* [PATCH 1/2] generic/461: Test RWF_NOWAIT
@ 2017-09-27 19:10 Goldwyn Rodrigues
2017-09-27 19:10 ` [PATCH 2/2] generic/462: Partial direct write test Goldwyn Rodrigues
2017-09-27 21:34 ` [PATCH 1/2] generic/461: Test RWF_NOWAIT Dave Chinner
0 siblings, 2 replies; 9+ messages in thread
From: Goldwyn Rodrigues @ 2017-09-27 19:10 UTC (permalink / raw)
To: fstests; +Cc: Goldwyn Rodrigues
From: Goldwyn Rodrigues <rgoldwyn@suse.com>
Tests the RWF_NOWAIT flag so the I/O returns immediately on
a new file, without any block allocations.
A new program which includes the pwritev2() call is used. This allows
passing flags for the I/O to be performed.
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
---
src/Makefile | 2 +-
src/pwritev2.c | 161 ++++++++++++++++++++++++++++++++++++++++++++++++++
tests/generic/461 | 80 +++++++++++++++++++++++++
tests/generic/461.out | 9 +++
tests/generic/group | 1 +
5 files changed, 252 insertions(+), 1 deletion(-)
create mode 100644 src/pwritev2.c
create mode 100755 tests/generic/461
create mode 100644 tests/generic/461.out
diff --git a/src/Makefile b/src/Makefile
index 7d1306bc..427c561e 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -23,7 +23,7 @@ LINUX_TARGETS = xfsctl bstat t_mtab getdevicesize preallo_rw_pattern_reader \
seek_copy_test t_readdir_1 t_readdir_2 fsync-tester nsexec cloner \
renameat2 t_getcwd e4compact test-nextquota punch-alternating \
attr-list-by-handle-cursor-test listxattr dio-interleaved t_dir_type \
- dio-invalidate-cache stat_test t_encrypted_d_revalidate
+ dio-invalidate-cache stat_test t_encrypted_d_revalidate pwritev2
SUBDIRS = log-writes
diff --git a/src/pwritev2.c b/src/pwritev2.c
new file mode 100644
index 00000000..28413910
--- /dev/null
+++ b/src/pwritev2.c
@@ -0,0 +1,161 @@
+#define _GNU_SOURCE
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/uio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <linux/fs.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <string.h>
+#include <sys/time.h>
+
+int openflag = O_RDWR | O_LARGEFILE;
+long datasz = 4096;
+long iovcnt = 1;
+long offset = 0;
+int readfile = 0;
+int measure_time = 0;
+int flags = 0;
+char *filename = "pwrite-file";
+unsigned int seed = 0xcdcdcdcd;
+
+int usage()
+{
+ fprintf(stderr, "Usage: pwritev2 \n");
+ fprintf(stderr, "\t-b [buffer size]: Size of each buffer\n");
+ fprintf(stderr, "\t-d: Direct write\n");
+ fprintf(stderr, "\t-f: create file if not exists\n");
+ fprintf(stderr, "\t-s [data size]: Size of each iovec data buffer\n");
+ fprintf(stderr, "\t-S [seed]: seed for the data buffer\n");
+ fprintf(stderr, "\t-t: measure time\n");
+ fprintf(stderr, "\t-o [offset]: Offset where to perform the pwritev2\n");
+ fprintf(stderr, "\t-V [num of iovecs]: Number of vectors\n");
+ fprintf(stderr, "\t-W: perform I/O with RWF_NOWAIT\n");
+ fprintf(stderr, "\t filename: The filename to perform pwritev2\n");
+ exit(1);
+}
+
+/*
+ * Scale value by kilo, mega, or giga.
+ */
+long long scale_by_kmg(long long value, char scale)
+{
+ switch (scale) {
+ case 'g':
+ case 'G':
+ value *= 1024;
+ case 'm':
+ case 'M':
+ value *= 1024;
+ case 'k':
+ case 'K':
+ value *= 1024;
+ break;
+ case '\0':
+ break;
+ default:
+ usage();
+ break;
+ }
+ return value;
+}
+
+
+int parse_args(int argc, char *argv[])
+{
+ int c;
+ extern int optind;
+ while ((c = getopt(argc, argv, "b:dfo:V:S:tW")) != -1) {
+ char *endp;
+ switch (c) {
+ case 'b':
+ datasz = strtol(optarg, &endp, 0);
+ datasz = (long)scale_by_kmg((long long)datasz,
+ *endp);
+ break;
+ case 'd':
+ openflag |= O_DIRECT;
+ break;
+ case 'f':
+ openflag |= O_CREAT;
+ break;
+ case 'o':
+ offset = strtol(optarg, &endp, 0);
+ break;
+ case 't':
+ measure_time = 1;
+ break;
+ case 'S':
+ seed = strtoul(optarg, &endp, 0);
+ break;
+ case 'V':
+ iovcnt = strtol(optarg, &endp, 0);
+ break;
+ case 'W':
+ flags |= RWF_NOWAIT;
+ break;
+ default:
+ usage();
+ }
+ }
+ filename = argv[optind];
+}
+
+
+
+int main(int argc, char *argv[])
+{
+ int i, fd;
+ size_t ret = 0;
+ struct iovec *iovec;
+ char *data;
+ struct timeval before, after;
+ struct timezone tz;
+
+ parse_args(argc, argv);
+
+ iovec = (struct iovec *)malloc(sizeof(struct iovec) * iovcnt);
+ if (!iovec) {
+ fprintf(stderr, "Could not allocate iovec\n");
+ exit(1);
+ }
+ data = aligned_alloc(datasz, datasz);
+ if (!data) {
+ fprintf(stderr, "Could not allocate data\n");
+ exit(1);
+ }
+
+ memset(data, seed, datasz);
+
+ for (i=0; i<iovcnt; i++) {
+ iovec[i].iov_base = data;
+ iovec[i].iov_len = datasz;
+ }
+
+ fd = open(filename, openflag, S_IRUSR | S_IWUSR);
+
+ gettimeofday(&before, &tz);
+ ret = pwritev2(fd, iovec, iovcnt, offset, flags);
+ gettimeofday(&after, &tz);
+ printf("pwritev2() ret = %ld %s(%d)\n", ret, strerror(errno), errno);
+
+ if (measure_time) {
+ unsigned long seconds, ms;
+ seconds = after.tv_sec - before.tv_sec;
+ if (after.tv_usec < before.tv_usec) {
+ seconds -= 1;
+ ms = 1000000 + after.tv_usec - before.tv_usec;
+ } else
+ ms = after.tv_usec - before.tv_usec;
+ printf("Time: %lu seconds %lu microseconds\n", seconds, ms);
+ }
+
+ close(fd);
+ return 0;
+
+}
+
+
diff --git a/tests/generic/461 b/tests/generic/461
new file mode 100755
index 00000000..7bbedd5d
--- /dev/null
+++ b/tests/generic/461
@@ -0,0 +1,80 @@
+#! /bin/bash
+# FS QA Test No. 461
+#
+# write() to a file opened with O_DIRECT with count > remaining
+# bytes. Result should be remaining bytes (to aligned bytes) instead
+# of ENOSPC error
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2015, Oracle and/or its affiliates. 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
+
+rm -f $seqres.full
+
+echo "Format and mount"
+_scratch_mkfs > $seqres.full 2>&1
+_scratch_mount >> $seqres.full 2>&1
+
+testdir=$SCRATCH_MNT/test-$seq
+mkdir $testdir
+
+# Create a file with pwrite nowait (will fail with EAGAIN)
+out=`src/pwritev2 -f -W -d -V 1 -b 128M -t $testdir/f1 | grep Time`
+microsecs=`echo $out | cut -f 4 -d " "`
+secs=`echo $out | cut -f 2 -d " "`
+if [ $microsecs -lt 50 -a $secs -eq 0 ]; then
+ echo "RWF_NOWAIT time is within limits."
+else
+ echo "RWF_NOWAIT took $secs seconds and $microsecs microsecs"
+fi
+
+# Write the file without nowait
+src/pwritev2 -S 0xaa -d -V 1 -b 128M $testdir/f1
+
+# Write the file with nowait
+src/pwritev2 -W -d -V 1 -b 128M $testdir/f1
+
+# Read the file to check if data is correct
+$XFS_IO_PROG -d -c "pread 0 4k -v" $testdir/f1 | _filter_xfs_io_unique
+
+# success, all done
+status=0
+exit
diff --git a/tests/generic/461.out b/tests/generic/461.out
new file mode 100644
index 00000000..1c331d5b
--- /dev/null
+++ b/tests/generic/461.out
@@ -0,0 +1,9 @@
+QA output created by 461
+Format and mount
+RWF_NOWAIT time is within limits.
+pwritev2() ret = 134217728 Success(0)
+pwritev2() ret = 134217728 Success(0)
+00000000: cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd ................
+*
+read 4096/4096 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 f922b496..6352064a 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -463,3 +463,4 @@
458 auto quick clone
459 auto dangerous
460 auto quick rw
+461 auto quick rw
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* [PATCH 2/2] generic/462: Partial direct write test
2017-09-27 19:10 [PATCH 1/2] generic/461: Test RWF_NOWAIT Goldwyn Rodrigues
@ 2017-09-27 19:10 ` Goldwyn Rodrigues
2017-09-27 21:34 ` [PATCH 1/2] generic/461: Test RWF_NOWAIT Dave Chinner
1 sibling, 0 replies; 9+ messages in thread
From: Goldwyn Rodrigues @ 2017-09-27 19:10 UTC (permalink / raw)
To: fstests; +Cc: 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.
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
---
tests/generic/462 | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++
tests/generic/462.out | 9 ++++++
tests/generic/group | 1 +
3 files changed, 100 insertions(+)
create mode 100755 tests/generic/462
create mode 100644 tests/generic/462.out
diff --git a/tests/generic/462 b/tests/generic/462
new file mode 100755
index 00000000..c6c5e7fc
--- /dev/null
+++ b/tests/generic/462
@@ -0,0 +1,90 @@
+#! /bin/bash
+# FS QA Test No. 461
+#
+# write() to a file opened with O_DIRECT with count > remaining
+# bytes. Result should be remaining bytes (to aligned bytes) instead
+# of ENOSPC error
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2015, Oracle and/or its affiliates. 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
+
+rm -f $seqres.full
+
+echo "Reformat with 320M size"
+umount $SCRATCH_MNT
+sz_bytes=$((320 * 1024 * 1024))
+_scratch_mkfs_sized $sz_bytes >> $seqres.full 2>&1
+_scratch_mount >> $seqres.full 2>&1
+testdir=$SCRATCH_MNT/test-$seq
+rm -rf $testdir
+mkdir $testdir
+sync
+
+#Almost fill the filesystem
+echo "Almost fill the filesystem"
+for i in `seq 1 5`; do
+ $XFS_IO_PROG -f -c "pwrite 0 40M" $testdir/file-$i > /dev/null
+done
+
+# Create a file using direct I/O which succeeds only partially
+$XFS_IO_PROG -f -d -c "pwrite -S 0xaa -b 128M -V 1 0 128M" $testdir/partial > /dev/null 2>&1
+partial_size=`stat $testdir/partial | grep "Size" | tr -s " " | cut -f 3 -d " "`
+if [ $partial_size -gt 0 ]; then
+ echo "File partial size is greater than zero."
+else
+ echo "File partial size is zero"
+fi
+
+# re-write the file again with default pattern using one single buffer
+write_size=`src/pwritev2 -d -V 1 -b 128M $testdir/partial | grep ret | cut -f 4 -d " "`
+if [ $write_size -gt 0 ]; then
+ echo "pwritev2() wrote more than zero bytes."
+else
+ echo "pwritev2() returned $write_size"
+fi
+
+# read the partial file to check if data written is correct
+$XFS_IO_PROG -d -c "pread 0 4k -v" $testdir/partial | _filter_xfs_io_unique
+
+# success, all done
+status=0
+exit
diff --git a/tests/generic/462.out b/tests/generic/462.out
new file mode 100644
index 00000000..117a64b0
--- /dev/null
+++ b/tests/generic/462.out
@@ -0,0 +1,9 @@
+QA output created by 462
+Reformat with 320M size
+Almost fill the filesystem
+File partial size is greater than zero.
+pwritev2() wrote more than zero bytes.
+00000000: cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd ................
+*
+read 4096/4096 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 6352064a..f6da2805 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -464,3 +464,4 @@
459 auto dangerous
460 auto quick rw
461 auto quick rw
+462 auto quick rw
--
2.14.1
^ permalink raw reply related [flat|nested] 9+ messages in thread* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-27 19:10 [PATCH 1/2] generic/461: Test RWF_NOWAIT Goldwyn Rodrigues
2017-09-27 19:10 ` [PATCH 2/2] generic/462: Partial direct write test Goldwyn Rodrigues
@ 2017-09-27 21:34 ` Dave Chinner
2017-09-27 21:39 ` Goldwyn Rodrigues
1 sibling, 1 reply; 9+ messages in thread
From: Dave Chinner @ 2017-09-27 21:34 UTC (permalink / raw)
To: Goldwyn Rodrigues; +Cc: fstests, Goldwyn Rodrigues
On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>
> Tests the RWF_NOWAIT flag so the I/O returns immediately on
> a new file, without any block allocations.
>
> A new program which includes the pwritev2() call is used. This allows
> passing flags for the I/O to be performed.
Rather than write a one-off test program for this that effectively
replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
flag support to xfs_io.
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-27 21:34 ` [PATCH 1/2] generic/461: Test RWF_NOWAIT Dave Chinner
@ 2017-09-27 21:39 ` Goldwyn Rodrigues
2017-09-27 21:51 ` Dave Chinner
0 siblings, 1 reply; 9+ messages in thread
From: Goldwyn Rodrigues @ 2017-09-27 21:39 UTC (permalink / raw)
To: Dave Chinner; +Cc: fstests, Goldwyn Rodrigues
On 09/27/2017 04:34 PM, Dave Chinner wrote:
> On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>
>> Tests the RWF_NOWAIT flag so the I/O returns immediately on
>> a new file, without any block allocations.
>>
>> A new program which includes the pwritev2() call is used. This allows
>> passing flags for the I/O to be performed.
>
> Rather than write a one-off test program for this that effectively
> replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
> flag support to xfs_io.
>
This one off program is required because xfs_io does not support partial
writes. It tries to do that within the loop and does not return the
number of bytes written. This is required for test generic/462.
--
Goldwyn
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-27 21:39 ` Goldwyn Rodrigues
@ 2017-09-27 21:51 ` Dave Chinner
2017-09-27 22:24 ` Goldwyn Rodrigues
0 siblings, 1 reply; 9+ messages in thread
From: Dave Chinner @ 2017-09-27 21:51 UTC (permalink / raw)
To: Goldwyn Rodrigues; +Cc: fstests, Goldwyn Rodrigues
On Wed, Sep 27, 2017 at 04:39:20PM -0500, Goldwyn Rodrigues wrote:
>
>
> On 09/27/2017 04:34 PM, Dave Chinner wrote:
> > On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
> >> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> >>
> >> Tests the RWF_NOWAIT flag so the I/O returns immediately on
> >> a new file, without any block allocations.
> >>
> >> A new program which includes the pwritev2() call is used. This allows
> >> passing flags for the I/O to be performed.
> >
> > Rather than write a one-off test program for this that effectively
> > replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
> > flag support to xfs_io.
> >
>
> This one off program is required because xfs_io does not support partial
> writes. It tries to do that within the loop and does not return the
> number of bytes written. This is required for test generic/462.
Then please also extend xfs_io to support partial reads and writes
in the manner you need.
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-27 21:51 ` Dave Chinner
@ 2017-09-27 22:24 ` Goldwyn Rodrigues
2017-09-28 1:51 ` Dave Chinner
0 siblings, 1 reply; 9+ messages in thread
From: Goldwyn Rodrigues @ 2017-09-27 22:24 UTC (permalink / raw)
To: Dave Chinner; +Cc: fstests
On 09/27/2017 04:51 PM, Dave Chinner wrote:
> On Wed, Sep 27, 2017 at 04:39:20PM -0500, Goldwyn Rodrigues wrote:
>>
>>
>> On 09/27/2017 04:34 PM, Dave Chinner wrote:
>>> On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
>>>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>>>
>>>> Tests the RWF_NOWAIT flag so the I/O returns immediately on
>>>> a new file, without any block allocations.
>>>>
>>>> A new program which includes the pwritev2() call is used. This allows
>>>> passing flags for the I/O to be performed.
>>>
>>> Rather than write a one-off test program for this that effectively
>>> replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
>>> flag support to xfs_io.
>>>
>>
>> This one off program is required because xfs_io does not support partial
>> writes. It tries to do that within the loop and does not return the
>> number of bytes written. This is required for test generic/462.
>
> Then please also extend xfs_io to support partial reads and writes
> in the manner you need.
>
That will break existing tests which rely on nothing but the error
returned in case of partial writes.
I don't mind fixing xfs_io. It could use some love especially in the
documentation.
--
Goldwyn
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-27 22:24 ` Goldwyn Rodrigues
@ 2017-09-28 1:51 ` Dave Chinner
2017-09-28 2:09 ` Goldwyn Rodrigues
0 siblings, 1 reply; 9+ messages in thread
From: Dave Chinner @ 2017-09-28 1:51 UTC (permalink / raw)
To: Goldwyn Rodrigues; +Cc: fstests
On Wed, Sep 27, 2017 at 05:24:49PM -0500, Goldwyn Rodrigues wrote:
>
>
> On 09/27/2017 04:51 PM, Dave Chinner wrote:
> > On Wed, Sep 27, 2017 at 04:39:20PM -0500, Goldwyn Rodrigues wrote:
> >>
> >>
> >> On 09/27/2017 04:34 PM, Dave Chinner wrote:
> >>> On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
> >>>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> >>>>
> >>>> Tests the RWF_NOWAIT flag so the I/O returns immediately on
> >>>> a new file, without any block allocations.
> >>>>
> >>>> A new program which includes the pwritev2() call is used. This allows
> >>>> passing flags for the I/O to be performed.
> >>>
> >>> Rather than write a one-off test program for this that effectively
> >>> replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
> >>> flag support to xfs_io.
> >>>
> >>
> >> This one off program is required because xfs_io does not support partial
> >> writes. It tries to do that within the loop and does not return the
> >> number of bytes written. This is required for test generic/462.
> >
> > Then please also extend xfs_io to support partial reads and writes
> > in the manner you need.
> >
>
> That will break existing tests which rely on nothing but the error
> returned in case of partial writes.
So trigger necessary partial write behaviour only when the CLI
option to use RWF_NOWAIT is present....
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-28 1:51 ` Dave Chinner
@ 2017-09-28 2:09 ` Goldwyn Rodrigues
2017-09-28 6:02 ` Dave Chinner
0 siblings, 1 reply; 9+ messages in thread
From: Goldwyn Rodrigues @ 2017-09-28 2:09 UTC (permalink / raw)
To: Dave Chinner; +Cc: fstests
On 09/27/2017 08:51 PM, Dave Chinner wrote:
> On Wed, Sep 27, 2017 at 05:24:49PM -0500, Goldwyn Rodrigues wrote:
>>
>>
>> On 09/27/2017 04:51 PM, Dave Chinner wrote:
>>> On Wed, Sep 27, 2017 at 04:39:20PM -0500, Goldwyn Rodrigues wrote:
>>>>
>>>>
>>>> On 09/27/2017 04:34 PM, Dave Chinner wrote:
>>>>> On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
>>>>>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>>>>>
>>>>>> Tests the RWF_NOWAIT flag so the I/O returns immediately on
>>>>>> a new file, without any block allocations.
>>>>>>
>>>>>> A new program which includes the pwritev2() call is used. This allows
>>>>>> passing flags for the I/O to be performed.
>>>>>
>>>>> Rather than write a one-off test program for this that effectively
>>>>> replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
>>>>> flag support to xfs_io.
>>>>>
>>>>
>>>> This one off program is required because xfs_io does not support partial
>>>> writes. It tries to do that within the loop and does not return the
>>>> number of bytes written. This is required for test generic/462.
>>>
>>> Then please also extend xfs_io to support partial reads and writes
>>> in the manner you need.
>>>
>>
>> That will break existing tests which rely on nothing but the error
>> returned in case of partial writes.
>
> So trigger necessary partial write behaviour only when the CLI
> option to use RWF_NOWAIT is present....
>
Partial write test case is not related to RWF_NOWAIT test case. These
are two separate test cases.
Anyways, I am working on implementing this. Would you prefer pwritev2 be
a separate subcommand calling pwritev2() or should I transform pwritev()
to pwritev2()? The system call is relatively new and there are
overlapping features such as RWF_DSYNC and RWF_SYNC. I am assuming the
former.
--
Goldwyn
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] generic/461: Test RWF_NOWAIT
2017-09-28 2:09 ` Goldwyn Rodrigues
@ 2017-09-28 6:02 ` Dave Chinner
0 siblings, 0 replies; 9+ messages in thread
From: Dave Chinner @ 2017-09-28 6:02 UTC (permalink / raw)
To: Goldwyn Rodrigues; +Cc: fstests
On Wed, Sep 27, 2017 at 09:09:55PM -0500, Goldwyn Rodrigues wrote:
>
>
> On 09/27/2017 08:51 PM, Dave Chinner wrote:
> > On Wed, Sep 27, 2017 at 05:24:49PM -0500, Goldwyn Rodrigues wrote:
> >>
> >>
> >> On 09/27/2017 04:51 PM, Dave Chinner wrote:
> >>> On Wed, Sep 27, 2017 at 04:39:20PM -0500, Goldwyn Rodrigues wrote:
> >>>>
> >>>>
> >>>> On 09/27/2017 04:34 PM, Dave Chinner wrote:
> >>>>> On Wed, Sep 27, 2017 at 02:10:02PM -0500, Goldwyn Rodrigues wrote:
> >>>>>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> >>>>>>
> >>>>>> Tests the RWF_NOWAIT flag so the I/O returns immediately on
> >>>>>> a new file, without any block allocations.
> >>>>>>
> >>>>>> A new program which includes the pwritev2() call is used. This allows
> >>>>>> passing flags for the I/O to be performed.
> >>>>>
> >>>>> Rather than write a one-off test program for this that effectively
> >>>>> replicates xfs_io pread/pwrite functionality, please add RWF_NOWAIT
> >>>>> flag support to xfs_io.
> >>>>>
> >>>>
> >>>> This one off program is required because xfs_io does not support partial
> >>>> writes. It tries to do that within the loop and does not return the
> >>>> number of bytes written. This is required for test generic/462.
> >>>
> >>> Then please also extend xfs_io to support partial reads and writes
> >>> in the manner you need.
> >>>
> >>
> >> That will break existing tests which rely on nothing but the error
> >> returned in case of partial writes.
> >
> > So trigger necessary partial write behaviour only when the CLI
> > option to use RWF_NOWAIT is present....
> >
>
> Partial write test case is not related to RWF_NOWAIT test case. These
> are two separate test cases.
>
> Anyways, I am working on implementing this. Would you prefer pwritev2 be
> a separate subcommand calling pwritev2() or should I transform pwritev()
> to pwritev2()? The system call is relatively new and there are
> overlapping features such as RWF_DSYNC and RWF_SYNC. I am assuming the
> former.
If pwritev2 exists at build time, build in support for it. If it
returns ENOSYS or it is not present at build time, fall back to
pwritev()...
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2017-09-28 6:04 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-09-27 19:10 [PATCH 1/2] generic/461: Test RWF_NOWAIT Goldwyn Rodrigues
2017-09-27 19:10 ` [PATCH 2/2] generic/462: Partial direct write test Goldwyn Rodrigues
2017-09-27 21:34 ` [PATCH 1/2] generic/461: Test RWF_NOWAIT Dave Chinner
2017-09-27 21:39 ` Goldwyn Rodrigues
2017-09-27 21:51 ` Dave Chinner
2017-09-27 22:24 ` Goldwyn Rodrigues
2017-09-28 1:51 ` Dave Chinner
2017-09-28 2:09 ` Goldwyn Rodrigues
2017-09-28 6:02 ` Dave Chinner
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox