lkml.org 
[lkml]   [2013]   [Sep]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/2] cpufreq: exynos4x12: Use the common clock framework to set APLL clock rate
Hi Yadwinder,

> Hi Tomasz,
>
> >> On Wed, Sep 25, 2013 at 4:52 PM, Lukasz Majewski
> >> <l.majewski@samsung.com> wrote:
> >> > In the exynos4x12_set_apll() function, the APLL frequency is set
> >> > with direct register manipulation.
> >> >
> >> > Such approach is not allowed in the common clock framework. The
> >> > frequency is changed, but the corresponding clock value is not
> >> > updated. This causes wrong frequency read from cpufreq's
> >> > cpuinfo_cur_freq sysfs attribute.
> >> >
> >>
> >> This patch looks incomplete, leaving the driver in untidy state,
> >> perhaps its doesn't fix the above stated problem completely. what
> >> about if (!exynos4x12_pms_change(old_index, new_index)) becomes
> >> true?
> >>
> >> IMHO, this driver needs lot more work in addition to this patch to
> >> cleanly and completely move the cpufreq driver to common clock
> >> framework.
> >
> > I agree that the other case needs to be handled as well. Basically
> > the whole conditional block dependent on exynos4x12_pms_change()
> > can be safely dropped, because this condition is already handled in
> > PLL driver.
> >
>
> Exactly!

After more corner case testing, I admit that corner cases with PLL s
parameter change are still broken (e.g. 1400000 -> 700000).

I will prepare v2 for it.

If I manage I will push it into -rc. If not those changes will be
included to exynos cpufreq rework.

Thanks guys for a thorough review.

>
> > Lukasz is already working on further rework of this driver to clean
> > it up from legacy code, but this will have to wait for 3.13, as
> > 3.12 is already in rc stage and only fixes can be accepted for it.
> >
> >> For fixing this issue urgently, setting CLK_GET_RATE_NOCACHE for
> >> apll in clk driver can also be quicker fix.
> >
> > Unfortunately this is not how this flag works. It only makes
> > clk_get_rate() call ->recalc_rate() operation of the clock instead
> > of instantly returning cached rate - it doesn't seem to work
> > recursively.
> >
>
> hmm.. yes it can't help in our case as it recursively walks only the
> subtree of clk but in our case we are changing rate of parent.
>
> Regards,
> Yadwinder



--
Best regards,

Lukasz Majewski

Samsung R&D Institute Poland (SRPOL) | Linux Platform Group


\
 
 \ /
  Last update: 2013-09-26 11:41    [W:0.220 / U:0.044 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site