tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: write alignment matters?

On Mon, Jun 27, 2011 at 02:00:54PM -0400, Paul Koning wrote:
> Sure it is fair.  The fact that some operating systems had extremely 
> simple/primitive file systems without any caching means that those systems 
> were capable of surviving file device pulls.  But it isn't reasonable to 
> consider that a feature.  RT-11 was in fact an unusual example in its days;
> few if any operating systems of that era had a "just yank it" property.  

I would consider it very reasonable for an OS to survive file device pulls.

Expecting the on-disk data to be consistent if you yank the device w/o telling
the OS to flush its cache is another issue entirely, but having the OS hold
on to the data (like Erik Fair wrote in a separate email) and be able to
flush it when/if the device is reconnected sounds pretty "reasonable" too.

Combine that w/ the ability notice that all cached data has been written and 
the device has been idle for a bit (perhaps 10 seconds?) and perform a
behind-the-scenes unmount + automatic remount if it's accessed again, and
I'd call that call that a pretty great feature.


Home | Main Index | Thread Index | Old Index