Re: [PATCH 04/18] Blackfin Serial Driver: Change hardware flowcontrol from poll to interrupt driven.

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 6 Mar 2009 17:49:13 -0500
Mike Frysinger <vapier.adi@gmail.com> wrote:

> On Fri, Mar 6, 2009 at 17:38, Andrew Morton wrote:
> > On Fri, __6 Mar 2009 14:42:46 +0800 Bryan Wu wrote:
> >> From: Sonic Zhang <sonic.zhang@analog.com>
> >>
> >> Only CTS bit is affected.
> >>
> >> Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
> >> Signed-off-by: Bryan Wu <cooloney@kernel.org>
> >
> > Another completely useless changelog.
> 
> the summary, while brief, describes the change: the flowcontrol is
> currently implemented by polling, but this changes it to an interrupt
> based system

yes, but it doesn't explain why the change was made.

> > Your [PATCH 00/18] email described all these patches as
> > needed-in-2.6.29 bugfixes. __As far as I can tell (based on almost zero
> > information), this patch is not a bugfix.
> 
> because the current implementation is polling (busy waiting), the
> kernel will basically hang when the remote asserts hardware
> flowcontrol pins.  if this is only asserted for short periods (less
> than 1 char time), then it's not really noticed ... but it still
> destroys latencies.

That.

Please send through replacement changelogs for all the patches which I
complained about.  Then I'll pick through it all and see what we can do
about a 2.6.29 merge.