00:51 <apritzel> so does anyone have a user manual for the T527, specifically? There is just a data sheet, which helps with the pins, but doesn't mention any clocks :-(

00:52 <apritzel> I got most from some BSP files (ccu-sun55iw3.c on Github), but those seem unreliable, and there are definitely mistakes in there

00:54 <apritzel> anyway, I got compiling clock drivers (for normal and PRCM) now, but will leave the testing for tomorrow ...

01:11 <dsimic> FWIW, I tried to find the T527 user manual, which yielded nothing

01:31 <MoeIcenowy> T527 should be (at least mostly) the same with A523

01:31 <MoeIcenowy> although I think A523 only claims one Ethernet MAC

01:36 <iscle> So, any of you know how reboot works in sunxi platforms? I'm trying out the port for V853 and for some reason "reboot" won't work, it gets stuck at "[ 11.685719] Reboot failed -- System halted"

01:37 <iscle> I've checked other devices (such as D1 which is what I'm basing it on) and can't see any references to "reset" or "reboot"

01:41 <dsimic> iscle: have a look how devm_register_sys_off_handler() is used in drivers/mfd/axp20x.c

01:41 <dsimic> s/a look/a look at/

01:42 <dsimic> oh, wait, you asked about restarting

01:44 <iscle> Yep, exactly. I've just seen that it is normally handled by the wdt, in the "sunxi_wdt_restart" function. I'll take a deeper look as I might have not configured it correctly for the SoC (I assumed it was 1:1 with the D1, which is what everything else is based on)

01:49 <iscle> Ah, this might be it: https://github.com/torvalds/linux/blob/de5cb0dcb74c294ec527eddfe5094acfdb21ff21/arch/riscv/boot/dts/allwinner/sunxi-d1s-t113.dtsi#L230

01:53 <iscle> Yep! Removing that line fixed it :)

01:53 <dsimic> good to hear :)

09:38 <apritzel> MoeIcenowy: yes, I was building on this "mostly the same with A523", and it's definitely helpful

09:39 <apritzel> but it only goes so far: the second EMAC is one example, but there is more: CAN, HDMI, second LVDS, ...

09:39 <apritzel> and for the clocks this does matter, and I'd rather have all the clocks in from the beginning

09:40 <apritzel> I see the obvious gaps in the memory map, and can compare this to the other recent SoCs (like D1 and H616), but it would be good to have it for real

09:40 <apritzel> for instance some clocks subtly changed in the A523 (like the already mentioned MMC clock), and I think there are others who lost a .P factor and such

09:42 <apritzel> and for instance the second EMAC seems to be a bit special, at least in the T527 datasheet it's described separately, and there is a second reset bit in a separate BGR register

09:43 <apritzel> so if someone saw a T527 user manual falling off a lorry, I'd be grateful to hear about it ;-)

12:48 <IlikeTech> If I want to use one of the X-Powers PMICs (over I2C), I just need to choose one that actually can provide the appropriate voltage rails, correct? Trying to figure out what goes with the F1C100s, and it looks like the AXP192.

12:48 <IlikeTech> though there's also the AXP199 which is maybe the same?

13:09 <apritzel> IlikeTech: there is no real technical relationship between AXP models and SoCs, it seems just that AW bundles pairs of them. I guess the rails provided somehow match the requirements of a particular SoC, maybe?

13:10 <IlikeTech> okay, that's what I figured

13:10 <IlikeTech> looks like the 192 should work for the F1C100 as it has 3 buck rails and a few LDOs

13:10 <apritzel> IlikeTech: I tried to summarise the bunch here: https://linux-sunxi.org/AXP_PMICs

13:11 <IlikeTech> only issue is that if the lithium battery is below 3.3v, the 3.3v rail will sag (and no boost for 5V OTG)

13:12 <IlikeTech> thanks

13:12 <apritzel> I think if you know what you are doing, you could choose any PMIC. I'd guess one of the more modern ones are better?

13:13 <apritzel> I *think* only the very recent ones have built-in boost converters? AXP717 definitely has, but is probably overkill for an F1C100s?

13:14 <IlikeTech> for the USB powered board I just used 3 TI switchers and an LDO which works fine

13:14 <apritzel> definitely the A64 boards with battery support had discrete boost converters, IIRC

13:14 <IlikeTech> but battery powered would be nice

13:15 <apritzel> yeah, I figured "charging" is a killer argument for a PMIC

13:16 <apritzel> seems like the AXP209 maybe? although it only has two DCDCs

13:36 <IlikeTech> I'll take a look!