06:46 <naobsd> mmind00: can you see anything about rk3288-timer on running system?

06:47 <naobsd> mmind00: only arch_sys_counter jiffies are available in /sys/devices/system/clocksource/clocksource0/available_clocksource

06:48 <naobsd> CONFIG_ROCKCHIP_TIMER=y and /sys/devices/platform/ff810000.timer is available...

06:55 <naobsd> mmind00: I'm using next-20150407 + some dts patches from your dts branch

07:02 <naobsd> mmind00: btw tsadc is working with hym8563 and tsadc dts patch :) do you still need tested-by for them?

07:05 <mmind00> naobsd: Tested-by-Tag is always helpful ... especially, as I want to try squeezing those two in for 4.1 still :-)

07:07 <mmind00> naobsd: as for the timer ... don't know at the moment, would need to check ... is /sys/devices/platform/ff810000.timer/driver present and lead to somewhere? [when the driver symlink is present, it means the device was bound to a driver]

07:08 <naobsd> oh, no ff810000.timer/driver :(

07:10 <mmind00> ok, strange ... I'll take a look after breakfast :-)

07:11 <rperier> naobsd: It worked yesterday btw, the right bootloader is started

07:12 <rperier> parameter works fine too, where should I write my kernel,btw ? after seek=0x2000 (which is parameter.img) ?

07:14 <naobsd> partition offset/size in parameter is _logical_ sector number. 0x2000th sector on SD card(and eMMC) is treated as logical sector 0 on u-boot (as same as RK proprietary loader)

07:15 <naobsd> rperier: e.g. if you use offset 0x8000 for kernel partition, you need to specify seek=$((0x2000+8000))

07:16 <rperier> and this partition should start at 0x8000 in the parameter file, right ?

07:18 <naobsd> rperier: yes

07:18 <rperier> I see

07:18 <naobsd> rperier: "if you use offset 0x8000 for kernel partition _in parameter_ ..." sorry ;)

07:18 <naobsd> ah

07:18 <rperier> ;)

07:18 <naobsd> seek=$((0x2000+0x8000)) <- needs "0x"

07:19 <rperier> yes, of course

07:28 <naobsd> mmind00: about rk3288-timer, I can see debug prints(I added), clockevents_config_and_register should be called which is in last line of rk_timer_init()

07:30 <mmind00> naobsd: ok ... i.e. it's registering a clockevents device, so I guess it is not meant to show up in clocksources? [never touched the timer stuff much]

07:37 <naobsd> ah

07:38 <naobsd> I roughly grep-ed arm arch/generic timer

07:38 <naobsd> there are clocksource_register_hz()

07:38 <naobsd> sched_clock_register()

07:39 <naobsd> timecounter_init()

07:39 <naobsd> but they don't exist in rockchip_timer.c

07:40 <naobsd> "it is not meant to show up in clocksources" may be right...

07:41 <naobsd> "This driver provides the basic timer functionnality"...what's basic timer func ;)

07:44 <naobsd> ok, at least I misunderstood about clockevents_config_and_register()

07:46 <naobsd> clockevent and clocksource are different thing

07:46 <mmind00> naobsd: the rockchip-timer is supposed to provide an alternative in the suspend case [when the architected timer might be turned off] ... same on the Cortex-A9 ones where we already have the Global/local timers

07:46 <mmind00> so it should be ok that its function is limited

07:46 <naobsd> I see

07:46 <naobsd> but I want to use rk3288-timer on RK3188 as for clocksource

07:47 <naobsd> arm_global_timer cannot be used with cpufreq

07:48 <naobsd> cannot be used -> is not fixed rate

07:49 <naobsd> rk3066 should be ok because it uses dw_apb_time which should provide clocksource

07:49 <naobsd> probably I should add rk3188-timer compat which provides clocksource

07:51 <naobsd> maybe for sched clock too?

07:55 <naobsd> (no idea timecounter... ETOOMANYTIMERSUBSYSTEMS!)

08:13 <mmind00> naobsd: not sure I understand ... my Radxa Rock runs perfectly fine with cpufreq and the global timer

08:13 <naobsd> mmm

08:13 <naobsd> I should check again with latest kernel...

08:13 <naobsd> I think "date; sleep 10; date" was wrong

08:14 <mmind00> naobsd: also the rate clocks for the timers (SCLK_TIMERx) are sourced from xin24m directly, so not dependent on the cpu rate at all

08:14 <mmind00> naobsd: bah ... scrap that last line :-)

08:15 <mmind00> naobsd: talking about the wrong timer block

08:15 <naobsd> I cannot remember about parent clock for global timer

08:18 <naobsd> CORE_PERI

08:18 <mmind00> yep ... but that was also only my guess ... the global-timer doesn't get mentioned in the docs at all

08:26 <naobsd> hmm... I want to reuse dw_apb_timer.c as rk_timer ;)

08:27 <naobsd> currently only lower 32bit part is used in rk_timer. it's basically same as dw_apb_timer except register offset ;)