public inbox for fstests@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/3] Check pwrite parameters
@ 2017-11-29 16:33 Goldwyn Rodrigues
  2017-11-29 16:33 ` [PATCH 2/3] generic/470: Test RWF_NOWAIT Goldwyn Rodrigues
                   ` (2 more replies)
  0 siblings, 3 replies; 8+ messages in thread
From: Goldwyn Rodrigues @ 2017-11-29 16:33 UTC (permalink / raw)
  To: darrick.wong, eguan; +Cc: fstests, Goldwyn Rodrigues

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

There are some parameters added with xfs_io. Check if the pwrite
parameters are available. For some cases, xfs_io now returns "command
-%c not supported", so added "not supported" to count as error.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>

Change since v2:
 - More comments
 - opts to become local variable

---
 common/rc | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/common/rc b/common/rc
index 4c053a53..90408b7e 100644
--- a/common/rc
+++ b/common/rc
@@ -2035,6 +2035,7 @@ _require_xfs_io_command()
 	shift
 	local param="$*"
 	local param_checked=0
+	local opts=""
 
 	testfile=$TEST_DIR/$$.xfs_io
 	case $command in
@@ -2079,6 +2080,15 @@ _require_xfs_io_command()
 		echo $testio | grep -q "invalid option" && \
 			_notrun "xfs_io $command support is missing"
 		;;
+	"pwrite")
+		# -N (RWF_NOWAIT) only works with direct I/O writes
+		if [ "$param" == "-N" ]; then
+			opts+=" -d"
+		fi
+		testio=`$XFS_IO_PROG -f $opts -c "pwrite $param 0 1M" \
+			$testfile 2>&1`
+		param_checked=1
+		;;
 	"scrub"|"repair")
 		testio=`$XFS_IO_PROG -x -c "$command probe 0" $TEST_DIR 2>&1`
 		echo $testio | grep -q "Inappropriate ioctl" && \
@@ -2109,7 +2119,9 @@ _require_xfs_io_command()
 		$XFS_IO_PROG -c "help $command" | grep -q "^ $param --" || \
 			_notrun "xfs_io $command doesn't support $param"
 	else
-		echo $testio | grep -q "invalid option" && \
+		# xfs_io could result in "command %c not supported" if it was
+		# built on kernels not supporting pwritev2() calls
+		echo $testio | grep -q "\(invalid option\|not supported\)" && \
 			_notrun "xfs_io $command doesn't support $param"
 	fi
 }
-- 
2.14.2


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 2/3] generic/470: Test RWF_NOWAIT
  2017-11-29 16:33 [PATCH 1/3] Check pwrite parameters Goldwyn Rodrigues
@ 2017-11-29 16:33 ` Goldwyn Rodrigues
  2017-12-06 10:05   ` Eryu Guan
  2017-11-29 16:33 ` [PATCH 3/3] generic/471: Partial direct write test Goldwyn Rodrigues
  2017-12-06  9:55 ` [PATCH 1/3] Check pwrite parameters Eryu Guan
  2 siblings, 1 reply; 8+ messages in thread
From: Goldwyn Rodrigues @ 2017-11-29 16:33 UTC (permalink / raw)
  To: darrick.wong, eguan; +Cc: fstests, Goldwyn Rodrigues

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

Tests the RWF_NOWAIT flag so the I/O returns immediately with -EAGAIN
on a new file since it requires block allocation.

It creates a file, syncs it, and overwrites the file with RWF_NOWAIT.
This should succeed.

Finally, read the contents to make sure the overwrite is successful.

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>

Changes since v1:
 - Fix testdir name
 - use of $XFS_IO_PROG instead of xfs_io
 - check pwrite accepts -N

Changes since v2:
 - corrected test description and improved documentation
 - new leaner look ;)
 - rw group
---
 tests/generic/470     | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++
 tests/generic/470.out | 13 +++++++++
 tests/generic/group   |  1 +
 3 files changed, 88 insertions(+)
 create mode 100755 tests/generic/470
 create mode 100644 tests/generic/470.out

diff --git a/tests/generic/470 b/tests/generic/470
new file mode 100755
index 00000000..5580718b
--- /dev/null
+++ b/tests/generic/470
@@ -0,0 +1,74 @@
+#! /bin/bash
+# FS QA Test No. 470
+#
+# write a file with RWF_NOWAIT and it would fail because there are no
+# blocks allocated. Create a file with direct I/O and re-write it
+# using RWF_NOWAIT. I/O should finish within 50 microsecods since
+# block allocations are already performed.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2017, SUSE Linux Products.  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!
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/populate
+. ./common/filter
+. ./common/attr
+
+# real QA test starts here
+_supported_os Linux
+_require_odirect
+_require_test
+_require_xfs_io_command pwrite -N
+
+# Remove reminiscence of previously run tests
+testdir=$TEST_DIR/$seq
+if [ -e $testdir ]; then
+	rm -Rf $testdir
+fi
+
+mkdir $testdir
+
+# Create a file with pwrite nowait (will fail with EAGAIN)
+$XFS_IO_PROG -f -d -c "pwrite -N -V 1 -b 1M 0 1M" $testdir/f1
+
+# Write the file without nowait
+$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -W -w -V 1 -b 1M 0 8M" $testdir/f1 | _filter_xfs_io
+
+time_taken=`$XFS_IO_PROG -d -c "pwrite -S 0xbb -N -V 1 -b 1M 2M 1M" $testdir/f1 | awk '/^1/ {print $5}'`
+
+# RWF_NOWAIT should finish within a short period of time. Anything longer
+# means it is waiting for something in the kernel which would be a fail.
+if (( $(echo "$time_taken < 0.05" | bc -l) )); then
+	echo "RWF_NOWAIT time is within limits."
+else
+	echo "RWF_NOWAIT took $time_taken seconds"
+fi
+
+$XFS_IO_PROG -c "pread -v 0 8M" $testdir/f1 | _filter_xfs_io_unique
+
+# success, all done
+status=0
+exit
diff --git a/tests/generic/470.out b/tests/generic/470.out
new file mode 100644
index 00000000..e46622cb
--- /dev/null
+++ b/tests/generic/470.out
@@ -0,0 +1,13 @@
+QA output created by 470
+pwrite: Resource temporarily unavailable
+wrote 8388608/8388608 bytes at offset 0
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+RWF_NOWAIT time is within limits.
+00000000:  aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa  ................
+*
+00200000:  bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb  ................
+*
+00300000:  aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa  ................
+*
+read 8388608/8388608 bytes at offset 0
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
diff --git a/tests/generic/group b/tests/generic/group
index 6c3bb03a..cf8529ea 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -472,3 +472,4 @@
 467 auto quick exportfs
 468 shutdown auto quick metadata
 469 auto quick
+470 auto quick rw
-- 
2.14.2


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 3/3] generic/471: Partial direct write test
  2017-11-29 16:33 [PATCH 1/3] Check pwrite parameters Goldwyn Rodrigues
  2017-11-29 16:33 ` [PATCH 2/3] generic/470: Test RWF_NOWAIT Goldwyn Rodrigues
@ 2017-11-29 16:33 ` Goldwyn Rodrigues
  2017-12-06  9:55 ` [PATCH 1/3] Check pwrite parameters Eryu Guan
  2 siblings, 0 replies; 8+ messages in thread
From: Goldwyn Rodrigues @ 2017-11-29 16:33 UTC (permalink / raw)
  To: darrick.wong, eguan; +Cc: fstests, Goldwyn Rodrigues

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

Getting an error midway through a direct write would return an error
and the error-code is returned in the write() call. However, partial
data is over-written during the call.

This tests simulates the ENOSPC error to check for partial direct
write consistency.

Changes since v1:
 - Check for option pwrite -O
 - Create a 200M file instead of multiple 40M files.

Changes since v2:
 - new leaner look ;)
 - use _fill_fs after reserving some space to create an almost full FS
 - _require_scratch for reseting system

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
---
 tests/generic/471     | 95 +++++++++++++++++++++++++++++++++++++++++++++++++++
 tests/generic/471.out |  9 +++++
 tests/generic/group   |  1 +
 3 files changed, 105 insertions(+)
 create mode 100755 tests/generic/471
 create mode 100644 tests/generic/471.out

diff --git a/tests/generic/471 b/tests/generic/471
new file mode 100755
index 00000000..51723548
--- /dev/null
+++ b/tests/generic/471
@@ -0,0 +1,95 @@
+#! /bin/bash
+# FS QA Test No. 471
+#
+# write() to a file opened with O_DIRECT with count > remaining
+# bytes. In case of a bug, the write returns ENOSPC wheras the
+# showing no data is written, but the file contents are updated.
+# Result should be the write should return remaining bytes
+# (to aligned bytes) instead of ENOSPC error.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2017, SUSE Linux Products.  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 -rf $tmp.* $testdir1
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/populate
+. ./common/filter
+. ./common/attr
+
+# real QA test starts here
+_supported_os Linux
+_require_odirect
+_require_xfs_io_command pwrite -O
+_require_scratch
+
+rm -f $seqres.full
+
+echo "Reformat with 320M size"
+sz_bytes=$((320 * 1024 * 1024))
+_scratch_mkfs_sized $sz_bytes >> $seqres.full 2>&1
+_scratch_mount >> $seqres.full 2>&1
+testdir=$SCRATCH_MNT/$seq
+mkdir $testdir
+
+# Reserve some space to remove later
+$XFS_IO_PROG -f -c "pwrite 0 2M" $testdir/reserve > /dev/null
+
+mkdir $testdir/fillspace
+echo "Fill the filesystem"
+_fill_fs $(( 1024 * 1024 * 1024 )) $testdir/fillspace 4096 0 > /dev/null 2>&1
+
+# Remove the reserve file to work on almost full filesystem
+rm $testdir/reserve
+sync
+
+# Create a file using buffered I/O which succeeds only partially
+$XFS_IO_PROG -f -c "pwrite -w -S 0xaa 0 4M" $testdir/partial
+
+# re-write using direct I/O with another pattern using one single buffer, once
+write_size=`$XFS_IO_PROG -d -c "pwrite -O -S 0xbb -V 1 -b 4M 0 4M" $testdir/partial | awk '/^wrote/ {split($2, bytes, "/"); print bytes[1]}'`
+if [ -z $write_size ]; then
+	write_size=0
+fi
+
+# If bug is not fixed, pwrite will return error and but still write data
+if [ $write_size -gt 0 ]; then
+	echo "pwrite wrote more than zero bytes."
+else
+	echo "pwrite wrote zero bytes"
+fi
+
+# read the partial file to check if data written is of last write
+$XFS_IO_PROG -c "pread 0 1M -v" $testdir/partial | _filter_xfs_io_unique
+
+# success, all done
+status=0
+exit
diff --git a/tests/generic/471.out b/tests/generic/471.out
new file mode 100644
index 00000000..b846eb32
--- /dev/null
+++ b/tests/generic/471.out
@@ -0,0 +1,9 @@
+QA output created by 471
+Reformat with 320M size
+Fill the filesystem
+pwrite: No space left on device
+pwrite wrote more than zero bytes.
+00000000:  bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb  ................
+*
+read 1048576/1048576 bytes at offset 0
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
diff --git a/tests/generic/group b/tests/generic/group
index cf8529ea..160e0e7b 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -473,3 +473,4 @@
 468 shutdown auto quick metadata
 469 auto quick
 470 auto quick rw
+471 auto quick rw
-- 
2.14.2


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* Re: [PATCH 1/3] Check pwrite parameters
  2017-11-29 16:33 [PATCH 1/3] Check pwrite parameters Goldwyn Rodrigues
  2017-11-29 16:33 ` [PATCH 2/3] generic/470: Test RWF_NOWAIT Goldwyn Rodrigues
  2017-11-29 16:33 ` [PATCH 3/3] generic/471: Partial direct write test Goldwyn Rodrigues
@ 2017-12-06  9:55 ` Eryu Guan
  2 siblings, 0 replies; 8+ messages in thread
From: Eryu Guan @ 2017-12-06  9:55 UTC (permalink / raw)
  To: Goldwyn Rodrigues; +Cc: darrick.wong, fstests, Goldwyn Rodrigues

On Wed, Nov 29, 2017 at 10:33:56AM -0600, Goldwyn Rodrigues wrote:
> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> 
> There are some parameters added with xfs_io. Check if the pwrite
> parameters are available. For some cases, xfs_io now returns "command
> -%c not supported", so added "not supported" to count as error.
> 
> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
> 
> Change since v2:
>  - More comments
>  - opts to become local variable
> 
> ---
>  common/rc | 14 +++++++++++++-
>  1 file changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/common/rc b/common/rc
> index 4c053a53..90408b7e 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -2035,6 +2035,7 @@ _require_xfs_io_command()
>  	shift
>  	local param="$*"
>  	local param_checked=0
> +	local opts=""
>  
>  	testfile=$TEST_DIR/$$.xfs_io
>  	case $command in
> @@ -2079,6 +2080,15 @@ _require_xfs_io_command()
>  		echo $testio | grep -q "invalid option" && \
>  			_notrun "xfs_io $command support is missing"
>  		;;
> +	"pwrite")
> +		# -N (RWF_NOWAIT) only works with direct I/O writes
> +		if [ "$param" == "-N" ]; then
> +			opts+=" -d"
> +		fi
> +		testio=`$XFS_IO_PROG -f $opts -c "pwrite $param 0 1M" \
> +			$testfile 2>&1`

Do we need to specify the vector count when testing "-N" support? e.g.

$XFS_IO_PROG -f $opts -c "pwrite $param -V 1 0 1M" ...

>From xfsprogs code, it seems to me that we need to do so

io/pwrite.c::do_pwrite
do_pwrite(
	...
{
	if (!vectors)
		return pwrite(fd, ...);
	return do_pwritev(fd, ..., pritev2_flags);
}

and 'vectors' is a global variable with 0 as default value (no vector
number specified via "-V N" option).

So without "-V N", we call pwrite(2) not pwritev2(2) and RWF_NOWAIT
doesn't get tested.

Thanks,
Eryu

> +		param_checked=1
> +		;;
>  	"scrub"|"repair")
>  		testio=`$XFS_IO_PROG -x -c "$command probe 0" $TEST_DIR 2>&1`
>  		echo $testio | grep -q "Inappropriate ioctl" && \
> @@ -2109,7 +2119,9 @@ _require_xfs_io_command()
>  		$XFS_IO_PROG -c "help $command" | grep -q "^ $param --" || \
>  			_notrun "xfs_io $command doesn't support $param"
>  	else
> -		echo $testio | grep -q "invalid option" && \
> +		# xfs_io could result in "command %c not supported" if it was
> +		# built on kernels not supporting pwritev2() calls
> +		echo $testio | grep -q "\(invalid option\|not supported\)" && \
>  			_notrun "xfs_io $command doesn't support $param"
>  	fi
>  }
> -- 
> 2.14.2
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 2/3] generic/470: Test RWF_NOWAIT
  2017-11-29 16:33 ` [PATCH 2/3] generic/470: Test RWF_NOWAIT Goldwyn Rodrigues
@ 2017-12-06 10:05   ` Eryu Guan
  2017-12-06 16:35     ` Goldwyn Rodrigues
  0 siblings, 1 reply; 8+ messages in thread
From: Eryu Guan @ 2017-12-06 10:05 UTC (permalink / raw)
  To: Goldwyn Rodrigues; +Cc: darrick.wong, fstests, Goldwyn Rodrigues

On Wed, Nov 29, 2017 at 10:33:57AM -0600, Goldwyn Rodrigues wrote:
> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> 
> Tests the RWF_NOWAIT flag so the I/O returns immediately with -EAGAIN
> on a new file since it requires block allocation.
> 
> It creates a file, syncs it, and overwrites the file with RWF_NOWAIT.
> This should succeed.
> 
> Finally, read the contents to make sure the overwrite is successful.
> 
> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
> 
> Changes since v1:
>  - Fix testdir name
>  - use of $XFS_IO_PROG instead of xfs_io
>  - check pwrite accepts -N
> 
> Changes since v2:
>  - corrected test description and improved documentation
>  - new leaner look ;)
>  - rw group
> ---
>  tests/generic/470     | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/generic/470.out | 13 +++++++++
>  tests/generic/group   |  1 +
>  3 files changed, 88 insertions(+)
>  create mode 100755 tests/generic/470
>  create mode 100644 tests/generic/470.out
> 
> diff --git a/tests/generic/470 b/tests/generic/470
> new file mode 100755
> index 00000000..5580718b
> --- /dev/null
> +++ b/tests/generic/470
> @@ -0,0 +1,74 @@
> +#! /bin/bash
> +# FS QA Test No. 470
> +#
> +# write a file with RWF_NOWAIT and it would fail because there are no
> +# blocks allocated. Create a file with direct I/O and re-write it
> +# using RWF_NOWAIT. I/O should finish within 50 microsecods since
> +# block allocations are already performed.
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2017, SUSE Linux Products.  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!
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/populate
> +. ./common/filter
> +. ./common/attr
> +
> +# real QA test starts here
> +_supported_os Linux
> +_require_odirect
> +_require_test
> +_require_xfs_io_command pwrite -N
> +
> +# Remove reminiscence of previously run tests
> +testdir=$TEST_DIR/$seq
> +if [ -e $testdir ]; then
> +	rm -Rf $testdir
> +fi
> +
> +mkdir $testdir
> +
> +# Create a file with pwrite nowait (will fail with EAGAIN)
> +$XFS_IO_PROG -f -d -c "pwrite -N -V 1 -b 1M 0 1M" $testdir/f1

It's weird that I got "Operation not supported" message here and test
failed as

@@ -1,5 +1,5 @@
 QA output created by 470
-pwrite: Resource temporarily unavailable
+pwrite: Operation not supported
 wrote 8388608/8388608 bytes at offset 0
 XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 RWF_NOWAIT time is within limits.

I was using latest for-next branch of upstream xfsprogs, and a
pre-v4.15-rc1 kernel shipped by Fedora rawhide. A simple test showed:

rm -f /mnt/xfs/testfile
xfs_io -fdc "pwrite -N -V 1 -b 4k 0 4k" /mnt/xfs/testfile
pwrite: Operation not supported

But an strace run showed pwritev2 did -1 and set EAGAIN as errno, but
printed "Operation not supported" as error message:
...
pwritev2(3, [{iov_base="\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315"..., iov_len=4096}], 1, 0, RWF_NOWAIT) = -1 EAGAIN (Resource temporarily unavailable)
dup(2)                                  = 4
fcntl(4, F_GETFL)                       = 0x402 (flags O_RDWR|O_APPEND)
openat(AT_FDCWD, "/usr/share/locale/en_US.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/share/locale/en_US.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/share/locale/en_US/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/share/locale/en.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/share/locale/en.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/usr/share/locale/en/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
fstat(4, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 0), ...}) = 0
write(4, "pwrite: Operation not supported\n", 32pwrite: Operation not supported
...

Not sure what happened yet, did I miss anything?

> +
> +# Write the file without nowait
> +$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -W -w -V 1 -b 1M 0 8M" $testdir/f1 | _filter_xfs_io

I think "-w" (fdatasync) can be removed? We already use "-W" (fsync) here.

> +
> +time_taken=`$XFS_IO_PROG -d -c "pwrite -S 0xbb -N -V 1 -b 1M 2M 1M" $testdir/f1 | awk '/^1/ {print $5}'`
> +
> +# RWF_NOWAIT should finish within a short period of time. Anything longer
> +# means it is waiting for something in the kernel which would be a fail.
> +if (( $(echo "$time_taken < 0.05" | bc -l) )); then

Better to describe where does this 0.05 come from.

Thanks,
Eryu

> +	echo "RWF_NOWAIT time is within limits."
> +else
> +	echo "RWF_NOWAIT took $time_taken seconds"
> +fi
> +
> +$XFS_IO_PROG -c "pread -v 0 8M" $testdir/f1 | _filter_xfs_io_unique
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/generic/470.out b/tests/generic/470.out
> new file mode 100644
> index 00000000..e46622cb
> --- /dev/null
> +++ b/tests/generic/470.out
> @@ -0,0 +1,13 @@
> +QA output created by 470
> +pwrite: Resource temporarily unavailable
> +wrote 8388608/8388608 bytes at offset 0
> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +RWF_NOWAIT time is within limits.
> +00000000:  aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa  ................
> +*
> +00200000:  bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb  ................
> +*
> +00300000:  aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa  ................
> +*
> +read 8388608/8388608 bytes at offset 0
> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> diff --git a/tests/generic/group b/tests/generic/group
> index 6c3bb03a..cf8529ea 100644
> --- a/tests/generic/group
> +++ b/tests/generic/group
> @@ -472,3 +472,4 @@
>  467 auto quick exportfs
>  468 shutdown auto quick metadata
>  469 auto quick
> +470 auto quick rw
> -- 
> 2.14.2
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 2/3] generic/470: Test RWF_NOWAIT
  2017-12-06 10:05   ` Eryu Guan
@ 2017-12-06 16:35     ` Goldwyn Rodrigues
  2017-12-07  4:17       ` Eryu Guan
  0 siblings, 1 reply; 8+ messages in thread
From: Goldwyn Rodrigues @ 2017-12-06 16:35 UTC (permalink / raw)
  To: Eryu Guan; +Cc: darrick.wong, fstests, Goldwyn Rodrigues



On 12/06/2017 04:05 AM, Eryu Guan wrote:
> On Wed, Nov 29, 2017 at 10:33:57AM -0600, Goldwyn Rodrigues wrote:
>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>
>> Tests the RWF_NOWAIT flag so the I/O returns immediately with -EAGAIN
>> on a new file since it requires block allocation.
>>
>> It creates a file, syncs it, and overwrites the file with RWF_NOWAIT.
>> This should succeed.
>>
>> Finally, read the contents to make sure the overwrite is successful.
>>
>> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>
>> Changes since v1:
>>  - Fix testdir name
>>  - use of $XFS_IO_PROG instead of xfs_io
>>  - check pwrite accepts -N
>>
>> Changes since v2:
>>  - corrected test description and improved documentation
>>  - new leaner look ;)
>>  - rw group
>> ---
>>  tests/generic/470     | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>  tests/generic/470.out | 13 +++++++++
>>  tests/generic/group   |  1 +
>>  3 files changed, 88 insertions(+)
>>  create mode 100755 tests/generic/470
>>  create mode 100644 tests/generic/470.out
>>
>> diff --git a/tests/generic/470 b/tests/generic/470
>> new file mode 100755
>> index 00000000..5580718b
>> --- /dev/null
>> +++ b/tests/generic/470
>> @@ -0,0 +1,74 @@
>> +#! /bin/bash
>> +# FS QA Test No. 470
>> +#
>> +# write a file with RWF_NOWAIT and it would fail because there are no
>> +# blocks allocated. Create a file with direct I/O and re-write it
>> +# using RWF_NOWAIT. I/O should finish within 50 microsecods since
>> +# block allocations are already performed.
>> +#
>> +#-----------------------------------------------------------------------
>> +# Copyright (c) 2017, SUSE Linux Products.  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!
>> +
>> +# get standard environment, filters and checks
>> +. ./common/rc
>> +. ./common/populate
>> +. ./common/filter
>> +. ./common/attr
>> +
>> +# real QA test starts here
>> +_supported_os Linux
>> +_require_odirect
>> +_require_test
>> +_require_xfs_io_command pwrite -N
>> +
>> +# Remove reminiscence of previously run tests
>> +testdir=$TEST_DIR/$seq
>> +if [ -e $testdir ]; then
>> +	rm -Rf $testdir
>> +fi
>> +
>> +mkdir $testdir
>> +
>> +# Create a file with pwrite nowait (will fail with EAGAIN)
>> +$XFS_IO_PROG -f -d -c "pwrite -N -V 1 -b 1M 0 1M" $testdir/f1
> 
> It's weird that I got "Operation not supported" message here and test
> failed as
> 
> @@ -1,5 +1,5 @@
>  QA output created by 470
> -pwrite: Resource temporarily unavailable
> +pwrite: Operation not supported
>  wrote 8388608/8388608 bytes at offset 0
>  XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>  RWF_NOWAIT time is within limits.
> 
> I was using latest for-next branch of upstream xfsprogs, and a
> pre-v4.15-rc1 kernel shipped by Fedora rawhide. A simple test showed:

What is the output of uname -a
It should be compatible with 4.13 onward. OTOH, _require_xfs_io_command
will catch that.

> 
> rm -f /mnt/xfs/testfile
> xfs_io -fdc "pwrite -N -V 1 -b 4k 0 4k" /mnt/xfs/testfile
> pwrite: Operation not supported

For kernels which do not support RWF_NOWAIT, it will return EOPNOTSUPP.

> 
> But an strace run showed pwritev2 did -1 and set EAGAIN as errno, but
> printed "Operation not supported" as error message:
> ...
> pwritev2(3, [{iov_base="\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315"..., iov_len=4096}], 1, 0, RWF_NOWAIT) = -1 EAGAIN (Resource temporarily unavailable)
> dup(2)                                  = 4
> fcntl(4, F_GETFL)                       = 0x402 (flags O_RDWR|O_APPEND)
> openat(AT_FDCWD, "/usr/share/locale/en_US.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> openat(AT_FDCWD, "/usr/share/locale/en_US.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> openat(AT_FDCWD, "/usr/share/locale/en_US/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> openat(AT_FDCWD, "/usr/share/locale/en.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> openat(AT_FDCWD, "/usr/share/locale/en.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> openat(AT_FDCWD, "/usr/share/locale/en/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> fstat(4, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 0), ...}) = 0
> write(4, "pwrite: Operation not supported\n", 32pwrite: Operation not supported
> ...
> 
> Not sure what happened yet, did I miss anything?

Did you trim any lines in between? If you had another error in between
before you collect -EAGAIN, it most likely overwrote the value. Not sure
what is going on though..

> 
>> +
>> +# Write the file without nowait
>> +$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -W -w -V 1 -b 1M 0 8M" $testdir/f1 | _filter_xfs_io
> 
> I think "-w" (fdatasync) can be removed? We already use "-W" (fsync) here.
> 
>> +
>> +time_taken=`$XFS_IO_PROG -d -c "pwrite -S 0xbb -N -V 1 -b 1M 2M 1M" $testdir/f1 | awk '/^1/ {print $5}'`
>> +
>> +# RWF_NOWAIT should finish within a short period of time. Anything longer
>> +# means it is waiting for something in the kernel which would be a fail.
>> +if (( $(echo "$time_taken < 0.05" | bc -l) )); then
> 
> Better to describe where does this 0.05 come from.
> 

No significance as such, but a conservative value for the operation to
be completed. If you want me to explicitly put 50 ms, I can.

> Thanks,
> Eryu
> 
>> +	echo "RWF_NOWAIT time is within limits."
>> +else
>> +	echo "RWF_NOWAIT took $time_taken seconds"
>> +fi
>> +
>> +$XFS_IO_PROG -c "pread -v 0 8M" $testdir/f1 | _filter_xfs_io_unique
>> +
>> +# success, all done
>> +status=0
>> +exit
>> diff --git a/tests/generic/470.out b/tests/generic/470.out
>> new file mode 100644
>> index 00000000..e46622cb
>> --- /dev/null
>> +++ b/tests/generic/470.out
>> @@ -0,0 +1,13 @@
>> +QA output created by 470
>> +pwrite: Resource temporarily unavailable
>> +wrote 8388608/8388608 bytes at offset 0
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> +RWF_NOWAIT time is within limits.
>> +00000000:  aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa  ................
>> +*
>> +00200000:  bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb  ................
>> +*
>> +00300000:  aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa  ................
>> +*
>> +read 8388608/8388608 bytes at offset 0
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> diff --git a/tests/generic/group b/tests/generic/group
>> index 6c3bb03a..cf8529ea 100644
>> --- a/tests/generic/group
>> +++ b/tests/generic/group
>> @@ -472,3 +472,4 @@
>>  467 auto quick exportfs
>>  468 shutdown auto quick metadata
>>  469 auto quick
>> +470 auto quick rw
>> -- 
>> 2.14.2
>>

-- 
Goldwyn

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 2/3] generic/470: Test RWF_NOWAIT
  2017-12-06 16:35     ` Goldwyn Rodrigues
@ 2017-12-07  4:17       ` Eryu Guan
  2017-12-07 11:57         ` Goldwyn Rodrigues
  0 siblings, 1 reply; 8+ messages in thread
From: Eryu Guan @ 2017-12-07  4:17 UTC (permalink / raw)
  To: Goldwyn Rodrigues; +Cc: darrick.wong, fstests, Goldwyn Rodrigues

On Wed, Dec 06, 2017 at 10:35:06AM -0600, Goldwyn Rodrigues wrote:
> 
> 
> On 12/06/2017 04:05 AM, Eryu Guan wrote:
> > On Wed, Nov 29, 2017 at 10:33:57AM -0600, Goldwyn Rodrigues wrote:
> >> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> >>
> >> Tests the RWF_NOWAIT flag so the I/O returns immediately with -EAGAIN
> >> on a new file since it requires block allocation.
> >>
> >> It creates a file, syncs it, and overwrites the file with RWF_NOWAIT.
> >> This should succeed.
> >>
> >> Finally, read the contents to make sure the overwrite is successful.
> >>
> >> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
> >>
> >> Changes since v1:
> >>  - Fix testdir name
> >>  - use of $XFS_IO_PROG instead of xfs_io
> >>  - check pwrite accepts -N
> >>
> >> Changes since v2:
> >>  - corrected test description and improved documentation
> >>  - new leaner look ;)
> >>  - rw group
> >> ---
> >>  tests/generic/470     | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++
> >>  tests/generic/470.out | 13 +++++++++
> >>  tests/generic/group   |  1 +
> >>  3 files changed, 88 insertions(+)
> >>  create mode 100755 tests/generic/470
> >>  create mode 100644 tests/generic/470.out
> >>
> >> diff --git a/tests/generic/470 b/tests/generic/470
> >> new file mode 100755
> >> index 00000000..5580718b
> >> --- /dev/null
> >> +++ b/tests/generic/470
> >> @@ -0,0 +1,74 @@
> >> +#! /bin/bash
> >> +# FS QA Test No. 470
> >> +#
> >> +# write a file with RWF_NOWAIT and it would fail because there are no
> >> +# blocks allocated. Create a file with direct I/O and re-write it
> >> +# using RWF_NOWAIT. I/O should finish within 50 microsecods since
> >> +# block allocations are already performed.
> >> +#
> >> +#-----------------------------------------------------------------------
> >> +# Copyright (c) 2017, SUSE Linux Products.  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!
> >> +
> >> +# get standard environment, filters and checks
> >> +. ./common/rc
> >> +. ./common/populate
> >> +. ./common/filter
> >> +. ./common/attr
> >> +
> >> +# real QA test starts here
> >> +_supported_os Linux
> >> +_require_odirect
> >> +_require_test
> >> +_require_xfs_io_command pwrite -N
> >> +
> >> +# Remove reminiscence of previously run tests
> >> +testdir=$TEST_DIR/$seq
> >> +if [ -e $testdir ]; then
> >> +	rm -Rf $testdir
> >> +fi
> >> +
> >> +mkdir $testdir
> >> +
> >> +# Create a file with pwrite nowait (will fail with EAGAIN)
> >> +$XFS_IO_PROG -f -d -c "pwrite -N -V 1 -b 1M 0 1M" $testdir/f1
> > 
> > It's weird that I got "Operation not supported" message here and test
> > failed as
> > 
> > @@ -1,5 +1,5 @@
> >  QA output created by 470
> > -pwrite: Resource temporarily unavailable
> > +pwrite: Operation not supported
> >  wrote 8388608/8388608 bytes at offset 0
> >  XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> >  RWF_NOWAIT time is within limits.
> > 
> > I was using latest for-next branch of upstream xfsprogs, and a
> > pre-v4.15-rc1 kernel shipped by Fedora rawhide. A simple test showed:
> 
> What is the output of uname -a
> It should be compatible with 4.13 onward. OTOH, _require_xfs_io_command
> will catch that.

Linux localhost 4.15.0-0.rc0.git6.1.fc28.x86_64 #1 SMP Mon Nov 20 18:08:48 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

_require_xfs_io_command doesn't run "pwrite -N" with "-V 1" option, so
test there passes, it only fails with vectors specified

# rm /mnt/xfs/testfile
# xfs_io -fdc "pwrite -N 0 4k" /mnt/xfs/testfile
wrote 4096/4096 bytes at offset 0
4 KiB, 1 ops; 0.0000 sec (787.402 KiB/sec and 196.8504 ops/sec)
# rm /mnt/xfs/testfile
# xfs_io -fdc "pwrite -N -V 1 0 4k" /mnt/xfs/testfile
pwrite: Operation not supported

That's why I was asking if we should include "-V 1" in
_require_xfs_io_command in patch 1.

> 
> > 
> > rm -f /mnt/xfs/testfile
> > xfs_io -fdc "pwrite -N -V 1 -b 4k 0 4k" /mnt/xfs/testfile
> > pwrite: Operation not supported
> 
> For kernels which do not support RWF_NOWAIT, it will return EOPNOTSUPP.
> 
> > 
> > But an strace run showed pwritev2 did -1 and set EAGAIN as errno, but
> > printed "Operation not supported" as error message:
> > ...
> > pwritev2(3, [{iov_base="\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315"..., iov_len=4096}], 1, 0, RWF_NOWAIT) = -1 EAGAIN (Resource temporarily unavailable)
> > dup(2)                                  = 4
> > fcntl(4, F_GETFL)                       = 0x402 (flags O_RDWR|O_APPEND)
> > openat(AT_FDCWD, "/usr/share/locale/en_US.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> > openat(AT_FDCWD, "/usr/share/locale/en_US.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> > openat(AT_FDCWD, "/usr/share/locale/en_US/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> > openat(AT_FDCWD, "/usr/share/locale/en.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> > openat(AT_FDCWD, "/usr/share/locale/en.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> > openat(AT_FDCWD, "/usr/share/locale/en/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
> > fstat(4, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 0), ...}) = 0
> > write(4, "pwrite: Operation not supported\n", 32pwrite: Operation not supported
> > ...
> > 
> > Not sure what happened yet, did I miss anything?
> 
> Did you trim any lines in between? If you had another error in between
> before you collect -EAGAIN, it most likely overwrote the value. Not sure
> what is going on though..

No, I didn't trim any lines in between.

> 
> > 
> >> +
> >> +# Write the file without nowait
> >> +$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -W -w -V 1 -b 1M 0 8M" $testdir/f1 | _filter_xfs_io
> > 
> > I think "-w" (fdatasync) can be removed? We already use "-W" (fsync) here.
> > 
> >> +
> >> +time_taken=`$XFS_IO_PROG -d -c "pwrite -S 0xbb -N -V 1 -b 1M 2M 1M" $testdir/f1 | awk '/^1/ {print $5}'`
> >> +
> >> +# RWF_NOWAIT should finish within a short period of time. Anything longer
> >> +# means it is waiting for something in the kernel which would be a fail.
> >> +if (( $(echo "$time_taken < 0.05" | bc -l) )); then
> > 
> > Better to describe where does this 0.05 come from.
> > 
> 
> No significance as such, but a conservative value for the operation to
> be completed. If you want me to explicitly put 50 ms, I can.

There should be a reason for this 50ms estimation. IMHO, as simple as "a
conservative value for the operation to be completed" would be better
than nothing :)

Thanks,
Eryu

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 2/3] generic/470: Test RWF_NOWAIT
  2017-12-07  4:17       ` Eryu Guan
@ 2017-12-07 11:57         ` Goldwyn Rodrigues
  0 siblings, 0 replies; 8+ messages in thread
From: Goldwyn Rodrigues @ 2017-12-07 11:57 UTC (permalink / raw)
  To: Eryu Guan, Goldwyn Rodrigues; +Cc: darrick.wong, fstests



On 12/06/2017 10:17 PM, Eryu Guan wrote:
> On Wed, Dec 06, 2017 at 10:35:06AM -0600, Goldwyn Rodrigues wrote:
>>
>>
>> On 12/06/2017 04:05 AM, Eryu Guan wrote:
>>> On Wed, Nov 29, 2017 at 10:33:57AM -0600, Goldwyn Rodrigues wrote:
>>>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>>>
>>>> Tests the RWF_NOWAIT flag so the I/O returns immediately with -EAGAIN
>>>> on a new file since it requires block allocation.
>>>>
>>>> It creates a file, syncs it, and overwrites the file with RWF_NOWAIT.
>>>> This should succeed.
>>>>
>>>> Finally, read the contents to make sure the overwrite is successful.
>>>>
>>>> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>>>
>>>> Changes since v1:
>>>>  - Fix testdir name
>>>>  - use of $XFS_IO_PROG instead of xfs_io
>>>>  - check pwrite accepts -N
>>>>
>>>> Changes since v2:
>>>>  - corrected test description and improved documentation
>>>>  - new leaner look ;)
>>>>  - rw group
>>>> ---
>>>>  tests/generic/470     | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>  tests/generic/470.out | 13 +++++++++
>>>>  tests/generic/group   |  1 +
>>>>  3 files changed, 88 insertions(+)
>>>>  create mode 100755 tests/generic/470
>>>>  create mode 100644 tests/generic/470.out
>>>>
>>>> diff --git a/tests/generic/470 b/tests/generic/470
>>>> new file mode 100755
>>>> index 00000000..5580718b
>>>> --- /dev/null
>>>> +++ b/tests/generic/470
>>>> @@ -0,0 +1,74 @@
>>>> +#! /bin/bash
>>>> +# FS QA Test No. 470
>>>> +#
>>>> +# write a file with RWF_NOWAIT and it would fail because there are no
>>>> +# blocks allocated. Create a file with direct I/O and re-write it
>>>> +# using RWF_NOWAIT. I/O should finish within 50 microsecods since
>>>> +# block allocations are already performed.
>>>> +#
>>>> +#-----------------------------------------------------------------------
>>>> +# Copyright (c) 2017, SUSE Linux Products.  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!
>>>> +
>>>> +# get standard environment, filters and checks
>>>> +. ./common/rc
>>>> +. ./common/populate
>>>> +. ./common/filter
>>>> +. ./common/attr
>>>> +
>>>> +# real QA test starts here
>>>> +_supported_os Linux
>>>> +_require_odirect
>>>> +_require_test
>>>> +_require_xfs_io_command pwrite -N
>>>> +
>>>> +# Remove reminiscence of previously run tests
>>>> +testdir=$TEST_DIR/$seq
>>>> +if [ -e $testdir ]; then
>>>> +	rm -Rf $testdir
>>>> +fi
>>>> +
>>>> +mkdir $testdir
>>>> +
>>>> +# Create a file with pwrite nowait (will fail with EAGAIN)
>>>> +$XFS_IO_PROG -f -d -c "pwrite -N -V 1 -b 1M 0 1M" $testdir/f1
>>>
>>> It's weird that I got "Operation not supported" message here and test
>>> failed as
>>>
>>> @@ -1,5 +1,5 @@
>>>  QA output created by 470
>>> -pwrite: Resource temporarily unavailable
>>> +pwrite: Operation not supported
>>>  wrote 8388608/8388608 bytes at offset 0
>>>  XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>>>  RWF_NOWAIT time is within limits.
>>>
>>> I was using latest for-next branch of upstream xfsprogs, and a
>>> pre-v4.15-rc1 kernel shipped by Fedora rawhide. A simple test showed:
>>
>> What is the output of uname -a
>> It should be compatible with 4.13 onward. OTOH, _require_xfs_io_command
>> will catch that.
> 
> Linux localhost 4.15.0-0.rc0.git6.1.fc28.x86_64 #1 SMP Mon Nov 20 18:08:48 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux
> 
> _require_xfs_io_command doesn't run "pwrite -N" with "-V 1" option, so
> test there passes, it only fails with vectors specified
> 
> # rm /mnt/xfs/testfile
> # xfs_io -fdc "pwrite -N 0 4k" /mnt/xfs/testfile
> wrote 4096/4096 bytes at offset 0
> 4 KiB, 1 ops; 0.0000 sec (787.402 KiB/sec and 196.8504 ops/sec)
> # rm /mnt/xfs/testfile
> # xfs_io -fdc "pwrite -N -V 1 0 4k" /mnt/xfs/testfile
> pwrite: Operation not supported
> 
> That's why I was asking if we should include "-V 1" in
> _require_xfs_io_command in patch 1.
> 

I am all for including your suggestion in patch 1. However, it does not
explain two things:
1. Why is the error code in strace changing from -EAGAIN to -EOPNOTSUPP?
2. Why is it happening on a kernel version greater than 4.14?

Anyways, I shall add the changes and send it across again.


>>
>>>
>>> rm -f /mnt/xfs/testfile
>>> xfs_io -fdc "pwrite -N -V 1 -b 4k 0 4k" /mnt/xfs/testfile
>>> pwrite: Operation not supported
>>
>> For kernels which do not support RWF_NOWAIT, it will return EOPNOTSUPP.
>>
>>>
>>> But an strace run showed pwritev2 did -1 and set EAGAIN as errno, but
>>> printed "Operation not supported" as error message:
>>> ...
>>> pwritev2(3, [{iov_base="\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315\315"..., iov_len=4096}], 1, 0, RWF_NOWAIT) = -1 EAGAIN (Resource temporarily unavailable)
>>> dup(2)                                  = 4
>>> fcntl(4, F_GETFL)                       = 0x402 (flags O_RDWR|O_APPEND)
>>> openat(AT_FDCWD, "/usr/share/locale/en_US.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
>>> openat(AT_FDCWD, "/usr/share/locale/en_US.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
>>> openat(AT_FDCWD, "/usr/share/locale/en_US/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
>>> openat(AT_FDCWD, "/usr/share/locale/en.UTF-8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
>>> openat(AT_FDCWD, "/usr/share/locale/en.utf8/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
>>> openat(AT_FDCWD, "/usr/share/locale/en/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
>>> fstat(4, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 0), ...}) = 0
>>> write(4, "pwrite: Operation not supported\n", 32pwrite: Operation not supported
>>> ...
>>>
>>> Not sure what happened yet, did I miss anything?
>>
>> Did you trim any lines in between? If you had another error in between
>> before you collect -EAGAIN, it most likely overwrote the value. Not sure
>> what is going on though..
> 
> No, I didn't trim any lines in between.
> 
>>
>>>
>>>> +
>>>> +# Write the file without nowait
>>>> +$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -W -w -V 1 -b 1M 0 8M" $testdir/f1 | _filter_xfs_io
>>>
>>> I think "-w" (fdatasync) can be removed? We already use "-W" (fsync) here.
>>>
>>>> +
>>>> +time_taken=`$XFS_IO_PROG -d -c "pwrite -S 0xbb -N -V 1 -b 1M 2M 1M" $testdir/f1 | awk '/^1/ {print $5}'`
>>>> +
>>>> +# RWF_NOWAIT should finish within a short period of time. Anything longer
>>>> +# means it is waiting for something in the kernel which would be a fail.
>>>> +if (( $(echo "$time_taken < 0.05" | bc -l) )); then
>>>
>>> Better to describe where does this 0.05 come from.
>>>
>>
>> No significance as such, but a conservative value for the operation to
>> be completed. If you want me to explicitly put 50 ms, I can.
> 
> There should be a reason for this 50ms estimation. IMHO, as simple as "a
> conservative value for the operation to be completed" would be better
> than nothing :)
> 
> Thanks,
> Eryu
> 

-- 
Goldwyn

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2017-12-07 11:57 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-11-29 16:33 [PATCH 1/3] Check pwrite parameters Goldwyn Rodrigues
2017-11-29 16:33 ` [PATCH 2/3] generic/470: Test RWF_NOWAIT Goldwyn Rodrigues
2017-12-06 10:05   ` Eryu Guan
2017-12-06 16:35     ` Goldwyn Rodrigues
2017-12-07  4:17       ` Eryu Guan
2017-12-07 11:57         ` Goldwyn Rodrigues
2017-11-29 16:33 ` [PATCH 3/3] generic/471: Partial direct write test Goldwyn Rodrigues
2017-12-06  9:55 ` [PATCH 1/3] Check pwrite parameters Eryu Guan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox