Re: [PATCH v2 3/4] btrfs: zoned: drop stranded dirty metadata buffers at unmount

On 7/23/26 7:02 PM, Boris Burkov wrote:
> Immediate fix looks fine, but the underlying issue is a little hacky.
> Can we detect that we have staled them from free_log_tree() and release
> the dirty there? Am I correct in understanding that this means we can
> overestimate dirty? By an arbitrary amount if I do lots of unsynced
> tree log updates?
I did try that, but it deadlocked pretty reliably :/