Port-arm archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: machdep.frequency.current handling incomplete ?



Hi.

On 20 Jul 2014, at 21:50, Manuel Bouyer <bouyer%antioche.eu.org@localhost> 
wrote:

> On Sun, Jul 20, 2014 at 08:29:03PM +0200, Radoslaw Kujawa wrote:
>> Hi!
>> 
>> On 20 Jul 2014, at 19:22, Manuel Bouyer <bouyer%antioche.eu.org@localhost> 
>> wrote:
>> 
>>> Hello,
>>> I already noticed that setting machdep.frequency.current to 1000 on the
>>> beaglebone would cause it to hang, even with an external 5V supply.
>>> Setting it to 900 used to work (modulo other problems, which are also
>>> present if I don't change the frequency).
>>> 
>>> Now that I have X11 running, I tried glxgears. Up to 800Mhz everything
>>> is fine, but with glxgears running, switching the frequency to 900Mhz
>>> causes the same hang. Note that I see this with both my beaglebone and
>>> beaglebone black, so I don't think I have broken hardware.
>>> 
>>> So I wonder if anyone is aware of something missing to properly switch
>>> the CPU to higher frequencies on the beaglebone ?
>>> Maybe some voltage adjustement in the SoC, or at the tps65217 level ?
>> 
>> I guess it?s voltage too low for operation at desired frequency. Currently 
>> tps65217pmic(4) does not support setting the voltage (only reading back), 
>> but if you compare outputs of envstat(8) when starting NetBSD, while already 
>> running at different frequency in U-Boot, you?ll notice the voltages are 
>> slightly different.
>> 
>> I think Dennis Ferguson already wrote about that in the past (circa 
>> august-sept 2013), so please search the mailing list archive.
>> 
> 
> thanks for the pointer, indeed I did find the post, which pointed me to the
> right part of the device tree.
> Now the problem is that looking for device trees I found conflicting infos.
> The dts supplied with linux 3.5 starts from 275 to only 720 Mhz.
> Can we run as low as 275Mhz ?
> I found other dts fragments with entries for 1Ghz, but they don't have the
> same voltage for the same frequencies.
> And we are using frequencies that are not in these device trees.

Obviously, entries up to 720MHz must be for original BeagleBone, while up to 
1GHz must be for BeagleBone Black.

> So now I'm trying to find some official documentation which would
> give the voltage for a given frequency ...

Search for information about OPP in the official OMAP3530 datasheet. Each OPP 
(operating performance point) level corresponds to a different DPLL clock speed 
setting. There’s also a table with recommended voltages for each OPP in the 
same datasheet.

http://www.ti.com/product/AM3358/datasheet/specifications#SPRS717-32

Note the different settings for ZCZ Package with Device Revision Code Blank 
(i.e. BB) and ZCZ Package with Device Revision Code A (i.e. BBB). 

So, yeah, we should extend the tps65217pmic(4) to allow messing with voltage 
settings after all.

-- 
Best regards,
Radoslaw Kujawa





Home | Main Index | Thread Index | Old Index