Re: [PATCH 2/4] xfs: rework min log size helper

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Jun 18, 2019 at 02:07:15PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> 
> The recent _scratch_find_xfs_min_logblocks helper has a major thinko in
> it -- it relies on feeding a too-small size to _scratch_do_mkfs so that
> mkfs will tell us the minimum log size.  Unfortunately, _scratch_do_mkfs
> will see that first failure and retry the mkfs without MKFS_OPTIONS,
> which means that we return the minimum log size for the default mkfs
> settings without MKFS_OPTIONS.
> 
> This is a problem if someone's running fstests with a set of
> MKFS_OPTIONS that affects the minimum log size.  To fix this, open-code
> the _scratch_do_mkfs retry behavior so that we only do the "retry
> without MKFS_OPTIONS" behavior if the mkfs failed for a reason other
> than the minimum log size check.
> 
> Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> ---
>  common/rc  |   13 ++++++++++---
>  common/xfs |   23 +++++++++++++++++++++--
>  2 files changed, 31 insertions(+), 5 deletions(-)
> 
> 
> diff --git a/common/rc b/common/rc
> index 25203bb4..a38b7f02 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -438,6 +438,14 @@ _scratch_mkfs_options()
>      echo $SCRATCH_OPTIONS $MKFS_OPTIONS $* $SCRATCH_DEV
>  }
>  
> +# Format the scratch device directly.  First argument is the mkfs command.
> +# Second argument are all the parameters.  stdout goes to $tmp.mkfsstd and
> +# stderr goes to $tmp.mkfserr.
> +__scratch_do_mkfs()
> +{
> +	eval "$1 $2 $SCRATCH_DEV" 2>$tmp.mkfserr 1>$tmp.mkfsstd

I'd prefer leaving stdout and stderr to caller to handle, because ..


> +}
> +
>  # Do the actual mkfs work on SCRATCH_DEV. Firstly mkfs with both MKFS_OPTIONS
>  # and user specified mkfs options, if that fails (due to conflicts between mkfs
>  # options), do a second mkfs with only user provided mkfs options.
> @@ -456,8 +464,7 @@ _scratch_do_mkfs()
>  
>  	# save mkfs output in case conflict means we need to run again.
>  	# only the output for the mkfs that applies should be shown
> -	eval "$mkfs_cmd $MKFS_OPTIONS $extra_mkfs_options $SCRATCH_DEV" \
> -		2>$tmp.mkfserr 1>$tmp.mkfsstd

it's easier to know the $tmp.mkfserr and $tmp.mkfsstd files should be
cleaned up, otherwise it's not that clear where these files come from.

> +	__scratch_do_mkfs "$mkfs_cmd" "$MKFS_OPTIONS $extra_mkfs_options"
>  	mkfs_status=$?
>  
>  	# a mkfs failure may be caused by conflicts between $MKFS_OPTIONS and
> @@ -471,7 +478,7 @@ _scratch_do_mkfs()
>  		) >> $seqres.full
>  
>  		# running mkfs again. overwrite previous mkfs output files
> -		eval "$mkfs_cmd $extra_mkfs_options $SCRATCH_DEV" \
> +		__scratch_do_mkfs "$mkfs_cmd" "$extra_mkfs_options" \
>  			2>$tmp.mkfserr 1>$tmp.mkfsstd

With the implemention in the patch, the "2>$tmp.mkfserr 1>$tmp.mkfsstd"
part should be removed too, but with the suggested implemention we can
keep it :)

>  		mkfs_status=$?
>  	fi
> diff --git a/common/xfs b/common/xfs
> index f8dafc6c..8733e2ae 100644
> --- a/common/xfs
> +++ b/common/xfs
> @@ -87,16 +87,33 @@ _scratch_find_xfs_min_logblocks()
>  	# minimum log size.
>  	local XFS_MIN_LOG_BYTES=2097152
>  
> -	_scratch_do_mkfs "$mkfs_cmd" "cat" $* -N -l size=$XFS_MIN_LOG_BYTES \
> -		2>$tmp.mkfserr 1>$tmp.mkfsstd
> +	# Try formatting the filesystem with all the options given and the
> +	# minimum log size.  We hope either that this succeeds or that mkfs
> +	# tells us the required minimum log size for the feature set.
> +	#
> +	# We cannot use _scratch_do_mkfs because it will retry /any/ failed
> +	# mkfs with MKFS_OPTIONS removed even if the only "failure" was that
> +	# the log was too small.
> +	local extra_mkfs_options="$* -N -l size=$XFS_MIN_LOG_BYTES"
> +	__scratch_do_mkfs "$mkfs_cmd" "$MKFS_OPTIONS $extra_mkfs_options"
>  	local mkfs_status=$?
>  
> +	# If the format fails for a reason other than the log being too small,
> +	# try again without MKFS_OPTIONS because that's what _scratch_do_mkfs
> +	# will do if we pass in the log size option.
> +	if [ $mkfs_status -ne 0 ] &&
> +	   ! grep -q 'log size.*too small, minimum' $tmp.mkfserr; then
> +		__scratch_do_mkfs "$mkfs_cmd" "$extra_mkfs_options"
> +		local mkfs_status=$?

We've already declared mkfs_status as local, no need to do it again
here.

Thanks,
Eryu

> +	fi
> +
>  	# mkfs suceeded, so we must pick out the log block size to do the
>  	# unit conversion
>  	if [ $mkfs_status -eq 0 ]; then
>  		local blksz="$(grep '^log.*bsize' $tmp.mkfsstd | \
>  			sed -e 's/log.*bsize=\([0-9]*\).*$/\1/g')"
>  		echo $((XFS_MIN_LOG_BYTES / blksz))
> +		rm -f $tmp.mkfsstd $tmp.mkfserr
>  		return
>  	fi
>  
> @@ -104,6 +121,7 @@ _scratch_find_xfs_min_logblocks()
>  	if grep -q 'minimum size is' $tmp.mkfserr; then
>  		grep 'minimum size is' $tmp.mkfserr | \
>  			sed -e 's/^.*minimum size is \([0-9]*\) blocks/\1/g'
> +		rm -f $tmp.mkfsstd $tmp.mkfserr
>  		return
>  	fi
>  
> @@ -111,6 +129,7 @@ _scratch_find_xfs_min_logblocks()
>  	echo "Cannot determine minimum log size" >&2
>  	cat $tmp.mkfsstd >> $seqres.full
>  	cat $tmp.mkfserr >> $seqres.full
> +	rm -f $tmp.mkfsstd $tmp.mkfserr
>  }
>  
>  _scratch_mkfs_xfs()
> 



[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux