10:36 <Darkmatter66_> Hi narmstrong ,I'm getting and error when rmmod lima and modprobe lima

10:36 <Darkmatter66_> https://0x0.st/zlP1.txt

10:37 <Darkmatter66_> narmstrong, any idea why is that

10:37 <Darkmatter66_> narmstrong, during boot the hw is proped and I get this https://0x0.st/zlPQ.txt

10:40 <xdarklight> Darkmatter66_: "IRQ ppmmuX not found" is to be expected, it's because the GPU only has three PP instances (but the driver supports up to 8 and informs us that 5 are not implemented by the HW)

10:42 <Darkmatter66_> xdarklight, yes i was guessing it's normal, the fail after modprobe is not normal though

10:43 <xdarklight> yep, that other error is strange indeed. I tested with my Le Potato board yesterday (S905X, so same silicon as your S905D) and didn't get that error

10:44 <Darkmatter66_> xdarklight, I don't suppose it could be a linux config error ? I use my own config

10:46 <xdarklight> so do I. what kernel version are you running?

10:49 <Darkmatter66_> 5.4.0 corrently at commit 81b6b96475ac7a4ebfceae9f16fb3758327adbfe

10:54 <xdarklight> afaerber is seeing the same error on his Realtek SoC: https://patchwork.ozlabs.org/cover/1188734/ (that was like 4 weeks ago)

10:57 <xdarklight> he's suspecting clocks, voltage supply and/or register sizes - I think for Amlogic that only leaves clocks because VDDEE is the voltage supply (if that was disabled the SoC wouldn't work) and register sizes seem to work for others

10:57 <xdarklight> (and in theory it could also be a bug in the lima driver)

11:01 <Darkmatter66_> xdarklight, if the problem is the clock shouldn't the driver crash at boot also

11:06 <xdarklight> Darkmatter66_: I've seen both. for example: if you try to access the VPU registers with it's clock disabled then the CPU locks up. however, if you do the same with the USB PHY registers on GXL then you simply read 0x0. I'm not sure how the Mali GPU is integrated and how it should behave in such case

11:17 <narmstrong> Darkmatter66_: you must have something missing in your config, check all the meson clock configs

11:20 <narmstrong> Darkmatter66_: to reduce the config, I do make defconfig, sed =m to =y, then I run menu config and I remove all the other arch (rockchip, ...) then other subsystems (pcie, acpi, SATA, ...) with this you should have a good start

11:20 <narmstrong> Then you can go into the subsystems and remove the useless drivers for other vendors

11:23 <Darkmatter66_> narmstrong, https://0x0.st/zlP6.txt

11:23 <Darkmatter66_> narmstrong, i think all clocks are enabled

11:48 <narmstrong> Darkmatter66_: can you dump the clk_summary on debugfs ?

11:50 <Darkmatter66_> narmstrong, sure, https://0x0.st/zlPR.txt

11:52 <narmstrong> Looks ok, did you enable clock measurer ?

11:52 <narmstrong> If yes can you dump it ?

12:03 <Darkmatter66_> narmstrong, https://0x0.st/zlZP.txt

12:16 <narmstrong> Darkmatter66_: could you tweak the Lima driver to return 0 right after enabling the clocks to check if these are really enabled ?

12:29 <Darkmatter66_> narmstrong, you mean like this https://0x0.st/zlZa.txt ?/

12:35 <Darkmatter66_> narmstrong, I used this patch https://0x0.st/zlZM.txt

12:35 <Darkmatter66_> narmstrong, and I get the same error

12:35 <Darkmatter66_> narmstrong, tell me if my patch is wrong

13:01 <narmstrong> Return 0 after Lima clk init in the main probe function

16:56 <Darkmatter66> narmstrong, after patching the driver i can modprobe the device and i get "[drm] Initialized lima 1.0.0 20190217 for d00c0000.gpu on minor 0"

16:56 <Darkmatter66> but if i rmmod it it shows a segfault

16:57 <Darkmatter66> narmstrong, and the following in dmesg

16:57 <Darkmatter66> narmstrong, https://0x0.st/zlNJ.txt

17:30 <narmstrong> Dump the clocks right after modprobe

17:50 <Darkmatter66> narmstrong, https://0x0.st/zlqN.txt

18:08 <narmstrong> They seem probed and enabled, and what about the clk measure to be sure the gp0 pll has locked

18:31 <Darkmatter66> narmstrong, https://0x0.st/zlqV.txt

18:41 <Darkmatter66> narmstrong, these logs i'm posting are from the first modeprobe, while before the patch it fails in the second modprobe, does that make a difference?

18:45 <narmstrong> Nop, the logs show the clock is right, so it’s something else

20:43 <Darkmatter66> narmstrong, Are those errors normal? https://0x0.st/zlbf.txt what do they mean and are they related to the probe problem ?