Re: [PATCH RFC 1/9] btrfs: zoned: do not select metadata BG as finish target

[Naohiro Aota <Naohiro.Aota@wdc.com>]

On Fri Jun 27, 2025 at 8:34 PM JST, Christoph Hellwig wrote:
> On Fri, Jun 27, 2025 at 11:19:06AM +0200, Johannes Thumshirn wrote:
>> From: Naohiro Aota <naohiro.aota@wdc.com>
>> 
>> We call btrfs_zone_finish_one_bg() to zone finish one block group and make
>> a room to activate another block group. Currently, we can choose a metadata
>> block group as a target. But, as we reserve an active metadata block group,
>> we no longer want to select a metadata block group. So, skip it in the
>> loop.
>
> Q: why do you finish a currently open zone to start with?  If you add
> an extra zones worth of over provisioning, you have enough slack to
> always be able to fill to the advertized capacity, and never need to
> finish an open zone before it is fully filled.  Which simplifies the
> implementation and reduces P/E cycles.

Basically, this is called when data extent allocation cannot activate a
new zone, so the number of active zones == max active zones. In this
case, it first call btrfs_zone_finish_one_bg() to try to finish a zone
with minimum free space. If it succeeds, we can allocate new block group
and allocate an extent from there. Or, it retries the allocation with a
smaller size. So, it just prefers zone finishing than filling with a
fragmented allocation.

Another usage is when it writes to metadata. While we reserve zones for
metadata, this can be an escape hatch to finish some zones and make a
room for the new writing.

>
>> Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
>
> You'll also need to add your signoff here when sending the patch on.