On 8/12/25 15:32, Christoph Hellwig wrote: > On Tue, Aug 12, 2025 at 02:28:46PM +0800, Chao Yu wrote: >> BTW, I suffered extremely long latency of checkpoint which may block every >> update operations when testing generic/299 w/ mode=lfs mount option in qemu, >> then I propose to use PREFLUSH instead of FUA to resolve this issue. >> >> "F2FS-fs (vdc): checkpoint was blocked for 24495 ms" >> >> I just realize that using cache=directsync option in qemu can avoid FUA hang >> issue, anyway, let me test more w/ this option. > > Well, for decent qemu performance you always want to use DIRECT I/O. > directsync is generally not a very good idea as it forces every write > to be synchronous and will give you very bad performance. Yeah, I think that may hurt the performance too, at least, I don't see any obvious change for time cost of generic/299 testcases, but still I need to run all my testcase to see what will happen. :) generic/299 115s ... 113s > > 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. Thanks,
