On 11/19/2013 08:26 AM, Viresh Kumar wrote: > On 19 November 2013 19:46, Nishanth Menon <nm@xxxxxx> wrote: >> Consider something like userspace governor selection -> the device at >> boot will probably remain in an unknown/"invalid" configuration till >> the very first transition attempt. I am less worried about the stats >> than not following what the hardware description is (as stated by >> device tree/other forms). >> >> I staunchly disagree that at a point of mismatch detection, we just >> refuse to load up cpufreq governor -even though we know from device >> tree/other alternative entries what the hardware behavior is supposed >> to be. To refuse to loadup to a known configuration is considering the >> "valid configuration" data provided to the driver is wrong - an >> equivalent(considering the i2c example) is that if i2c driver sees bus >> configured for 3.4MHz and was asked to use 100KHz, it just refuses to >> load up! > > CPU looks to be a bit different in that aspect as compared to I2C. We > aren't really sure if I2C will work at the existing freq but we are 100% > sure that current freq of CPU is valid enough, otherwise we wouldn't > have reached to this point.. :) > Not completely true - reaching probe after boot in a few seconds may not mean that system will remain stable at that frequency for longer duration. From a silicon vendor perspective, I do know that we gaurentee the discrete frequencies in the data manual (and that gets populated in devicetree and hence in freq_table), but we will not guarentee any other frequency to be functional for any length of time. in short, if a actual product is manufactured and operational at a frequency we do not "officially support", there is a risk associated with that. just a boot on a few development systems do not ever guarentee productization capability. >> The above two are fair comments -> but that implies that policy->cur >> population should no longer be the responsibility of cpufreq drivers >> and be the responsibility of cpufreq core. are we stating we want to >> move that to cpufreq core? > > I am sure you want to have a look at this: my bad. I missed this one. So, to summarize: what is our overall strategy here? to move to a frequency matched in freq_table OR just giveup? I can try and respin accordingly. > > commit da60ce9f2faca87013fd3cab1c3bed5183608c3d > Author: Viresh Kumar <viresh.kumar@xxxxxxxxxx> > Date: Thu Oct 3 20:28:30 2013 +0530 > > cpufreq: call cpufreq_driver->get() after calling ->init() > > Almost all drivers set policy->cur with current CPU frequency in > their ->init() > part. This can be done for all of them at core level and so they > wouldn't need > to do it. > > This patch adds supporting code in cpufreq core for calling get() > after we have > called init() for a policy. > > Signed-off-by: Viresh Kumar <viresh.kumar@xxxxxxxxxx> > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx> > --- > drivers/cpufreq/cpufreq.c | 11 +++++++++++ > 1 file changed, 11 insertions(+) > -- Regards, Nishanth Menon -- To unsubscribe from this list: send the line "unsubscribe cpufreq" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
