[lkml]   [2009]   [Feb]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH F 06/12] OMAP2/3 clock: every clock must have a clkdm
Hello Russell,

On Sat, 31 Jan 2009, Russell King - ARM Linux wrote:

> On Wed, Jan 28, 2009 at 12:35:15PM -0700, Paul Walmsley wrote:
> > Every OMAP2/3 clock must now have an assigned clockdomain, so we can
> > remove the checks from clk_enable()/clk_disable(). Instead, verify
> > that the clockdomain is present only at clock init time via the
> > arch_clock clk_register() hook.
> I don't see the point of requiring all clocks to have a clockdomain field.

For physical clocks, the idea is to match the OMAP2/3 hardware, in which
nearly every clock is associated with a clockdomain. The point for
virtual clocks is to discourage the use of virtual clocks.

> Given that we have virtual clocks, it is quite reasonable for some clocks
> not to have a clock domain associated with them.

Virtual clocks should soon be eliminated from the OMAP clock tree. (
"Virtual clocks" is here used to refer a clock that does not have a
referent in the hardware; the usage of the term in the code is loose.)
So far as I know, all of the OMAP virtual clocks have either turned out to
be superfluous (and prone to spinlock recursion bugs), or to be better
implemented outside of the clock framework (such as the virtual OPP
clocks). We've eliminated the former. We should be able to eliminate the
latter in a few months.

Only physical hardware clocks will then be left in the clock tree. For
OMAP2/3, it makes sense to keep those grouped into clockdomains, since
that is how they are implemented in the hardware.

So I'd recommend keeping this requirement.

- Paul

 \ /
  Last update: 2009-02-03 10:29    [from the cache]
©2003-2014 Jasper Spaans. hosted at Digital Ocean