Page 107 of 139

Re: New kernels

Posted: Sat Feb 22, 2025 8:19 am
by xeno74
Hypex wrote: Fri Feb 21, 2025 6:23 pm Currently running now. Seems stable. Booted from my loader with ramdisk and it had no issues finding root PCI device info:
Great, that the stable long-term kernel 5.10.234 works with your loader.
Hypex wrote: Fri Feb 21, 2025 6:23 pm Feb 22 03:35:38 ubuntu modprobe: FATAL: Could not read '/lib/modules/5.10.234-powerpc64-smp/kernel/net/bluetooth/bluetooth.ko.xz': Invalid module format
I think, I should not release the kernel with compressed kernel modules.
Hypex wrote: Fri Feb 21, 2025 6:23 pm I also booted the 6.14 RC2 which managed to work. But it struggled as it was waiting over a minute for devices to appear. Then the desktop managed to load. I'm struggling with Fienix with it as it has no desktop. So surviving on menu and shell. But at times I want to open a folder but can't.
I don't have this issue. Thank you for testing the RC2.

Re: New kernels

Posted: Sat Feb 22, 2025 4:32 pm
by Hypex
xeno74 wrote: Sat Feb 22, 2025 8:19 am Great, that the stable long-term kernel 5.10.234 works with your loader.
The LTS kernels are very reliable with my Ubuntu system. :-)
I think, I should not release the kernel with compressed kernel modules.
Well, aside from the BT error in log, I had no issue with it. Seems to function normally with no noticeable slowdown. But, if you like, you could release an XZ compressed kernel image and I can copy it right in. :-D
I don't have this issue. Thank you for testing the RC2.
I'm going to do a test and set up a new partition with a Fresh Fienix. Yes that name will do nicely. It just has too many faults, even though it was a fresh install from last year or close to.

I've just booted and tested the 6.14 RC1 again from CFE. It booted to desktop and swiftly. Even in use it feels fast. But, it still has the missing address and size cells on boot, so still crashes. In addition, the Network Manager keeps crashing and I had no network. So I just tried removing network-manager and reinstalling which worked just then. After booting the RC2 from my loader last night the RC3 is installed and ready to go soon. Also, because the latest kernels are much larger, my boot loader is unable to load them in GZ format from a zImage. Even without ramdisk they are too large. Over 16MB. This is no problem for me, as I personally use XZ format to store kernels, which allows both kernel and ramdisks to be loaded in. But, if anyone wanting to use my boot manager leaves it at default settings, it will use GZ and install a zImage if found, which will then break. AmigaBoot will either stall or immediately break to CFE, depending on version, which all seem to lack size checks. This can be enough for people to give up. I know myself, as if I can't get something to work in a short amount of time, I tend to give up. :-)

Now, I'm still intrigued by the errors in my pxp io-bridge blob. Can I ask you for a favor please? With your fine working kernel can you grab me a DTB dump? My root node and "pxp@0,e0000000" node has #address and #size properties. But my "io-bridge@0" node does not and I wonder if it should. When I Google about missing #address and #size properties my results indicate these are needed for a parent node but not all child nodes. Some excepts from my tree:
sudo dtc -I dtb /sys/firmware/fdt

Code: Select all

/ {
	#interrupt-cells = <0x02>;
	#address-cells = <0x02>;
	#size-cells = <0x02>;
	model = "pasemi,nemo";
	device_type = "bootrom";
	compatible = "pasemi,nemo\0pasemi,pa6t-1682m\0PA6T-1682M\0pasemi,pwrficient\0pasemi";
	platform-open-pic = <0x00 0xfc000000 0x00 0x41000>;
	phandle = <0x7fdff018>;

Code: Select all

	pxp@0,e0000000 {
		interrupt-controller;
		device_type = "pci";
		#interrupt-cells = <0x01>;
		#address-cells = <0x03>;
		#size-cells = <0x02>;
		model = "pa";
		compatible = "pasemi,rootbus\0pa-pxp";
		interrupt-map-mask = <0x00 0x00 0x00 0xff>;
		interrupt-map = <0x00 0x00 0x00 0x30 0x7fe2f6e8 0x30 0x01 0x00 0x00 0x00 0x31 0x7fe2f6e8 0x31 0x01 0x00 0x00 0x00 0x32 0x7fe2f6e8 0x32 0x01 0x00 0x00 0x00 0x33 0x7fe2f6e8 0x33 0x01>;
		bus-range = <0x00 0xff>;
		ranges = <0x00 0x00 0xe0000000 0x00 0xe0000000 0x00 0x10000000 0x1000000 0x00 0x00 0x00 0xfc800000 0x00 0x800000 0x2000000 0x00 0x80000000 0x00 0x80000000 0x00 0x60000000 0x2000000 0x00 0xe0000000 0x00 0xe0000000 0x00 0x100000 0x2000000 0x00 0xfd800000 0x00 0xfd800000 0x00 0x1000 0x2000000 0x800 0x00 0x800 0x00 0x400 0x00>;
		reg = <0x00 0xe0000000 0x00 0x10000000>;
		phandle = <0x7fe2ffa8>;

Code: Select all

		io-bridge@0 {
			device_type = "isa";
			vendor-id = <0x1959>;
			device-id = <0xa001>;
			revision-id = <0x12>;
			class-code = <0x60000>;
			reg = <0x00 0x00 0x00 0x00 0x00 0x82000000 0x00 0xe0000000 0x00 0x2000 0xc2000000 0x00 0xfd800000 0x00 0x1000>;
			assigned-addresses = <0x82000000 0x00 0xe0000000 0x00 0x2000 0xc2000000 0x00 0xfd800000 0x00 0x1000>;
			interrupt-parent = <0x7fe2f6e8>;
			interrupts = <0x56 0x01>;
			compatible = "pasemi,1682m-iob\0pasemi,io-bridge";
			phandle = <0x7fe32728>;
		};
I also see this in dmesg:

Code: Select all

[    0.234573] OF: /localbus@f0000000/cf@1000000: could not find phandle 14
[    0.234599] OF: /localbus@f0000000/cf@1000000: could not find phandle 1
[    0.234619] OF: /localbus@f0000000/cf@1000000: could not find phandle 10
[    0.234648] OF: /localbus@f0000000/cf@1000000: could not find phandle 11

Re: New kernels

Posted: Sun Feb 23, 2025 9:44 am
by xeno74
Hypex wrote: Sat Feb 22, 2025 4:32 pm Can I ask you for a favor please? With your fine working kernel can you grab me a DTB dump?
Is it possible to load another dtb in the CFE?

In my point of view you can't change the dtb like on the X5000. You have to live with your dtb.

Re: New kernels

Posted: Sun Feb 23, 2025 2:15 pm
by Hypex
xeno74 wrote: Sun Feb 23, 2025 9:44 am Is it possible to load another dtb in the CFE?
Not that I know of.
In my point of view you can't change the dtb like on the X5000. You have to live with your dtb.
You can change it? Well, I didn't want to change it, I just wanted to look at yours which works for you. :D

I'm just wondering how mine is becoming corrupted with missing properties, or, why the kernel thinks they are missing and sets off recoverable traps. I get the fault regardless of launching the kernel from CFE or my loader. I've suspected I could have bad RAM again, though it's not that old. But the 5.x kernels boot fine including the latest ones you compiled. Do these have the DTB changes for fixing up the Nemo PCI bus type?

It could even be a corruption on my Fienix volume. But the message looks like it indicated it was kernel alone when it was about missing DTB properties. AFAIK these would be read in from the table it created above kernel in memory. Would like to know where in kernel source it's getting stuck on this. :-?

Re: New kernels

Posted: Sun Feb 23, 2025 3:52 pm
by xeno74
Hypex,

Here is my Nemo DTB dump: nemo.dts

Cheers,
Christian

Re: New kernels

Posted: Sun Feb 23, 2025 4:03 pm
by xeno74
Hypex wrote: Sat Feb 22, 2025 4:32 pm With your fine working kernel can you grab me a DTB dump? My root node and "pxp@0,e0000000" node has #address and #size properties. But my "io-bridge@0" node does not and I wonder if it should.
From my nemo.dts:

Code: Select all

io-bridge@0 {
			device_type = "isa";
			vendor-id = <0x1959>;
			device-id = <0xa001>;
			revision-id = <0x12>;
			class-code = <0x60000>;
			reg = <0x00 0x00 0x00 0x00 0x00 0x82000000 0x00 0xe0000000 0x00 0x2000 0xc2000000 0x00 0xfd800000 0x00 0x1000>;
			assigned-addresses = <0x82000000 0x00 0xe0000000 0x00 0x2000 0xc2000000 0x00 0xfd800000 0x00 0x1000>;
			interrupt-parent = <0x7fe2f6e8>;
			interrupts = <0x56 0x01>;
			compatible = "pasemi,1682m-iob\0pasemi,io-bridge";
			phandle = <0x7fe32728>;
		};

Re: New kernels

Posted: Mon Feb 24, 2025 5:08 am
by Hypex
xeno74 wrote: Sun Feb 23, 2025 3:52 pm Hypex,

Here is my Nemo DTB dump: nemo.dts
Great, thanks! So from your nemo.dts I don't see any difference in io-bridge@0. I also found where in source it is coming from:
https://github.com/torvalds/linux/blob/ ... base.c#L95

Not sure why it's crashing. Do we need to be excluded on the list? So on your X1000 you don't see the warning nor a crash? :-?

Looking a bit more it has a default 1 cell for address and size. But, according to the below spec it should be 2 for address. Since the PASemi is 64 bit it should be using 2 for address. But, IMHO, it shouldn't use any defaults since it's meant to exactly describe hardware. The root node contain this info so why doesn't it grab it from there? On top of that the pxp@0 parent device has these properties so why isn't using them?

https://devicetree-specification.readth ... asics.html

Anyway, I booted the 6.14 RC3 a few times. One time it booted to desktop. Another time terminal. I just tested it with Ubuntu MATE desktop and it booted fine to desktop. In all boots, regardless of desktop working or not, is missing warning and crashes. I don't get it. :-?

I booted my Ubuntu Precise with kernel 5.10.234. But found my system was mounted read only. Don't why what happened there. Perhaps XZ modules are causing havoc. Rebooted and fsck fixed next boot. Checked all other filesystems with no errors. Ran Disk Utility and checked SMART data. No issues detected. Ran a small test on SSD. No issues. Checked my HDD. Not even a bad block.

Be good to check ram next. Need a PPC ram checker. Or perhaps pull ram out and give a wipe over with some isoprophyl in case there is a build up. :-)

Re: New kernels

Posted: Mon Feb 24, 2025 1:49 pm
by xeno74
Hypex wrote: Mon Feb 24, 2025 5:08 am Be good to check ram next. Need a PPC ram checker.
ramsmp

Re: New kernels

Posted: Mon Feb 24, 2025 2:16 pm
by xeno74
Hi All,

Here is the RC4 of kernel 6.14 for testing.

Download and further information: github.com

BrassMonkey works great with the RC4 and Adélie.

You can see here the HTML editor including a good working spell checker.

Image

Please test the kernels.

Thanks,
Christian

Re: New kernels

Posted: Mon Feb 24, 2025 4:03 pm
by Hypex
xeno74 wrote: Mon Feb 24, 2025 1:49 pm ramsmp
Thanks.

So ran through all tests I could see. No issues reported. Reported results to thread including floats. Also running memtester now with no issues found yet.