svn commit: r340187 - head/sys/geom

Warner Losh imp at bsdimp.com
Wed Nov 7 20:31:53 UTC 2018


On Wed, Nov 7, 2018 at 1:20 PM Poul-Henning Kamp <phk at phk.freebsd.dk> wrote:

> --------
> In message <
> CANCZdfrK-7vBETkhfQ9KkM4usrvjVh8wS_OZJiDmH5336DTsqQ at mail.gmail.com>
> , Warner Losh writes:
>
> >BIO_FLUSH is primarily done to force ordering points,
>
> Originally BIO_FLUSH was defined the way it is, to make it possible
> to flush an isolated specified range on providers which support that
> so that fsync(2) could be implemented that way.
>
> I can't remember the exact semantics of the two "magic" flush
> operations (off=0,len=0) and (off=end,len=0) but they were different
> from each other in some important aspect (Pawel?)
>

I couldn't find any place where the different types of flush are
differentiated (nothing in CAM for sure), though I suppose I could have
missed something. All the CAM drivers just use it to push data to the
device and don't look at the flags.

Warner


More information about the svn-src-all mailing list