11:31 <uis> What difference between hclk and aclk?

11:31 <uis> And sclk

12:13 <punit> and pclk

12:27 <mmind00> don't quote me on actual terms ... but in general hclk/pclk are the ones supplying the circuits of a peripheral, so accessing their registers needs these to be active

14:35 <wens> my guess is a/h/p refer to axi/ahb/apb

14:36 <wens> sclk is special clock or module clock

14:40 <mmind00> what wens said :-)

14:41 <mmind00> mriesch: we're going up to -rc3 now, so I guess for the gmac issue it would definitly be time for some sort of revert

14:42 <mmind00> mriesch: not sure about the pm_runtime details, but can't we just remove the pm_runtime_enable/_disable parts to rely on the stmmac core to do those?

14:42 <mmind00> mriesch: though dwmac_mediatek also does the full enable+get stuff

16:08 <uis> If trm says crypto max freq is 150m means max sclk rate is 150?

16:10 <uis> mmind00

16:44 <diederik> I stumbled upon this yesterday: https://github.com/LibreELEC/LibreELEC.tv/blob/master/projects/Rockchip/patches/linux/default/linux-1002-for-libreelec.patch#L647

16:44 <diederik> no idea if it's useful or even relevant, but thought I'd mention it anyway

18:24 <uis> What purpose of trng_flush?

19:17 <mriesch> mmind00: makes sense. due to vacations and the lack of rk3399 hardware (both conditions are planned to change in three weeks or so) i am afraid i cannot be of much help, but let's see...

19:20 <mriesch> mmind00: ...i guess the partial revert you suggested would get rid of the annoying warning, yet should leave the ethernet on the rk3399 intact

19:28 <mriesch> mmind00: if this partial revert works it would be at least a nice intermediate steps. i think we still need to take a look on what is going on here (and maybe compare it to other glue drivers such as the dwmac_mediathek), but the clock is not ticking on that one then

19:28 <mriesch> s/steps/step

19:34 <mriesch> mmind00: btw thanks for the replies to the rfc series. yes, pretty sure that wake-on-lan will break -- at least the internal phy will not be left powered on. not sure what happens to any external phy, though

19:34 <mmind00> uis: I'd guess so, but have never looked too deeply into the crypto module at all

19:36 <mriesch> mmind00: as stated in the cover letter, i could use some pointers here. i'll catch up the work on this cleanup some time in october, if all goes as planned

19:37 <mmind00> mriesch: and yep, that partial revert worked for me on rk3399-puma ... gmac "broken" before and adding the pm_runtime_get_sync + put back in alone (without the enable) did work for me at least

19:37 <uis> Is 100% monobits healthy for TRNG?

21:15 <uis> Waait. Is HCLK_CRYPTO_SLV slave AHB for DMA? If so, then why upstream rk3228 crypto enable DMAC1 clk?

22:06 <punit> diederik: That's the commit that broke ethernet on rk3399.

22:08 <diederik> ah ok, thx.

22:16 <punit> mmind00: The partial revert also gets ethernet going on the rockpro64 for me.

22:17 <punit> Looking at my debug prints, the additional reference prevents the device from going through multiple suspend/resume cycles

22:19 <punit> It effectively disables runtime idle

22:19 <punit> ... for the gmac