04:10 <MoeIcenowy> dracc: unfortunately AXM-8-256 is Volcanic

04:10 <MoeIcenowy> and Volcanic support is a big TODO

06:40 <mwalle> NishanthMenon: thanks! looks like you get it to work. I've git the same of_clk_set_defaults() hack, but it didn't work. I'll compare yours with mine later

06:41 <mwalle> NishanthMenon: and that was my initial question, wether the dm firmware does something different on the j722s regarding the clocking

06:42 <mwalle> NishanthMenon: it seems that on the j721s2 you're able to set the clock speed although the device isn't turned on yet. But not on the j722s

06:52 <mwalle> NishanthMenon: ah I see. you've mentioned that the GPU on the j722s has its own PLL. so for the j721s2 it just worked by accident? :)

07:10 <dracc> MoeIcenowy: Oh, bummer. Thanks for the reply!

07:27 <mwalle> NishanthMenon: mhh, still doesn't work for me, even with your patches. What firmware are you using?

10:04 <NishanthMenon> mwalle: the upstream linux-firmware.

10:10 <NishanthMenon> mwalle https://gitlab.com/kernel-firmware/linux-firmware/-/tree/main/powervr?ref_type=heads

11:40 <mwalle> NishanthMenon: https://gist.github.com/mwalle/dadf805ed3eb483ab0cc82ed8584c305 that's my bootlog. You were using kernel modules, right? I'm using everything needed compiled in (incl. the firmware) I'm just building your kernel with arch defconfig

11:41 <NishanthMenon> mwalle: yes, i used the defconfig as is (is modules)

11:43 <NishanthMenon> mwalle: to answer your question if DM operates differently, from what I understand, jacinto DM+bootloader sequences prioritize enabling all PLLs early on to save on boot time Vs Sitara DM which will enable PLLs only on need to save power (different priorities)

11:43 <NishanthMenon> mwalle: my firmware is in the rootfs, kernel modules are in rootfs as well. i have'nt tried to built-in drivers or firmware

11:44 <mwalle> NishanthMenon: yeah, I'll try that one one

11:45 <mwalle> NishanthMenon: maybe anything in my bootlog will catch your eye

11:47 <NishanthMenon> mwalle: might be disable the oldi dt node :P

11:48 <NishanthMenon> "[ 7.717573] powervr fd80000.gpu: [drm] *ERROR* Firmware failed to boot" that is weird..

11:48 <NishanthMenon> fd9fb9f378ced6a3c08d6ecc2298d90eae6317b3cba271bed02baca00fae005d rogue_33.15.11.3_v1.fw

11:48 <NishanthMenon> /lib/firmware/powervr$ sha256sum *

11:48 <NishanthMenon> 3075875ab63de9c5ae6d42da70a35a843ac22fe7c9d564b18ef657efae0d3273 rogue_36.53.104.796_v1.fw

11:49 <mwalle> NishanthMenon: oldi node.. yeah but i don't think that will stop the GPU from probing ;)

11:50 <mwalle> $ md5sum powervr/*

11:50 <mwalle> fbaa8bf2d40c7cfdc5103186ee19bd88 powervr/rogue_36.53.104.796_v1.fw

11:50 <mwalle> err

11:52 <mwalle> NishanthMenon: i've just downloaded the binary from both git.kernel.org linux-firmware and from your link above

11:52 <mwalle> ahh sha256sum.

11:52 <mwalle> $ sha256sum ~/Downloads/rogue_36.53.104.796_v1.fw

11:52 <mwalle> 3075875ab63de9c5ae6d42da70a35a843ac22fe7c9d564b18ef657efae0d3273 /home/mwalle/Downloads/rogue_36.53.104.796_v1.fw

11:58 <NishanthMenon> mwalle: to cross check i just kick started test on am62p, j722s etc.. let me see what the results look like..

12:15 <NishanthMenon> mwalle: https://gist.github.com/nmenon/d3957983d25cfae02823b92c2283ff88 -> ignore the j721s2 fail - the networking issue is known though not rootcaused

14:04 <mwalle> NishanthMenon: still doesn't work https://gist.github.com/mwalle/fe9ae7519fa7e8215125463a50a5c100 even with your exact kernel and modules

14:04 <mwalle> NishanthMenon: what are your TF-a, optee and tidm firmware versions?

14:05 <NishanthMenon> On the tip

14:28 <NishanthMenon> Sigh.. FW_BOOT_TIMEOUT_USEC -> what would happen if we made it a longer duration.. it is already 5 seconds.. not really sure how fwif_sysinit->firmware_started gets set :(

16:16 <MoeIcenowy> simon-perretta-img: looks like clc is dropped on dev/bxs?

16:16 <MoeIcenowy> (and some kmsro zink enablement disappeared

16:19 <simon-perretta-img> MoeIcenowy: clc bits should still be there under src/imagination/pco/usc{gen,lib} IIRC?

16:20 <simon-perretta-img> I think some zink bits may have disappeared as they can't be tested/used anymore post-MR 34822

16:25 <MoeIcenowy> oops I didn't see commit message on clc

16:25 <MoeIcenowy> simon-perretta-img: oops then dynamic rendering needs to be implemented for imagination driver to support zink ?

16:31 <simon-perretta-img> Yeah, we're planning on adding support for the extension and could possibly end up switching over to using the common runtime impl for the renderpass stuff as part of it too