Most of the cpufreq table is sorted in descending order.
But when cpufreq table is made from dev_pm_opp_init_cpufreq_table,
it is sorted in ascending order. So in some case to make cpufreq table
with descending order it needs an extra work.
Signed-off-by: Jonghwan Choi <[email protected]>
---
drivers/base/power/opp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/base/power/opp.c b/drivers/base/power/opp.c
index 2553867..8d81a46 100644
--- a/drivers/base/power/opp.c
+++ b/drivers/base/power/opp.c
@@ -649,7 +649,7 @@ int dev_pm_opp_init_cpufreq_table(struct device *dev,
return -ENOMEM;
}
- list_for_each_entry(opp, &dev_opp->opp_list, node) {
+ list_for_each_entry_reverse(opp, &dev_opp->opp_list, node) {
if (opp->available) {
freq_table[i].driver_data = i;
freq_table[i].frequency = opp->rate / 1000;
--
1.7.10.4
On Tue, Apr 29, 2014 at 10:54 AM, Jonghwan Choi <[email protected]> wrote:
> Most of the cpufreq table is sorted in descending order.
Which part of kernel says that?
> But when cpufreq table is made from dev_pm_opp_init_cpufreq_table,
> it is sorted in ascending order. So in some case to make cpufreq table
> with descending order it needs an extra work.
Order of this table shouldn't matter at all in cpufreq layer..
Hi~
> > Most of the cpufreq table is sorted in descending order.
>
> Which part of kernel says that?
>
-> I couldn't check all. Sorry.
> > But when cpufreq table is made from dev_pm_opp_init_cpufreq_table, it
> > is sorted in ascending order. So in some case to make cpufreq table
> > with descending order it needs an extra work.
>
> Order of this table shouldn't matter at all in cpufreq layer..
-> You are right. But 5440 cpufreq driver write an index number instead of
clk divider value
for change DVFS. And our another(will submit) also write an index number for
changing DVFS.
As you said, order of table shouldn't matter at all in cpufreq layer. Then,
could this can be applied?
In our case, We want to use index 0 for P0 and index 1 for P1.....
Thanks
Best Ragrds.
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of
> Viresh Kumar
> Sent: Tuesday, April 29, 2014 2:33 PM
> To: Jonghwan Choi
> Cc: open list; Rafael J. Wysocki; Len Brown
> Subject: Re: [PATCH 1/2] PM / OPP: Use list_for_each_entry_reverse instead
> of list_for_each_entry
>
> On Tue, Apr 29, 2014 at 10:54 AM, Jonghwan Choi <[email protected]>
> wrote:
> > Most of the cpufreq table is sorted in descending order.
>
> Which part of kernel says that?
>
> > But when cpufreq table is made from dev_pm_opp_init_cpufreq_table, it
> > is sorted in ascending order. So in some case to make cpufreq table
> > with descending order it needs an extra work.
>
> Order of this table shouldn't matter at all in cpufreq layer..
On 29 April 2014 11:20, Jonghwan Choi <[email protected]> wrote:
> -> You are right. But 5440 cpufreq driver write an index number instead of
> clk divider value
> for change DVFS. And our another(will submit) also write an index number for
> changing DVFS.
> As you said, order of table shouldn't matter at all in cpufreq layer. Then,
> could this can be applied?
These two patches aren't going to fly I believe. Depending on the order
of table for setting hardware is inviting trouble.
> In our case, We want to use index 0 for P0 and index 1 for P1.....
What I would recommend is, use .driver_data field to hold what has to
be written to hardware for any frequency. And then simply use
driver_data instead of index.