search

kdave / btrfs-progs

Development of userspace BTRFS tools
GNU General Public License v2.0
557 stars 242 forks source link

'btrfs device stats' output 0 as corruption_errs, but 2 is expected(pre-release 5.15-rc1) #417

Closed wangyugui closed 2 years ago

wangyugui commented 2 years ago

xfstests(btrfs/215) is broken with progs pre-release 5.15-rc1

It seems not a feature change, just a bug.

'btrfs device stats' output 0 as corruption_errs, but 2 is expected.

wangyugui commented 2 years ago

mkfs.btrfs (pre-release 5.15-rc1, 5.14.2) & 'btrfs device stat'(pre-release 5.15-rc1, 5.14.2) : xfstests(btrfs/215) FAIL mkfs.btrfs -O no-holes -R free-space-tree(5.10) & 'btrfs device stat'(pre-release 5.15-rc1, 5.14.2) : xfstests(btrfs/215) PASS mkfs.btrfs -O no-holes -R free-space-tree(5.10) & 'btrfs device stat'( 5.10) : xfstests(btrfs/215) PASS

'btrfs device stat'(pre-release 5.15-rc1, 5.14.2) correctly output the value that get from kernel.

Is this a feature change OR bug of 'mkfs.btrfs (pre-release 5.15-rc1, 5.14.2)'?

wangyugui commented 2 years ago

This is because mkfs.btrfs '-m DUP' default for single SSD too since this release.

Add '-m single' to xfstests(btrfs/215) test script to avoid this failure.

adam900710 commented 2 years ago

No, it's not that simple.

If you go misc-next branch, that would cause a different problem.

In fact btrfs/215 will always fail due to nospace_cache mount option will prevent us to mount an fs with v2 cache.

I have sent a fix for the bug anyway. Thanks for the report.

kdave commented 2 years ago

Right, the defaults in mkfs could change test results like you observe, it's expected but good to know in which cases in particular.