On Tue, 24 Apr 2001, Russell King - ARM Linux wrote:
> On Tue, Apr 24, 2001 at 03:04:43AM -0700, Russ Dill wrote:
> > --- linux/Documentation/Configure.help Tue Apr 24 03:01:39 2001
> > +++ linux.curr/Documentation/Configure.help Tue Apr 24 02:55:11 2001
> > @@ -16802,8 +16802,9 @@
> > address divisible by 4. On 32-bit ARM processors, these non-aligned
> > fetch/store instructions will be emulated in software if you say
> > here, which has a severe performance impact. This is necessary for
> > - correct operation of some network protocols. With an IP-only
> > - configuration it is safe to say N, otherwise say Y.
> > + correct operation of some non-IP network protocols as well as the
> > + proper operation of mtd (Memory Technology Device Subsystem). If
> > + unsure, say Y here.
>
> If that is true, then we have a problem.
>
> I can imagine that some MTD chips would require a complete word to be written
> _in one write_ rather than several. I'd like to confirm why the MTD layer
> requires this before putting this in.
Actually, it's not the MTD layer (at least not for the flash I personally
use). JFFS2 however has some rare cases where unaligned buffers are
being passed to the MTD layer which for some reasons would rander the code
unnecessarily complicated to work around, and/or impact performances for the
99% cases where data is correctly aligned.
That's what started the discussion about the alignment fixup code on the
iPAQ mailing list where Alan Cox concluded with the affirmation that it
shouldn't be possible to disable the alignment trap at all in the kernel.
Nicolas
_______________________________________________
http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
Please visit the above address for information on this list.
