Skip to contentSkip to conversation

generic: test reporting of wb errors via syncfs

RFC PATCH Superseded Jeff Layton 2 participants 7 messages Apr 14, 2020 – Apr 21, 2020 VFS

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 generic: test reporting of wb errors via syncfs

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

---
 tests/generic/999     | 98 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 ++++
 tests/generic/group   |  1 +
 3 files changed, 107 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out
tests/generic/999 new file +98 -0
new file mode 100755
index 000000000000..7383ce24c8fd
@@ -0,0 +1,98 @@
+#! /bin/bash
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2020, Jeff Layton <[email protected]>
+#
+# 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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
+_dmerror_init
+_dmerror_mount
+
+datalen=65536
+_require_fs_space $SCRATCH_MNT $datalen
+
+# use fd 5 to hold file open
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+exec 5>$testfile
+
+# write some data to file and fsync it out
+$XFS_IO_PROG -c "pwrite -W -q 0 $datalen" $testfile
+
+# flip device to non-working mode
+_dmerror_load_error_table
+
+# rewrite the data, and do fdatasync
+$XFS_IO_PROG -c "pwrite -w -q 0 $datalen" $testfile
+
+# heal the device error
+_dmerror_load_working_table
+
+# open again and call syncfs twice
+echo "One of the following syncfs calls should fail with EIO:"
+$XFS_IO_PROG -r -c syncfs -c syncfs $testfile
+echo "done"
+
+echo "This syncfs call should succeed:"
+$XFS_IO_PROG -r -c syncfs $testfile
+echo "done"
+
+# close file
+exec 5>&-
+
+# success, all done
+_dmerror_cleanup
+
+status=0
+exit
tests/generic/999.out new file +8 -0
new file mode 100644
index 000000000000..950a2ba42503
@@ -0,0 +1,8 @@
+QA output created by 999
+Format and mount
+fdatasync: Input/output error
+One of the following syncfs calls should fail with EIO:
+syncfs: Input/output error
+done
+This syncfs call should succeed:
+done
tests/generic/group +1 -0
index 99d06c9ad945..028cdbd0d52b 100644
@@ -598,3 +598,4 @@
 594 auto quick quota
 595 auto quick encrypt
 596 auto quick
+999 auto quick
Signature · 1 line
2.25.2
#1 lore raw
Quoted 6 lines · Jeff Layton

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

Quoted 9 lines
---
 tests/generic/999     | 98 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 ++++
 tests/generic/group   |  1 +
 3 files changed, 107 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out

I should note that I sent this as an RFC as it will fail until this series is merged:

Subject: [PATCH v4 RESEND 0/2] vfs: have syncfs() return error when there are writeback errors

...which I've just recently re-posted to linux-fsdevel, et. al.

Quoted 127 lines
tests/generic/999 new file +98 -0
new file mode 100755
index 000000000000..7383ce24c8fd
@@ -0,0 +1,98 @@
+#! /bin/bash
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2020, Jeff Layton <[email protected]>
+#
+# 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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
+_dmerror_init
+_dmerror_mount
+
+datalen=65536
+_require_fs_space $SCRATCH_MNT $datalen
+
+# use fd 5 to hold file open
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+exec 5>$testfile
+
+# write some data to file and fsync it out
+$XFS_IO_PROG -c "pwrite -W -q 0 $datalen" $testfile
+
+# flip device to non-working mode
+_dmerror_load_error_table
+
+# rewrite the data, and do fdatasync
+$XFS_IO_PROG -c "pwrite -w -q 0 $datalen" $testfile
+
+# heal the device error
+_dmerror_load_working_table
+
+# open again and call syncfs twice
+echo "One of the following syncfs calls should fail with EIO:"
+$XFS_IO_PROG -r -c syncfs -c syncfs $testfile
+echo "done"
+
+echo "This syncfs call should succeed:"
+$XFS_IO_PROG -r -c syncfs $testfile
+echo "done"
+
+# close file
+exec 5>&-
+
+# success, all done
+_dmerror_cleanup
+
+status=0
+exit
tests/generic/999.out new file +8 -0
new file mode 100644
index 000000000000..950a2ba42503
@@ -0,0 +1,8 @@
+QA output created by 999
+Format and mount
+fdatasync: Input/output error
+One of the following syncfs calls should fail with EIO:
+syncfs: Input/output error
+done
+This syncfs call should succeed:
+done
tests/generic/group +1 -0
index 99d06c9ad945..028cdbd0d52b 100644
@@ -598,3 +598,4 @@
 594 auto quick quota
 595 auto quick encrypt
 596 auto quick
+999 auto quick
Signature · 1 line
Jeff Layton <[email protected]>
#2 lore raw
In reply to Jeff Layton
Quoted 36 lines · Jeff Layton

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

---
 tests/generic/999     | 98 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 ++++
 tests/generic/group   |  1 +
 3 files changed, 107 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out
tests/generic/999 new file +15 -0
new file mode 100755
index 000000000000..7383ce24c8fd
@@ -0,0 +1,98 @@
+#! /bin/bash
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2020, Jeff Layton <[email protected]>
+#
+# 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.

tests/generic/999 line 25

Quoted 10 lines
tests/generic/999 new file +10 -0
+#
+# 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
+#-----------------------------------------------------------------------

I think the big copyright hunk has been replaced with the SPDX-License-Identifier thing (see other tests for reference).

Quoted 29 lines
tests/generic/999 new file +29 -0
+
+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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error

tests/generic/999 line 64

Quoted 10 lines
tests/generic/999 new file +10 -0
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
+_dmerror_init
+_dmerror_mount
+
+datalen=65536
+_require_fs_space $SCRATCH_MNT $datalen

That seems unnecessary for such a small value. BTW, any reason this needs to write more than a page?

tests/generic/999 line 69

Quoted 5 lines
tests/generic/999 new file +5 -0
+
+# use fd 5 to hold file open
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+exec 5>$testfile
+

Also what's the reason for holding an fd on the test file like this? Does this affect error reporting behavior in some way? Otherwise the rest looks reasonable to me.

Brian

Quoted 55 lines
tests/generic/999 new file +29 -0
+# write some data to file and fsync it out
+$XFS_IO_PROG -c "pwrite -W -q 0 $datalen" $testfile
+
+# flip device to non-working mode
+_dmerror_load_error_table
+
+# rewrite the data, and do fdatasync
+$XFS_IO_PROG -c "pwrite -w -q 0 $datalen" $testfile
+
+# heal the device error
+_dmerror_load_working_table
+
+# open again and call syncfs twice
+echo "One of the following syncfs calls should fail with EIO:"
+$XFS_IO_PROG -r -c syncfs -c syncfs $testfile
+echo "done"
+
+echo "This syncfs call should succeed:"
+$XFS_IO_PROG -r -c syncfs $testfile
+echo "done"
+
+# close file
+exec 5>&-
+
+# success, all done
+_dmerror_cleanup
+
+status=0
+exit
tests/generic/999.out new file +8 -0
new file mode 100644
index 000000000000..950a2ba42503
@@ -0,0 +1,8 @@
+QA output created by 999
+Format and mount
+fdatasync: Input/output error
+One of the following syncfs calls should fail with EIO:
+syncfs: Input/output error
+done
+This syncfs call should succeed:
+done
tests/generic/group +1 -0
index 99d06c9ad945..028cdbd0d52b 100644
@@ -598,3 +598,4 @@
 594 auto quick quota
 595 auto quick encrypt
 596 auto quick
+999 auto quick
Signature · 1 line
2.25.2
#3 lore raw
Quoted 42 lines · Brian Foster
Quoted 41 lines · Jeff Layton

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

---
 tests/generic/999     | 98 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 ++++
 tests/generic/group   |  1 +
 3 files changed, 107 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out
tests/generic/999 new file +20 -0
new file mode 100755
index 000000000000..7383ce24c8fd
@@ -0,0 +1,98 @@
+#! /bin/bash
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2020, Jeff Layton <[email protected]>
+#
+# 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.
Quoted 9 lines

tests/generic/999 line 25

Quoted 5 lines
tests/generic/999 new file +5 -0
+#
+# 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
+#-----------------------------------------------------------------------

I think the big copyright hunk has been replaced with the SPDX-License-Identifier thing (see other tests for reference).

Thanks. Will fix.

Quoted 34 lines
Quoted 34 lines
tests/generic/999 new file +34 -0
+
+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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
Quoted 9 lines

tests/generic/999 line 64

Quoted 5 lines
tests/generic/999 new file +5 -0
+_dmerror_init
+_dmerror_mount
+
+datalen=65536
+_require_fs_space $SCRATCH_MNT $datalen

That seems unnecessary for such a small value. BTW, any reason this needs to write more than a page?

No reason for that size. I think I just copied that from generic/487 since I started with that one as a template. I'll cut it down to a page.

Should I not bother calling _require_fs_space here since it's so small? I wasn't sure how strict that was...

Quoted 10 lines

tests/generic/999 line 69

Quoted 5 lines
tests/generic/999 new file +5 -0
+
+# use fd 5 to hold file open
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+exec 5>$testfile
+

Also what's the reason for holding an fd on the test file like this? Does this affect error reporting behavior in some way? Otherwise the rest looks reasonable to me.

Again, copied from 487. It's not necessary for this test. I'll switch that to just "touch testfile" at the start and get rid of the follow-on close.

Thanks for the review!

Signature · 1 line
Jeff Layton <[email protected]>
#4 lore raw
Quoted 99 lines · Jeff Layton
Quoted 42 lines · Brian Foster
Quoted 41 lines · Jeff Layton

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

---
 tests/generic/999     | 98 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 ++++
 tests/generic/group   |  1 +
 3 files changed, 107 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out
tests/generic/999 new file +20 -0
new file mode 100755
index 000000000000..7383ce24c8fd
@@ -0,0 +1,98 @@
+#! /bin/bash
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2020, Jeff Layton <[email protected]>
+#
+# 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.
Quoted 9 lines

tests/generic/999 line 25

Quoted 5 lines
tests/generic/999 new file +5 -0
+#
+# 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
+#-----------------------------------------------------------------------

I think the big copyright hunk has been replaced with the SPDX-License-Identifier thing (see other tests for reference).

Thanks. Will fix.

Quoted 34 lines
Quoted 34 lines
tests/generic/999 new file +34 -0
+
+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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
Quoted 9 lines

tests/generic/999 line 64

Quoted 5 lines
tests/generic/999 new file +5 -0
+_dmerror_init
+_dmerror_mount
+
+datalen=65536
+_require_fs_space $SCRATCH_MNT $datalen

That seems unnecessary for such a small value. BTW, any reason this needs to write more than a page?

Quoted 6 lines

No reason for that size. I think I just copied that from generic/487 since I started with that one as a template. I'll cut it down to a page.

Should I not bother calling _require_fs_space here since it's so small? I wasn't sure how strict that was...

I'm not sure there's a rule for when to use it or not (it's probably just up to Eryu's preference), but yeah, it's probably overkill for a test that's only writing a handful of pages.

Brian

Quoted 19 lines
Quoted 10 lines

tests/generic/999 line 69

Quoted 5 lines
tests/generic/999 new file +5 -0
+
+# use fd 5 to hold file open
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+exec 5>$testfile
+

Also what's the reason for holding an fd on the test file like this? Does this affect error reporting behavior in some way? Otherwise the rest looks reasonable to me.

Again, copied from 487. It's not necessary for this test. I'll switch that to just "touch testfile" at the start and get rid of the follow-on close.

Thanks for the review!

Signature · 1 line
Jeff Layton <[email protected]>
#5 lore raw

RFC PATCH v2 1/1 generic: test reporting of wb errors via syncfs

In reply to Jeff Layton

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

---
 tests/generic/999     | 79 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 +++++
 tests/generic/group   |  1 +
 3 files changed, 88 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out

v2:
- update license comment
- only write a page of data
- don't bother testing for enough scratch space
- don't hold file open over test

Thanks to Brian Foster for the review! This is testing a proposed behavior change and is dependent on this patchset being merged:

vfs: have syncfs() return error when there are writeback errors

We'll probably want to wait until its fate is clear before merging this.

tests/generic/999 new file +79 -0
new file mode 100755
index 000000000000..cdc0772d0774
@@ -0,0 +1,79 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright (c) 2020, Jeff Layton. All rights reserved.
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+
+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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
+_dmerror_init
+_dmerror_mount
+
+
+# create file
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+touch $testfile
+
+# write a page of data to file, and call fsync
+datalen=$(getconf PAGE_SIZE)
+$XFS_IO_PROG -c "pwrite -W -q 0 $datalen" $testfile
+
+# flip device to non-working mode
+_dmerror_load_error_table
+
+# rewrite the data and call fdatasync
+$XFS_IO_PROG -c "pwrite -w -q 0 $datalen" $testfile
+
+# heal the device error
+_dmerror_load_working_table
+
+# open again and call syncfs twice
+echo "One of the following syncfs calls should fail with EIO:"
+$XFS_IO_PROG -r -c syncfs -c syncfs $testfile
+echo "done"
+
+echo "This syncfs call should succeed:"
+$XFS_IO_PROG -r -c syncfs $testfile
+echo "done"
+
+# success, all done
+_dmerror_cleanup
+
+status=0
+exit
tests/generic/999.out new file +8 -0
new file mode 100644
index 000000000000..950a2ba42503
@@ -0,0 +1,8 @@
+QA output created by 999
+Format and mount
+fdatasync: Input/output error
+One of the following syncfs calls should fail with EIO:
+syncfs: Input/output error
+done
+This syncfs call should succeed:
+done
tests/generic/group +1 -0
index 718575baeef9..9bcf296fc3dd 100644
@@ -598,3 +598,4 @@
 594 auto quick quota
 595 auto quick encrypt
 596 auto quick
+999 auto quick
Signature · 1 line
2.25.3
Reviewed-by on v2 1/1
#6 lore raw
Quoted 8 lines · Jeff Layton

From: Jeff Layton <[email protected]>

Add a test for new syncfs error reporting behavior. When an inode fails to be written back, ensure that a subsequent call to syncfs() will also report an error.

---

Quoted 131 lines
 tests/generic/999     | 79 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/999.out |  8 +++++
 tests/generic/group   |  1 +
 3 files changed, 88 insertions(+)
 create mode 100755 tests/generic/999
 create mode 100644 tests/generic/999.out

v2:
- update license comment
- only write a page of data
- don't bother testing for enough scratch space
- don't hold file open over test

Thanks to Brian Foster for the review! This is testing a proposed behavior change and is dependent on this patchset being merged:

vfs: have syncfs() return error when there are writeback errors

We'll probably want to wait until its fate is clear before merging this.

tests/generic/999 new file +79 -0
new file mode 100755
index 000000000000..cdc0772d0774
@@ -0,0 +1,79 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright (c) 2020, Jeff Layton. All rights reserved.
+# FS QA Test No. 999
+#
+# Open a file and write to it and fsync. Then, flip the data device to throw
+# errors, write to it again and do an fdatasync. Then open an O_RDONLY fd on
+# the same file and call syncfs against it and ensure that an error is reported.
+# Then call syncfs again and ensure that no error is reported. Finally, repeat
+# the open and syncfs and ensure that there is no error reported.
+
+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.*
+	_dmerror_cleanup
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/dmerror
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_nocheck
+# This test uses "dm" without taking into account the data could be on
+# realtime subvolume, thus the test will fail with rtinherit=1
+_require_no_rtinherit
+_require_dm_target error
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
+_dmerror_init
+_dmerror_mount
+
+
+# create file
+testfile=$SCRATCH_MNT/syncfs-reports-errors
+touch $testfile
+
+# write a page of data to file, and call fsync
+datalen=$(getconf PAGE_SIZE)
+$XFS_IO_PROG -c "pwrite -W -q 0 $datalen" $testfile
+
+# flip device to non-working mode
+_dmerror_load_error_table
+
+# rewrite the data and call fdatasync
+$XFS_IO_PROG -c "pwrite -w -q 0 $datalen" $testfile
+
+# heal the device error
+_dmerror_load_working_table
+
+# open again and call syncfs twice
+echo "One of the following syncfs calls should fail with EIO:"
+$XFS_IO_PROG -r -c syncfs -c syncfs $testfile
+echo "done"
+
+echo "This syncfs call should succeed:"
+$XFS_IO_PROG -r -c syncfs $testfile
+echo "done"
+
+# success, all done
+_dmerror_cleanup
+
+status=0
+exit
tests/generic/999.out new file +8 -0
new file mode 100644
index 000000000000..950a2ba42503
@@ -0,0 +1,8 @@
+QA output created by 999
+Format and mount
+fdatasync: Input/output error
+One of the following syncfs calls should fail with EIO:
+syncfs: Input/output error
+done
+This syncfs call should succeed:
+done
tests/generic/group +1 -0
index 718575baeef9..9bcf296fc3dd 100644
@@ -598,3 +598,4 @@
 594 auto quick quota
 595 auto quick encrypt
 596 auto quick
+999 auto quick
Signature · 1 line
2.25.3