On 3/25/2013 2:15 PM, Viresh Kumar wrote:
On 25 March 2013 14:06, Sekhar Nori nsekhar@ti.com wrote:
There is a line in the code a little above the ones you deleted that also sets these same variables. I guess you were relying on that line to set policy->cur, but that also sets policy->{min, max} which can be cleaned up.
This code is rather confusing or wrong, this was the state of code before this patch:
policy->cur = policy->min = policy->max = davinci_getspeed(0);
if (freq_table) { result = cpufreq_frequency_table_cpuinfo(policy, freq_table); if (!result) cpufreq_frequency_table_get_attr(freq_table, policy->cpu); } else { policy->cpuinfo.min_freq = policy->min; policy->cpuinfo.max_freq = policy->max; }
policy->min = policy->cpuinfo.min_freq; policy->max = policy->cpuinfo.max_freq; policy->cur = davinci_getspeed(0);The tricky part is if/else, where if don't return error if cpufreq_frequency_table_cpuinfo() fails. We want to set ->min[max] and cpuinfo.min[max] always. And i can see this code not doing that for some case even with my patch.
Possible scenarios:
- Valid freq_table: My patch + what you suggested is required.
- Invalid freq_table: We never set cpuinfo.min[max] with or without my patch
- No freq_table: Only my patch is required.
If i do what you suggested then 2 and 3 would fail... If you want to return error in case cpufreq_frequency_table_cpuinfo(), then i can fix it properly.
So down in the cpufreq driver probe below, we bail out if freq_table is not provided. So all this checking for freq_table in the code you pasted above is superfluous. If you can clean that part up and add checking for cpufreq_frequency_table_cpuinfo() as you proposed, I will be glad to test it out ;)
Thanks, Sekhar