DC-IRC has quit [Remote host closed the connection]
DC-IRC has joined #armbian-allwinner
<DC-IRC> <viraniac> Interesting that this driver keeps coming up. Other distributions including debian own kernel keeps this disabled. We have it enabled. I added a patch to fix some slowness issue for this driver in https://github.com/armbian/build/pull/5679
<DC-IRC> <viraniac> I have Orange Pi Prime (H5). Let me check if I can reproduce the problem on it as well.
<DC-IRC> <viraniac> I have Orange Pi Prime (H5). I will try to reproduce the problem and get back to you
<DC-IRC> <viraniac> 6.6 kernel, 15 minutes since bootup. USB seems to be working fine for me. I am writing 100 MB file in a loop with data coming from /dev/urandom on a usb flash drive. I lost my uart though after about 10 minutes, it can be my buggy uart dongle, not sure. But ssh is working fine and so is usb
<DC-IRC> <viraniac> After running the test for 45 minutes. I am shutting it down. USB is working fine for me on 6.6 kernel on allwinner H5 (Orange Pi Prime).
juri__ has joined #armbian-allwinner
juri_ has quit [Ping timeout: 246 seconds]
juri__ has quit [Ping timeout: 268 seconds]
juri_ has joined #armbian-allwinner
<DC-IRC> <c0rnelius77> @viraniac Interesting. So it doesn't seem to effect all H5's the same.
<DC-IRC> <viraniac> can it be the power supply you used? Just confirming to remove that from the list of possible culprits
<DC-IRC> <viraniac> In my case, I was using raspberry pi 5v 2.5 amps power supply along with usb to 4.0 barrel jack cable.
<DC-IRC> <c0rnelius77> I've tried several diff PSU ranging in diff amps. Its always the same.
<DC-IRC> <c0rnelius77> I've even powered it off the header which was the last test I just ran using an Armbian img.
<DC-IRC> <viraniac> Hmm.. interesting indeed. So for me both current and edge works fine. For lanefu, current works fine but edge is broken and for you both kernels are broken?
<DC-IRC> <viraniac> Hmm.. interesting indeed. So for me both current and edge works fine. For lanefu, current works fine but edge is broken and for you both current and edge kernels are broken?
<DC-IRC> <c0rnelius77> Yeah
<DC-IRC> <c0rnelius77> This doesn't mean it happens everytime right away, but it does eventually happen.
<DC-IRC> <c0rnelius77> With the Tritium I find it usually dies quick. Quicker than my NEOPLUS2.
<DC-IRC> <c0rnelius77> ```
<DC-IRC> <c0rnelius77> +&mbus {
<DC-IRC> <c0rnelius77> + compatible = "allwinner,sun8i-h3-mbus";
<DC-IRC> <c0rnelius77> + reg = <0x01c62000 0x1000>;
<DC-IRC> <c0rnelius77> + /delete-property/ reg-names;
<DC-IRC> <c0rnelius77> + clocks = <&ccu CLK_MBUS>;
<DC-IRC> <c0rnelius77> + /delete-property/ clock-names;
<DC-IRC> <c0rnelius77> + #address-cells = <1>;
<DC-IRC> <c0rnelius77> + #size-cells = <1>;
<DC-IRC> <c0rnelius77> + dma-ranges = <0x00000000 0x40000000 0xc0000000>;
<DC-IRC> <c0rnelius77> + #interconnect-cells = <1>;
<DC-IRC> <c0rnelius77> +};
<DC-IRC> <c0rnelius77> +
<DC-IRC> <c0rnelius77> ```
<DC-IRC> <c0rnelius77> I started adding that to my DTS.
<DC-IRC> <c0rnelius77> problem goes away.
<DC-IRC> <viraniac> Does blacklisting the sun8i-a33-mbus module solves the issue without modifying dts, or does it only get solved after adding the dt node?
<DC-IRC> <c0rnelius77> Honestly I'm not sure. I never tried to blacklist it. The mbus bit is basically me reverting the node back to what it was in the `sunxi-h3-h5.dtsi` before the change. I also have an A64, but I haven't checked that unit yet. I could try blacklisting it? Not a big fan of the idea though.
<DC-IRC> <c0rnelius77> If debian just stopped including the driver, than blacklisting it would probs work.
<DC-IRC> <c0rnelius77> That `network slowness issue` patch is for eth?
<DC-IRC> <viraniac> why are you deleting reg-names and clock-names cell? Also in the upstream kernel, there are 3 clocks and 2 regs in use. But in your dts, Only one clock and one reg is there
<DC-IRC> <viraniac> we started debugging with ethernet. Even though the board had gigabit ethernet, it was giving 40mbps speed. While debugging, the problem found to be in the mbus driver. It was not scaling up ram that quickly from the lowest value which was causing the slowness. This seemed to have been caused by the autorefresh block in the code. And commenting that solved the issue
<DC-IRC> <c0rnelius77> If you leave those you'll get visible errors in dmesg. those weren't apart of the original node.
<DC-IRC> <viraniac> what is the visible error you get?
<DC-IRC> <viraniac> changing compatible itself would have resulted into the driver not loading, as the upstream driver doesn't support h3
<DC-IRC> <viraniac> changing compatible string itself would have resulted into the driver not loading, as the upstream driver doesn't support h3
<DC-IRC> <c0rnelius77> Beats me know? I was messing with this a week or so ago.
<DC-IRC> <c0rnelius77> Thats what I thought too, but I was still seeing error msgs I believe ,so I just added whole node as it was. I can test it again this weekend and see if that string can just be used and if not what the error was.
<DC-IRC> <viraniac> `I just added whole node as it was` what do you mean by as it was?
<DC-IRC> <c0rnelius77> before the change.
<DC-IRC> <viraniac> hmm.. interesting indeed.
<DC-IRC> <c0rnelius77> Beats me now? I was messing with this a week or so ago.