FreeLists / haiku-development / [haiku-development] Re: BFS and write back problems

["Urias McCullough" <umccullough@xxxxxxxxx>]

On 02/04/2008, Marcus Overhagen <marcusoverhagen@xxxxxxxx> wrote:
> Stephan Assmus schrieb:
>
> > Alexander G. M. Smith wrote:
> >
> > > dolgov wrote on Wed, 02 Apr 2008 13:19:19 +0300:
> > >
> > > > I thought about adding flush() in several places in code related to
> file transactions with ifdef BeOS [...]
> > > >
> > > We did that in the mail daemon replacement, after it had finished
> receiving a new batch of e-mail.  Makes things a bit more reliable in BeOS
> R5, particularly if a mail fetch happened while debugging other crashy code.
>  Though it does annoyingly spin up all sleeping hard drives with mounted
> file systems.
> > >
> >
> > Actually, I think something like that is absolutely justified. It's the
> users email after all. I would go as far as add such behaviour to text
> editors in general. ;-)
> >
>
>  I think that we need to be very careful with this. Applications shouldn't
> ever need to flush or sync. The kernel must handle this
>  automatically.
>
>  The kernel really needs to flush dirty data to the disc once the
>  system is "idle" after a disk operation (to be defined) for some
>  time (i.e. 500ms for removeable media, 3000 ms for permanent storage).

Disclaimer: I'm probably over my head for this discussion.

Travis was mentioning in #haiku yesterday about some differences
between how say OSX does write-backs vs. Linux.

Apparently OS X has very specific intent to flush ASAP after a file
close to ensure the highest probability that the data is written. I
don't know the details, or I'm probably misunderstanding what he meant
- but it would seem for a desktop OS that user sanity should hold a
little more weight than possible performance gains...

Linux apparently is quite lazy by comparison.

I would think Haiku should take a more sane and furious-write approach
vs. higher performance. Again, my uninformed opinion :)

- Urias