Re: [PATCH] xfs: Check if rt summary/bitmap buffers are logged with correct xfs_buf type

[Chandan Babu R <chandanrlinux@gmail.com>]

On Monday 14 September 2020 8:43:46 PM IST Darrick J. Wong wrote:
> On Mon, Sep 14, 2020 at 06:34:56PM +0800, Zorro Lang wrote:
> > On Mon, Sep 14, 2020 at 02:30:53PM +0530, Chandan Babu R wrote:
> > > This commit adds a test to check if growing a real-time device can end
> > > up logging an xfs_buf with the "type" subfield of
> > > bip->bli_formats->blf_flags set to XFS_BLFT_UNKNOWN_BUF. When this
> > > occurs the following call trace is printed on the console,
> > > 
> > > XFS: Assertion failed: (bip->bli_flags & XFS_BLI_STALE) || (xfs_blft_from_flags(&bip->__bli_format) > XFS_BLFT_UNKNOWN_BUF && xfs_blft_from_flags(&bip->__bli_format) < XFS_BLFT_MAX_BUF), file: fs/xfs/xfs_buf_item.c, line: 331
> > > Call Trace:
> > >  xfs_buf_item_format+0x632/0x680
> > >  ? kmem_alloc_large+0x29/0x90
> > >  ? kmem_alloc+0x70/0x120
> > >  ? xfs_log_commit_cil+0x132/0x940
> > >  xfs_log_commit_cil+0x26f/0x940
> > >  ? xfs_buf_item_init+0x1ad/0x240
> > >  ? xfs_growfs_rt_alloc+0x1fc/0x280
> > >  __xfs_trans_commit+0xac/0x370
> > >  xfs_growfs_rt_alloc+0x1fc/0x280
> > >  xfs_growfs_rt+0x1a0/0x5e0
> > >  xfs_file_ioctl+0x3fd/0xc70
> > >  ? selinux_file_ioctl+0x174/0x220
> > >  ksys_ioctl+0x87/0xc0
> > >  __x64_sys_ioctl+0x16/0x20
> > >  do_syscall_64+0x3e/0x70
> > >  entry_SYSCALL_64_after_hwframe+0x44/0xa9
> > > 
> > > The kernel patch "xfs: Set xfs_buf type flag when growing summary/bitmap
> > > files" is required to fix this issue.
> > > 
> > > Signed-off-by: Chandan Babu R <chandanrlinux@gmail.com>
> > > ---
> > >  tests/xfs/260     | 52 +++++++++++++++++++++++++++++++++++++++++++++++
> > >  tests/xfs/260.out |  2 ++
> > >  tests/xfs/group   |  1 +
> > >  3 files changed, 55 insertions(+)
> > >  create mode 100755 tests/xfs/260
> > >  create mode 100644 tests/xfs/260.out
> > > 
> > > diff --git a/tests/xfs/260 b/tests/xfs/260
> > > new file mode 100755
> > > index 00000000..5fc1a5fc
> > > --- /dev/null
> > > +++ b/tests/xfs/260
> > > @@ -0,0 +1,52 @@
> > > +#! /bin/bash
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +# Copyright (c) 2020 Chandan Babu R.  All Rights Reserved.
> > > +#
> > > +# FS QA Test 260
> > > +#
> > > +# Test to check if growing a real-time device can end up logging an
> > > +# xfs_buf with the "type" subfield of bip->bli_formats->blf_flags set
> > > +# to XFS_BLFT_UNKNOWN_BUF.
> 
> Please state explicitly that this is a regression test for "xfs: Set
> xfs_buf type flag when growing summary/bitmap files".
>

Sure. I will add a description to the comment header.

> > > +#
> > > +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
> > > +
> > > +# remove previous $seqres.full before test
> > > +rm -f $seqres.full
> > > +
> > > +# real QA test starts here
> > > +
> > > +# Modify as appropriate.
> >      ^^^^
> > I think this comment line is useless

Sorry, I missed out on removing that.

> > 
> > > +_supported_fs generic
> > > +_supported_os Linux
> > > +_require_realtime
> > > +
> > > +MKFS_OPTIONS="-f -m reflink=0,rmapbt=0 -r rtdev=${SCRATCH_RTDEV},size=10M" \
> > > + 	    _mkfs_dev $SCRATCH_DEV >> $seqres.full
> > 
> > Hmm... if you need a sized rtdev, the _scratch_mkfs really can't help that
> > for now. You have to use _mkfs_dev(as you did) or make the helper to support
> > rtdev size:)
> 
> Does "_scratch_mkfs -r size=10M" not work here?

It works! Thanks for the suggestion.

> 
> > I don't know why you need "reflink=0,rmapbt=0", but not old xfsprogs doesn't
> > supports this two options, so you might need _scratch_mkfs_xfs_supported()
> > to check that. If they're not supported, they won't be enabled either. And
> > better to add comment to explain why make sure reflink and rmapbt are disabled.
> 
> That's a bug in mkfs.xfs, which should be fixed by "mkfs: fix
> reflink/rmap logic w.r.t.  realtime devices and crc=0 support".

I have replaced this part with just
_scratch_mkfs -r size=10M >> $seqres.full.

IMHO, the tester has to specify reflink=0 explicitly when testing realtime
configurations when using xfsprogs which support reflink and don't have
"mkfs: fix reflink/rmap logic w.r.t.  realtime devices and crc=0 support"
patch applied . Otherwise, the call to _scratch_mkfs() inside "check" script
will fail.

> 
> > > +_scratch_mount -o rtdev=$SCRATCH_RTDEV
> > 
> > As I known, xfstests deal with SCRATCH_RTDEV things in common/rc _scratch_options()
> > properly, _require_realtime with _scratch_mount are enough, don't need the
> > "-o rtdev=$SCRATCH_RTDEV".

You are right. I will remove it before posting the next version.

> > 
> > > +
> > > +$XFS_GROWFS_PROG $SCRATCH_MNT >> $seqres.full
> > > +
> > > +_scratch_unmount
> > > +
> > > +echo "Silence is golden"
> > > +
> > > +# success, all done
> > > +status=0
> > > +exit
> > > diff --git a/tests/xfs/260.out b/tests/xfs/260.out
> > > new file mode 100644
> > > index 00000000..18ca517c
> > > --- /dev/null
> > > +++ b/tests/xfs/260.out
> > > @@ -0,0 +1,2 @@
> > > +QA output created by 260
> > > +Silence is golden
> > > diff --git a/tests/xfs/group b/tests/xfs/group
> > > index ed0d389e..6f30a2e7 100644
> > > --- a/tests/xfs/group
> > > +++ b/tests/xfs/group
> > > @@ -257,6 +257,7 @@
> > >  257 auto quick clone
> > >  258 auto quick clone
> > >  259 auto quick
> > > +260 auto
> > 
> > Better to add 'growfs' group, and if the case is quick enough, 'quick' is acceptable:)
> 
> And maybe the 'realtime' group.
>

Sure, will do.

Darrick and Zorro, Thanks for your review comments.

> 
> > >  261 auto quick quota
> > >  262 dangerous_fuzzers dangerous_scrub dangerous_online_repair
> > >  263 auto quick quota
> > 
> 


-- 
chandan