linux-unionfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Eryu Guan <eguan@linux.alibaba.com>
To: Amir Goldstein <amir73il@gmail.com>
Cc: Eryu Guan <guaneryu@gmail.com>, Icenowy Zheng <icenowy@aosc.io>,
	Chengguang Xu <cgxu519@mykernel.net>,
	Xiao Yang <yangx.jy@cn.fujitsu.com>,
	Miklos Szeredi <miklos@szeredi.hu>,
	overlayfs <linux-unionfs@vger.kernel.org>,
	fstests <fstests@vger.kernel.org>
Subject: Re: [PATCH 4/4] overlay: Test lost immutable/append-only flags on copy-up
Date: Wed, 27 Jan 2021 10:57:52 +0800	[thread overview]
Message-ID: <20210127025752.GG58500@e18g06458.et15sqa> (raw)
In-Reply-To: <CAOQ4uxiXUN7LkzNLZto6iK2YuDdxp7PGoQMGCm89p1kNWUf=YA@mail.gmail.com>

On Mon, Jan 25, 2021 at 03:24:07PM +0200, Amir Goldstein wrote:
> On Sat, Jan 16, 2021 at 6:56 PM Amir Goldstein <amir73il@gmail.com> wrote:
> >
> > Chengguang Xu reported [1] that append-only flag is lost on copy-up.
> > I had noticed that for directories, immutable flag can also be lost on
> > copy up (when parent is copied up). That's an old overlayfs bug.
> >
> > Overlayfs added the ability to set inode flags (e.g. chattr +i) in
> > kernel 5.10 by commit 61536bed2149 ("ovl: support [S|G]ETFLAGS and
> > FS[S|G]ETXATTR ioctls for directories").
> > Icenowy Zheng reported [2] a regression in that commit that causes
> > a deadlock when setting inode flags on lower dir.
> >
> > There is a commented line in the test that triggers this deadlock,
> > but it has been left commented out until a fix is merged upstream.
> >
> 
> Re-iterate in correct thread:
> 
> The fix for above is in overlayfs-next:
> 
> * 147ec02b8705 - ovl: avoid deadlock on directory ioctl
> 
> But I wouldn't uncomment that line in the test just yet.

Then I'd prefer wait for the deadlock fix land in upstream first, and
merge the test with the deadlock trigger in place.

Or as you mentioned in previous thread, we could seperate the deadlock
case as a new test (also remove it from current overlay/075), so we
could merge the [s,g]etxflags case first, then the deadlock case only
when the fix is upstreamd.

Either way works for me, I just want to avoid merging the test without
the deadlock trigger, then uncomment it when the fix is available.

Thanks,
Eryu

> 
> To be clear, this test does not pass in master and there is no clear
> estimation on when the reported issues will be fixed.
> 
> Miklos,
> 
> Do you plan to followup on your VFS implementation for the
> {s,g}etxflags methods?
> 
> Thanks,
> Amir.
> 
> > [1] https://lore.kernel.org/linux-unionfs/20201226104618.239739-1-cgxu519@mykernel.net/
> > [2] https://lore.kernel.org/linux-unionfs/20210101201230.768653-1-icenowy@aosc.io/
> >
> > Signed-off-by: Amir Goldstein <amir73il@gmail.com>
> > ---
> >  tests/overlay/075     | 97 +++++++++++++++++++++++++++++++++++++++++++
> >  tests/overlay/075.out | 11 +++++
> >  tests/overlay/group   |  1 +
> >  3 files changed, 109 insertions(+)
> >  create mode 100755 tests/overlay/075
> >  create mode 100644 tests/overlay/075.out
> >
> > diff --git a/tests/overlay/075 b/tests/overlay/075
> > new file mode 100755
> > index 00000000..bcdc8d4e
> > --- /dev/null
> > +++ b/tests/overlay/075
> > @@ -0,0 +1,97 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (C) 2021 CTERA Networks. All Rights Reserved.
> > +#
> > +# FS QA Test No. 075
> > +#
> > +# Run the t_immutable test program for immutable/append-only files
> > +# and directories that exist in overlayfs lower layer.
> > +#
> > +# This test is similar and was derived from generic/079, but instead
> > +# of creating new files which are created in upper layer, prepare
> > +# the test area in lower layer before running the t_immutable test on
> > +# the overlayfs mount.
> > +#
> > +seq=`basename $0`
> > +seqres=$RESULT_DIR/$seq
> > +echo "QA output created by $seq"
> > +
> > +here=`pwd`
> > +timmutable=$here/src/t_immutable
> > +lowerdir=$OVL_BASE_SCRATCH_MNT/$OVL_LOWER
> > +upperdir=$OVL_BASE_SCRATCH_MNT/$OVL_UPPER
> > +tmp=/tmp/$$
> > +status=1       # failure is the default!
> > +trap "_cleanup; exit \$status" 0 1 2 3 15
> > +
> > +_cleanup()
> > +{
> > +       # -r will fail to remove test dirs, because we added subdirs
> > +       # we just need to remove the flags so use -R
> > +       $timmutable -R $upperdir/testdir &> /dev/null
> > +       $timmutable -R $lowerdir/testdir &> /dev/null
> > +       rm -f $tmp.*
> > +}
> > +
> > +# get standard environment, filters and checks
> > +. ./common/rc
> > +. ./common/filter
> > +
> > +_supported_fs overlay
> > +
> > +_require_chattr iaA
> > +_require_test_program "t_immutable"
> > +_require_scratch
> > +
> > +_scratch_mkfs
> > +
> > +# Preparing test area files in lower dir and check chattr support of base fs
> > +mkdir -p $lowerdir
> > +mkdir -p $upperdir
> > +$timmutable -C $lowerdir/testdir >$tmp.out 2>&1
> > +if grep -q -e 'Operation not supported' -e "Inappropriate ioctl" $tmp.out; then
> > +       _notrun "Setting immutable/append flag not supported"
> > +fi
> > +# Remove the immutable/append-only flags and create subdirs
> > +$timmutable -R $lowerdir/testdir >$tmp.out 2>&1
> > +for dir in $lowerdir/testdir/*.d; do
> > +       mkdir $dir/subdir
> > +done
> > +# Restore the immutable/append-only flags
> > +$timmutable -C $lowerdir/testdir >$tmp.out 2>&1
> > +
> > +_scratch_mount
> > +
> > +# Test immutability of files in overlay
> > +echo "Before directories copy up"
> > +$timmutable $SCRATCH_MNT/testdir 2>&1
> > +
> > +# Trigger copy-up of immutable/append-only dirs by touching their subdirs
> > +# inode flags are not copied-up, so immutable/append-only flags are lost
> > +for dir in $SCRATCH_MNT/testdir/*.d; do
> > +       # chattr on dir fails (not supported) on kernel < 5.10.
> > +       # chattr on lower dir will deadlock on kernel 5.10 with commit 61536bed2149
> > +       # ("ovl: support [S|G]ETFLAGS and FS[S|G]ETXATTR ioctls for directories"),
> > +       # so this line is commented out until a fix is merged
> > +       # $CHATTR_PROG +A $dir/subdir > /dev/null 2>&1
> > +       touch $dir/subdir
> > +done
> > +
> > +# Trigger copy-up of append-only files by touching them
> > +# inode flags are not copied-up, so append-only flags are lost
> > +# touch on the immutable files is expected to fail, so immutable
> > +# flags will not be lost
> > +for file in $SCRATCH_MNT/testdir/*.f; do
> > +       touch $file > /dev/null 2>&1
> > +done
> > +
> > +# immutable/append-only flags still exist on the overlay in-core inode
> > +# After mount cycle, flags are forever lost
> > +_scratch_cycle_mount
> > +
> > +# Test immutability of files in overlay after directories copy-up
> > +echo "After directories copy up"
> > +$timmutable $SCRATCH_MNT/testdir 2>&1
> > +
> > +status=$?
> > +exit
> > diff --git a/tests/overlay/075.out b/tests/overlay/075.out
> > new file mode 100644
> > index 00000000..ab39c6b8
> > --- /dev/null
> > +++ b/tests/overlay/075.out
> > @@ -0,0 +1,11 @@
> > +QA output created by 075
> > +Before directories copy up
> > +testing immutable...PASS.
> > +testing append-only...PASS.
> > +testing immutable as non-root...PASS.
> > +testing append-only as non-root...PASS.
> > +After directories copy up
> > +testing immutable...PASS.
> > +testing append-only...PASS.
> > +testing immutable as non-root...PASS.
> > +testing append-only as non-root...PASS.
> > diff --git a/tests/overlay/group b/tests/overlay/group
> > index 047ea046..cfc75bb1 100644
> > --- a/tests/overlay/group
> > +++ b/tests/overlay/group
> > @@ -77,6 +77,7 @@
> >  072 auto quick copyup hardlink
> >  073 auto quick whiteout
> >  074 auto quick exportfs dangerous
> > +075 auto quick perms
> >  100 auto quick union samefs
> >  101 auto quick union nonsamefs
> >  102 auto quick union nonsamefs xino
> > --
> > 2.25.1
> >

  parent reply	other threads:[~2021-01-27  4:56 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-16 16:56 [PATCH 0/4] Tests for overlayfs immutable/append-only files Amir Goldstein
2021-01-16 16:56 ` [PATCH 1/4] overlay/030: Update comment w.r.t upstream kernel Amir Goldstein
2021-01-16 16:56 ` [PATCH 2/4] src/t_immutable: factor out some helpers Amir Goldstein
2021-01-24 15:09   ` Eryu Guan
2021-01-24 15:29     ` Amir Goldstein
2021-01-25 12:35       ` Eryu Guan
2021-01-16 16:56 ` [PATCH 3/4] src/t_immutable: Allow setting flags on existing files Amir Goldstein
2021-01-24 15:14   ` Eryu Guan
2021-01-24 15:32     ` Amir Goldstein
2021-01-25 12:46       ` Eryu Guan
2021-01-25 13:17         ` Amir Goldstein
2021-01-16 16:56 ` [PATCH 4/4] overlay: Test lost immutable/append-only flags on copy-up Amir Goldstein
2021-01-25 13:24   ` Amir Goldstein
2021-01-26  8:47     ` Miklos Szeredi
2021-01-27  2:57     ` Eryu Guan [this message]
2021-01-28  6:45       ` Amir Goldstein

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210127025752.GG58500@e18g06458.et15sqa \
    --to=eguan@linux.alibaba.com \
    --cc=amir73il@gmail.com \
    --cc=cgxu519@mykernel.net \
    --cc=fstests@vger.kernel.org \
    --cc=guaneryu@gmail.com \
    --cc=icenowy@aosc.io \
    --cc=linux-unionfs@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=yangx.jy@cn.fujitsu.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).