Re: [PATCH v2] generic/793: limit number of overwrites

On 24/04/2026 13:48, Johannes Thumshirn wrote:
> Testcase generic/793 overwrites a 1GB file "number of sequential zones"
> times in order to trigger a bug in zoned BTRFS' garbage collection. But on
> scratch devices with a huge number of sequential zones (like SMR drives)
> this can cause very long runtimes.
> 
> As the testcase also limits the filesystem size to 16GB, constraint the
> number of overwrite loops to the number of sequential zones in the
> filesystem. This still guarantees that the file will be overwritten often
> enough to trigger the bug while not exploding the overall test runtime.
> 
> Reported-by: Shin'ichiro Kawasaki <[email protected]>
> Signed-off-by: Johannes Thumshirn <[email protected]>
> ---
>  tests/generic/793 | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/tests/generic/793 b/tests/generic/793
> index a59e84694309..fcfbc4983d7c 100755
> --- a/tests/generic/793
> +++ b/tests/generic/793
> @@ -11,6 +11,7 @@
>  _begin_fstest auto quick zone
>  
>  . ./common/filter
> +. ./common/zoned
>  
>  _require_scratch_size $((16 * 1024 * 1024))
>  
> @@ -21,7 +22,6 @@ else
>  fi
>  
>  _require_zoned_device "$zoned_dev"
> -_require_command "$BLKZONE_PROG" blkzone

_zone_capacity uses BLKZONE_PROG, so this needs to stay.
With that fixed, this looks good.

Reviewed-by: Hans Holmberg <[email protected]>

>  
>  # This test requires specific data space usage, skip if we have compression
>  # enabled.
> @@ -34,12 +34,14 @@ _fixed_by_fs_commit btrfs 258e46a6385c \
>  _fixed_by_fs_commit btrfs e2a7fd22378f \
>  	"btrfs: zoned: add zone reclaim flush state for DATA space_info"
>  
> -_scratch_mkfs_sized $((16 * 1024 * 1024 * 1024)) &>>$seqres.full
> +fssize=$((16 * 1024 * 1024 * 1024))
> +_scratch_mkfs_sized ${fssize} &>>$seqres.full
>  _scratch_mount
>  
> -# Overwrite the data "number of seq zones" times, this is arbitrarily chosen
> +# Overwrite the data "number of zones in fs" times, this is arbitrarily chosen
>  # but triggers the original bug reliably
> -loops=$($BLKZONE_PROG report $zoned_dev | grep -c "SEQ_WRITE_REQUIRED")
> +zsize=$(_zone_capacity 0 $zoned_dev)
> +loops=$(( fssize/zsize))
>  
>  for (( i = 0; i < $loops; i++ )); do
>  	dd if=/dev/zero of=$SCRATCH_MNT/test bs=1M count=1024 status=none