Re: [PATCH v3] f2fs: introduce flush_policy sysfs entry

[Chao Yu <chao@kernel.org>]

On 8/12/25 15:59, Christoph Hellwig wrote:
> On Tue, Aug 12, 2025 at 03:53:54PM +0800, Chao Yu wrote:
>>> What did you use before?  At least for older qemu the default was
>>> buffered I/O, which can lead to very expensive fua or flush calls.
>>
>> Previously, I didn't use any cache= option, as manual described, it
>> should equal to cache=wrteback.
> 
> Modern qemu actually split the cache option.  You absolute want
> cache.direct=on.  If you don't do simulated power fail testing by

Yes,

> killing qemu (or run real workloads for the matter, but who does that
> :)) it might make sense to just ignore the flushes with cache.no-flush=on

Yes, I don't care whether data can be persisted to host devices or not,
nor killing qemu for test, so cache.no-flush=on looks good to me as well.

> as well, which is what I do for my test VMs on the laptop.

Thanks for sharing this, it helps. :)

Thanks,