Page 109 of 139
Re: New kernels
Posted: Fri Feb 28, 2025 4:45 pm
by Hypex
xeno74 wrote: Thu Feb 27, 2025 6:41 pm
I don't have really much knowledge about patching the dtb during the boot.
I'm still learning myself but writing the X1 loader taught me some quirks about the OF API works.
Maybe this can help but I am not sure if it is correct:
I'll make some comments.
Code: Select all
+ prom_setprop(iob, NULL, "#address-cells", "3", sizeof(3));
+ prom_setprop(iob, NULL, "#size-cells", "2", sizeof(2));
+ prom_setprop(iob, NULL, "#interrupt-cells", "1", sizeof(1));
What do you think?
That's a good start. I'm not sure if it will reject properties that don't exist as it needs to add them but it should work fine. The only issue is the hash values are integers so will need be sent as such. So the following (untested) example may be needed:
Code: Select all
arch/powerpc/kernel/prom_init.c:
/*
* The io-bridge has device_type set to 'io-bridge' change it to 'isa'
* so that generic isa-bridge code can add the SB600 and its on-board
* peripherals.
*/
name = "/pxp@0,e0000000/io-bridge@0";
iob = call_prom("finddevice", 1, 1, ADDR(name));
if (!PHANDLE_VALID(iob))
return;
/* device_type is already set, just change it. */
prom_printf("Changing device_type of SB600 node...\n");
prom_setprop(iob, name, "device_type", "isa", sizeof("isa"));
+ u32 addresss_cells = 3, size_cells =2, interrupt_cells = 1;
+ prom_setprop(iob, name, "#address-cells", &address_cells, sizeof(u32));
+ prom_setprop(iob, name, "#size-cells", &size_cells, sizeof(u32));
+ prom_setprop(iob, name, "#interrupt-cells", &interrupt_cells, sizeof(u32));
Though I think it would be better to pull the cell sizes from the parent pxp device, rather than use direct values.
Re: New kernels
Posted: Fri Feb 28, 2025 5:48 pm
by xeno74
xeno74 wrote: Fri Feb 28, 2025 10:39 am
Hi All,
I created a
patch because of the missing properties today.
After that, I compiled a new RC4 with this patch.
Download:
v6.14.0-rc4-2
Please test this kernel.
If it boots then please post the output of
dtc -I dtb /sys/firmware/fdt.
Thanks,
Christian
Hypex,
Thank you for your comments. Could you also please test the kernel
v6.14.0-rc4-2?
After that, I will replace my code with yours.
Thanks,
Christian
Re: New kernels
Posted: Sat Mar 01, 2025 8:20 am
by xeno74
Hypex wrote: Fri Feb 28, 2025 4:45 pm
That's a good start. I'm not sure if it will reject properties that don't exist as it needs to add them but it should work fine. The only issue is the hash values are integers so will need be sent as such.
I tested the kernel 6.14.0-rc4-2 today and you're right with that's a good start.
The output of missing properties is gone. That means, we can patch the dtb during the boot.
But you're also right with sending the hash values as integers.
Currently the values are:
Code: Select all
io-bridge@0 {
#interrupt-cells = <0x31000000>;
#size-cells = <0x32000000>;
#address-cells = <0x33000000>;
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>;
The values are incorrect.
Re: New kernels
Posted: Sat Mar 01, 2025 11:30 am
by xeno74
Hypex,
Your code works! I created a new patch and compiled a new RC4 with this patch today.
Download and further information:
v6.14.0-rc4-3
@All
Please test this kernel.
Thanks,
Christian
Re: New kernels
Posted: Sat Mar 01, 2025 12:25 pm
by Hypex
xeno74 wrote: Sat Mar 01, 2025 11:30 am
Hypex,
Your code works! I created a new patch and compiled a new RC4 with this patch today.
Great!
Sorry I had to dog sit over the week end so away from my X1000. Back tonight. Will also test the patch for interest but did the rc4-2 crash?
Another way would be add Nemo to the excluded default cells list like below. But it would silence the warning and use default values which may still be faulty. So I think your solution of fixing up the tree to be compliant is better.
https://github.com/torvalds/linux/commi ... 72fb3bc895
Re: New kernels
Posted: Sat Mar 01, 2025 1:50 pm
by xeno74
Hypex wrote: Sat Mar 01, 2025 12:25 pm
Will also test the patch for interest but did the rc4-2 crash?
No, it didn't crash.
Re: New kernels
Posted: Sun Mar 02, 2025 9:55 am
by musa
Hi
vmlinux-6.14_rc4-2 and vmlinux-6.14_rc4-3 running on my system.
Re: New kernels
Posted: Sun Mar 02, 2025 12:24 pm
by xeno74
musa wrote: Sun Mar 02, 2025 9:55 am
Hi
vmlinux-6.14_rc4-2 and vmlinux-6.14_rc4-3 running on my system.
Great! Thanks a lot for testing!
Re: New kernels
Posted: Sun Mar 02, 2025 3:50 pm
by xeno74
Re: New kernels
Posted: Mon Mar 03, 2025 3:32 am
by Hypex
xeno74 wrote: Sat Mar 01, 2025 1:50 pm
No, it didn't crash.
Confirmed. Ran both patches. Seems to work fine and soon onto the RC5.