[snip]
> 
> This looks like the same problem I reported earlier this month, and
> also filed a bug for at
> https://bugzilla.kernel.org/show_bug.cgi?id=202717
> 
> In my case I did a scrub and check before clearing space cache v1. No
> problems reported. And then clearing space cache v1 crashed. And then
> check reports corruption.
> 
> Bug 1, for sure btrfs check clear cache crashing is a bug
> Bug 2, btrfs check appears to do non-COW overwrite of the extent tree,
> which might be fine as long as it doesn't crash but it seems risky
> considering how fragile the extent tree is anyway

In fact I think this is a bigger problem.

This breaks the basic metadata CoW, so I'm going to look into this bug.

We have enough tool to investigate this problem, but I'm afraid it may
need certain images to trigger.

If anyone is trying to try clear space cache, please try to take a
binary dump and provide it (if small enough).

Thanks,
Qu

> 
> Clear cache right now is not fail safe near as I can tell. It can make
> an error free file system corrupt.
> 
> If there's some problem already, before clearing space cache, that
> means more bugs:
> 
> Bug 3, btrfs check doesn't find the problem, reports no errors
> Bug 4, btrfs kernel code doesn't find the problem during scrub,
> reports no errors
> 
> 

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to