4, 348078080, 16384) = 0 <0.000048>
readahead(4, 348110848, 16384) = 0 <0.000045>
readahead(4, 348127232, 16384) = 0 <0.000044>
(It's been running 20min now, I'm going to kill it and go to bed.)
--D
>
> --D
>
> > +
> > +# success, all done
> > +status=0
> > +exit
> > diff --git a/tests/generic/352.out b/tests/generic/352.out
> > new file mode 100644
> > index 0000000..a87c507
> > --- /dev/null
> > +++ b/tests/generic/352.out
> > @@ -0,0 +1,5 @@
> > +QA output created by 352
> > +wrote 131072/131072 bytes at offset 0
> > +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> > +0: [0..2097151]: 0x2000
> > +1: [2097152..2097407]: 0x2001
> > diff --git a/tests/generic/group b/tests/generic/group
> > index 36fb759..3f00386 100644
> > --- a/tests/generic/group
> > +++ b/tests/generic/group
> > @@ -354,3 +354,4 @@
> > 349 blockdev quick rw
> > 350 blockdev quick rw
> > 351 blockdev quick rw
> > +352 auto clone
> > --
> > 2.8.2
> >
> >
> >
> --
> 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 fstests" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
