[TUHS] sync; sync; sync; halt

[Tom Ivar Helbekkmo]

Noel Chiappa <jnc at mercury.lcs.mit.edu> writes:

> It all depends on the particular driver; some drivers (e.g. the RP
> driver) re-organized the waiting request queue to optimize head
> motion, using a so-called 'elevator algorithm'.

Ah, yes, now that you point it out, I see the sorting in the RP driver.

> I really think the whole triple-sync thing is mythology.

It looks that way, yes.  The fact that it does a synchronous write of
the superblocks at the start of the sync seems to imply that two
successive calls should ensure a proper flushing of everything, and may
even have been written like that with that effect in mind, but once disk
drivers started optimizing I/O ordering, the effect was lost - which
might just be why the synchronous superblock write was gone in V7.

-tih
-- 
Popularity is the hallmark of mediocrity.  --Niles Crane, "Frasier"