| Message ID | 20220909214810.761928-2-mcgrof@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs-progs: minor mkfs fs size error message enhancements | expand |
> With this now the user sees: > > ERROR: Size 512.00MiB is too small to make a usable filesystem. > ERROR: The minimum size for a zoned btrfs filesystem requires > ERROR: 5 dedicated zones. This device's zone size is 128.00MiB so > ERROR: the minimum size for a btrfs filesystem for this zoned > ERROR: storage device (/dev/nvme5n1) is 640.00MiB > > The following fstests fail because of this issue, and at first glance > it was not clear why: > > * brfs/132 > * generic/226 > * generic/416 > * generic/650 > > Now it's clear what the issue is and what needs to get done to > fix those respective tests. But note, that if working on a sequential > zone then parallel writes are expected to fail, so tests which require > that and use parallel writes must also check for sequential zones > and bail if one is to be used. Each of these tests needs to be adapted > a bit for zoned devices to work properly. > > Signed-off-by: Luis Chamberlain <mcgrof@kernel.org> > --- Looks good, Reviewed-by: Pankaj Raghav <p.raghav@samsung.com>
diff --git a/mkfs/main.c b/mkfs/main.c index eaecdc061dfe..ebf462587bd5 100644 --- a/mkfs/main.c +++ b/mkfs/main.c @@ -1401,10 +1401,14 @@ int BOX_MAIN(mkfs)(int argc, char **argv) * 1 zone for a data block group */ if (zoned && block_count && block_count < 5 * zone_size(file)) { - error("size %llu is too small to make a usable filesystem", - block_count); - error("minimum size for a zoned btrfs filesystem is %llu", - min_dev_size); + error("Size %s is too small to make a usable filesystem.", + pretty_size_mode(block_count, UNITS_DEFAULT)); + error("The minimum size for a zoned btrfs filesystem requires"); + error("5 dedicated zones. This device's zone size is %s so", + pretty_size_mode(zone_size(file), UNITS_DEFAULT)); + error("the minimum size for a btrfs filesystem for this zoned"); + error("storage device (%s) is %s", + file, pretty_size_mode(5 * zone_size(file), UNITS_DEFAULT)); goto error; }
The minimum size for a filesystem on a zone storage device actually depends on the size of a devices' zone size, given 5 dedicated zones are required, regardless of their fulll size of the drive: * 2 zones for the primary superblock * 1 zone for the system block group * 1 zone for a metadata block group * 1 zone for a data block group So the minimum size for a btrfs filesystem for a zone storage device is completely device specific. The check for this however complains about the minimum size required for btrfs in general, not for the above consideration. Fix the error message so that users understand the exact reason why mkfs might fail if you want to create a filesystem which cannot possibly fit into less than 5 zones. Also clarify how the minimum size then is completely device specific. Without this a user can end up scratching their heads if they tried to create say a 512 MiB filesystem on a 100 GiB NVMe ZNS drive with 128 MiB zone size. In this case 5 * 128 = 640 MiB is the minimum required. Without this patch a user will end up with the following error message: ERROR: size 268435456 is too small to make a usable filesystem ERROR: minimum size for a zoned btrfs filesystem is 47185920 Clearly 268435456 (256 MiB) > 47185920 (45 MiB). Fix the error message so that the user knows what to do if they want the mimimum filesystem for btrfs zone storage device. With this now the user sees: ERROR: Size 512.00MiB is too small to make a usable filesystem. ERROR: The minimum size for a zoned btrfs filesystem requires ERROR: 5 dedicated zones. This device's zone size is 128.00MiB so ERROR: the minimum size for a btrfs filesystem for this zoned ERROR: storage device (/dev/nvme5n1) is 640.00MiB The following fstests fail because of this issue, and at first glance it was not clear why: * brfs/132 * generic/226 * generic/416 * generic/650 Now it's clear what the issue is and what needs to get done to fix those respective tests. But note, that if working on a sequential zone then parallel writes are expected to fail, so tests which require that and use parallel writes must also check for sequential zones and bail if one is to be used. Each of these tests needs to be adapted a bit for zoned devices to work properly. Signed-off-by: Luis Chamberlain <mcgrof@kernel.org> --- mkfs/main.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)