On Tue, Apr 08, 2025 at 11:10:43PM +0800, Anand Jain wrote:
>
>
> On 8/4/25 00:58, Darrick J. Wong wrote:
> > On Mon, Apr 07, 2025 at 11:48:16AM +0800, Anand Jain wrote:
> > > Ensure logs don’t write to a `.full` file when `_set_fs_sysfs_attr()`
> > > is called during setup (before a testcase) in XFS due to unset seqres.
> > >
> > > Signed-off-by: Anand Jain <anand.jain@xxxxxxxxxx>
> > > ---
> > > common/rc | 11 +++++++++--
> > > 1 file changed, 9 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/common/rc b/common/rc
> > > index e89eee5de840..ca1d13ca1f0b 100644
> > > --- a/common/rc
> > > +++ b/common/rc
> > > @@ -5201,6 +5201,13 @@ _set_fs_sysfs_attr()
> > > local attr=$1
> > > shift
> > > local content="$*"
> > > + local logfile="/dev/null"
> >
> > If we're not actually running a test, should the error output go to
> > $check.full instead?
> >
>
> Ah. I didn't know. It seems like no one is using check.full.
Oh, they do, but it's subtle:
$ git grep seqres=
check:852: seqres="$check"
check:1120: seqres="$check"
--D
> -----
> $ find . -type f -exec grep check\.full {} \; -print
> rm -f $check.full
> ./check
> -----
>
> Thanks, Anand
>
>
> > --D
> >
> > > +
> > > + # This function may be called outside a testcase during setup,
> > > + # so seqres might not be set.
> > > + if [[ -v seqres ]]; then
> > > + logfile="$seqres.full"
> > > + fi
> > > if [ ! -b "$dev" -o -z "$attr" -o -z "$content" ];then
> > > _fail "Usage: _set_fs_sysfs_attr <mounted_device> <attr> <content>"
> > > @@ -5208,8 +5215,8 @@ _set_fs_sysfs_attr()
> > > local dname=$(_fs_sysfs_dname $dev)
> > > - echo "echo '$content' 2>&1 > /sys/fs/${FSTYP}/${dname}/${attr}" >> $seqres.full
> > > - echo "$content" 2>&1 > /sys/fs/${FSTYP}/${dname}/${attr} | tee -a $seqres.full
> > > + echo "echo '$content' 2>&1 > /sys/fs/${FSTYP}/${dname}/${attr}" >> $logfile
> > > + echo "$content" 2>&1 > /sys/fs/${FSTYP}/${dname}/${attr} | tee -a $logfile
> > > }
> > > # Print the content of /sys/fs/$FSTYP/$DEV/$ATTR
> > > --
> > > 2.47.0
> > >
> > >
>
>
