Re: fsync error reporting

[Jason Thorpe <thorpej%me.com@localhost>]

> On Feb 19, 2021, at 5:33 AM, Greg Troxel <gdt%lexort.com@localhost> wrote:
> 
> I thought NCQ was supposed to give acks for actual writing, but allow
> them to be perhaps ordered and multiple in flight, so that one could use
> that instead of the big-hammer inscrutable writeback cache.

Certainly in the universe of SCSI tagged command queueing, signaling completion does NOT guarantee that the data is actually encoded in rust.  There is a mode page that controls the caching behavior of the drive, and if “write cache enable” is set (as it is by default with most drives because drive manufacturers want to look good in performance benchmarks), then accepting the write into the drive’s cache is sufficient to report a successful completion.

I would be shocked, honestly, if SATA NCQ made stronger guarantees than SCSI.

-- thorpej