[PATCH 2/9] cpm2: Fix off-by-one error in setbrg().

[Vitaly Bordug]

On Thu, 30 Aug 2007 15:13:12 -0500
Scott Wood wrote:

> On Thu, Aug 30, 2007 at 02:09:07AM +0400, Vitaly Bordug wrote:
> > On Tue, 28 Aug 2007 15:19:21 -0500
> > Scott Wood wrote:
> > 
> > > The hardware adds one to the BRG value to get the divider, so it
> > > must be subtracted by software.
> > 
> > Prolly a note why it used to work, or what exactly this is
> > resulting in the code. IIRC this was just fw-ported so arch/ppc
> > should have this as well.
> 
> It *didn't* work before -- hence the fix.
> 
It used to work at least at the time I did initial port, but I am not going to argue.

> The failure mode from being off by just one isn't total
> nonfunctionality, but rather a corrupted character now and then,
> which could explain why it wasn't fixed before.
Can this be added to description please? Someone may be grepping for such a weirdness in some custom code forked off
some kernel.org revision, and will be happy to know it is addressed.

> 
> As for arch/ppc, I'm just trying to not break it more than it already
> is.
I mean, when powerpc/ was nearly empty, same issue in ppc/ went unnoticed for years...

-- 
Sincerely, Vitaly