generic: add a less thorough testing mode for fsync-err program
PATCH 1/1 generic: add a less thorough testing mode for fsync-err program
From: Jeff Layton <[email protected]>
Currently we just have this test run on a whitelist of filesystems, but it would be best to be able to run it on all of them. The problem is that a lot of filesystems basically shut down once they hit metadata errors.
Allow the fsync-err testcase to operate in two different modes. One mode just does basic testing to ensure that we get an error back on all fd's when we fsync. The other does a more thorough test to ensure that we get back 0 on subsequent fsyncs when there hasn't been any write activity.
For now, we just opt-in to the more thorough testing on certain
filesystems: xfs, ext3 and ext4 on the generic test. All other filesystems will run in simple mode.
--- common/rc | 9 +++++++++ src/fsync-err.c | 51 +++++++++++++++++++++++++++++++++------------------ tests/generic/441 | 22 +++++++++++++++------- 3 files changed, 57 insertions(+), 25 deletions(-)
common/rc +9 -0
@@ -1738,6 +1738,15 @@ _require_test() touch ${RESULT_DIR}/require_test } +_has_logdev() +{ + ret=0 + [ -z "$SCRATCH_LOGDEV" -o ! -b "$SCRATCH_LOGDEV" ] && + [ "$USE_EXTERNAL" != yes ] && ret=1 + + return $ret +} + # this test needs a logdev # _require_logdev()
src/fsync-err.c +33 -18
88 lines not on this page · raw message
tests/generic/441 +15 -7
@@ -45,15 +45,23 @@ _cleanup() . ./common/dmerror # real QA test starts here -_supported_fs ext2 ext3 ext4 xfs _supported_os Linux _require_scratch -# Generally, we want to avoid journal errors in this test. Ensure that -# journalled fs' have a logdev. -if [ "$FSTYP" != "ext2" ]; then - _require_logdev -fi +# Generally, we want to avoid journal errors on the extended testcase. Only +# set the -r flag if we have a logdev +sflag='-s' +case $FSTYP in + btrfs) + _notrun "btrfs has a specialized test for this" + ;; + ext3|ext4|xfs) + # Do the more thorough test if we have a logdev + _has_logdev && sflag='' + ;; + *) + ;; +esac _require_dm_target error _require_test_program fsync-err
@@ -70,7 +78,7 @@ _require_fs_space $SCRATCH_MNT 65536 testfile=$SCRATCH_MNT/fsync-err-test -$here/src/fsync-err -d $here/src/dmerror $testfile +$here/src/fsync-err $sflag -d $here/src/dmerror $testfile # success, all done _dmerror_load_working_table
Signature · 1 line
2.13.0
Quoted 29 lines · On Thu, Jun 29, 2017 at 09:29:47AM -0400, [email protected] wrote:
From: Jeff Layton <[email protected]>
Currently we just have this test run on a whitelist of filesystems, but it would be best to be able to run it on all of them. The problem is that a lot of filesystems basically shut down once they hit metadata errors.
Allow the fsync-err testcase to operate in two different modes. One mode just does basic testing to ensure that we get an error back on all fd's when we fsync. The other does a more thorough test to ensure that we get back 0 on subsequent fsyncs when there hasn't been any write activity.
For now, we just opt-in to the more thorough testing on certain
filesystems: xfs, ext3 and ext4 on the generic test. All other filesystems will run in simple mode.Signed-off-by Jeff Layton <[email protected]>--- common/rc | 9 +++++++++ src/fsync-err.c | 51 +++++++++++++++++++++++++++++++++------------------ tests/generic/441 | 22 +++++++++++++++------- 3 files changed, 57 insertions(+), 25 deletions(-)common/rc +0
-0
common/rc _require_test() · line 1745
Quoted 8 lines
common/rc +5
-0@@ -1738,6 +1738,15 @@ _require_test() touch ${RESULT_DIR}/require_test } +_has_logdev() +{ + ret=0 + [ -z "$SCRATCH_LOGDEV" -o ! -b "$SCRATCH_LOGDEV" ] && + [ "$USE_EXTERNAL" != yes ] && ret=1
This doesn't seem right. _has_logdev = ($SCRATCH_LOGDEV is block device) && ($USE_EXTERNAL set to yes), so we should set ret=1 if any of these conditions is not met. But this code returns 0 (true) if I only have SCRATCH_LOGDEV set but not USE_EXTERNAL.
And once we have _has_logdev, I think we can refactor _require_logdev to use _has_logdev too.
Quoted 115 lines
common/rc +4
-0+ + return $ret +} + # this test needs a logdev # _require_logdev()src/fsync-err.c +33
-18@@ -13,6 +13,7 @@#include <string.h> #include <unistd.h> #include <getopt.h> +#include <stdbool.h> /* * btrfs has a fixed stripewidth of 64k, so we need to write enough data to @@ -25,7 +26,7 @@static void usage () { -printf ("Usage: fsync-err [ -b bufsize ] [ -n num_fds ] -d dmerror path <filename>\n" ); +printf ("Usage: fsync-err [ -b bufsize ] [ -n num_fds ] [ -s ] -d dmerror path <filename>\n" ); }int main (int argc,char **argv)@@ -35,8 +36,9 @@ int main(int argc, char **argv)char *dmerror_path =NULL ;char *cmdbuf;size_t cmdsize, bufsize =DEFAULT_BUFSIZE ; +bool simple_mode = false; -while ((i =getopt (argc, argv,"b:d:n:" )) !=-1 ) { +while ((i =getopt (argc, argv,"b:d:n:s" )) !=-1 ) {switch (i) {case 'b' : bufsize =strtol (optarg, &buf,0 );@@ -55,6 +57,15 @@ int main(int argc, char **argv)return 1 ; }break ; +case 's' : +/* +* Many filesystems will continue to throw errors after +* fsync has already advanced to the current error, +* due to metadata writeback failures or other +* issues. Allow those fs' to opt out of more thorough +* testing. +*/ + simple_mode = true; } }@@ -154,16 +165,18 @@ int main(int argc, char **argv) } } -for (i =0 ; i < numfds; ++i) { - ret =fsync (fd[i]); -if (ret <0 ) { -/* -* We did a failed write and fsync on each fd before. -* Now the error should be clear since we've not done -* any writes since then. -*/ -printf ("Third fsync on fd[%d] failed: %m\n" , i); -return 1 ; +if (!simple_mode) { +for (i =0 ; i < numfds; ++i) { + ret =fsync (fd[i]); +if (ret <0 ) { +/* +* We did a failed write and fsync on each fd before. +* Now the error should be clear since we've not done +* any writes since then. +*/ +printf ("Third fsync on fd[%d] failed: %m\n" , i); +return 1 ; + } } }@@ -185,12 +198,14 @@ int main(int argc, char **argv)return 1 ; } -for (i =0 ; i < numfds; ++i) { - ret =fsync (fd[i]); -if (ret <0 ) { -/* The error should still be clear */ -printf ("fsync after healing device on fd[%d] failed: %m\n" , i); -return 1 ; +if (!simple_mode) { +for (i =0 ; i < numfds; ++i) { + ret =fsync (fd[i]); +if (ret <0 ) { +/* The error should still be clear */ +printf ("fsync after healing device on fd[%d] failed: %m\n" , i); +return 1 ; + } } }tests/generic/441 +0
-1@@ -45,15 +45,23 @@ _cleanup() . ./common/dmerror # real QA test starts here -_supported_fs ext2 ext3 ext4 xfs _supported_os Linux
tests/generic/441 _cleanup() · line 53
Quoted 10 lines
tests/generic/441 +3
-5_require_scratch -# Generally, we want to avoid journal errors in this test. Ensure that -# journalled fs' have a logdev. -if [ "$FSTYP" != "ext2" ]; then - _require_logdev -fi +# Generally, we want to avoid journal errors on the extended testcase. Only +# set the -r flag if we have a logdev +sflag='-s'
This part seems confusing, comments said "set the -r flag" but '-s' is actually used.
Quoted 15 lines
tests/generic/441 +11
-0+case $FSTYP in + btrfs) + _notrun "btrfs has a specialized test for this" + ;; + ext3|ext4|xfs) + # Do the more thorough test if we have a logdev + _has_logdev && sflag='' + ;; + *) + ;; +esac _require_dm_target error _require_test_program fsync-err
tests/generic/441 _require_fs_space $SCRATCH_MNT 65536 · line 81
Quoted 5 lines
tests/generic/441 +1
-1@@ -70,7 +78,7 @@ _require_fs_space $SCRATCH_MNT 65536 testfile=$SCRATCH_MNT/fsync-err-test -$here/src/fsync-err -d $here/src/dmerror $testfile +$here/src/fsync-err $sflag -d $here/src/dmerror $testfile
Better to append this command to $seqres.full too for debug purpose, so we know if we're running with or without '-s' option.
Thanks,
Eryu
Quoted 6 lines
tests/generic/441 +0
-0# success, all done _dmerror_load_working_tableSignature · 1 line
2.13.0
PATCH v2 1/1 generic: add a less thorough testing mode for fsync-err program
In reply to [email protected]From: Jeff Layton <[email protected]>
Currently we just have this test run on a whitelist of filesystems, but it would be best to be able to run it on all of them. The problem is that a lot of filesystems basically shut down once they hit metadata errors.
Allow the fsync-err testcase to operate in two different modes. One mode just does basic testing to ensure that we get an error back on all fd's when we fsync. The other does a more thorough test to ensure that we get back 0 on subsequent fsyncs when there hasn't been any write activity.
For now, we just opt-in to the more thorough testing on certain
filesystems: xfs, ext3 and ext4 on the generic test. All other filesystems will run in simple mode.
--- common/rc | 9 +++++++++ src/fsync-err.c | 51 +++++++++++++++++++++++++++++++++------------------ tests/generic/441 | 23 ++++++++++++++++------- 3 files changed, 58 insertions(+), 25 deletions(-)
v2: fix _has_logdev test, append command to seqres.full, fix comments
common/rc +9 -0
@@ -1738,6 +1738,15 @@ _require_test() touch ${RESULT_DIR}/require_test } +_has_logdev() +{ + ret=0 + [ -z "$SCRATCH_LOGDEV" -o ! -b "$SCRATCH_LOGDEV" ] && ret=1 + [ "$USE_EXTERNAL" != yes ] && ret=1 + + return $ret +} + # this test needs a logdev # _require_logdev()
src/fsync-err.c +33 -18
88 lines not on this page · raw message
tests/generic/441 +16 -7
@@ -45,15 +45,23 @@ _cleanup() . ./common/dmerror # real QA test starts here -_supported_fs ext2 ext3 ext4 xfs _supported_os Linux _require_scratch -# Generally, we want to avoid journal errors in this test. Ensure that -# journalled fs' have a logdev. -if [ "$FSTYP" != "ext2" ]; then - _require_logdev -fi +# Generally, we want to avoid journal errors on the extended testcase. Only +# unset the -s flag if we have a logdev +sflag='-s' +case $FSTYP in + btrfs) + _notrun "btrfs has a specialized test for this" + ;; + ext3|ext4|xfs) + # Do the more thorough test if we have a logdev + _has_logdev && sflag='' + ;; + *) + ;; +esac _require_dm_target error _require_test_program fsync-err
@@ -70,7 +78,8 @@ _require_fs_space $SCRATCH_MNT 65536 testfile=$SCRATCH_MNT/fsync-err-test -$here/src/fsync-err -d $here/src/dmerror $testfile +echo "$here/src/fsync-err $sflag -d $here/src/dmerror $testfile" >> $seqres.full +$here/src/fsync-err $sflag -d $here/src/dmerror $testfile # success, all done _dmerror_load_working_table
Signature · 1 line
2.13.0