On Thu, Mar 25, 2021 at 7:26 PM Chris Murphy lists@colorremedies.com wrote:
The problem is well understood for some time. https://lwn.net/Articles/572911/
This is an update on that 8 year old story. Last year writebehind patches were proposed, and a discussion ensued.
https://lore.kernel.org/linux-mm/CAHk-=whf2BQ8xqVBF8YuxRznByrP-oTgcHSY9DgDnr...
I don't think an fsync/fdatasync centric workload is going to be affected by these knobs. And it'd take testing to find out if they affect anything we care about, but it's interesting that changing them could improve real world performance while negatively impacting synthetic benchmarking.