Re: kernel loading

[Robert Swindells <rjs%fdy2.co.uk@localhost>]

Nick Hudson <nick.hudson%gmx.co.uk@localhost> wrote:
> On 11 Nov 2018, at 11:01, Robert Swindells <rjs%fdy2.co.uk@localhost> wrote:
>> 
>> Nick Hudson <nick.hudson%gmx.co.uk@localhost> wrote:
>>>> On 10/11/2018 19:45, Robert Swindells wrote:
>>>> 
>>>> How is loading and running of a 32-bit kernel by u-boot expected to
>>>> work ?
>>>> 
>>>> I have copied std.generic and mk.generic to use with fdt zynq sources.
>>>> 
>>>> The kernel_noload flag to mkubootimage(1) looks to make u-boot execute
>>>> the kernel in place. What address should I load netbsd.ub at ?
>>> 
>>> your uboot "should" have a kernel_addr_r and a fdt_addr_r variable for 
>>> where it'll load the kernel and FDT.  The kernel should be aligned to a 
>>> least L1_S_SIZE (10x00100000).  I guess your old u-boot doesn't.
>>> 
>>> http://src.illumos.org/source/xref/netbsd-src/sys/arch/arm/include/arm32/pte.h#100
>>> 
>>> The netbsd.ub is loaded at the aligned address with the header and the 
>>> code is straight after the header, ie. at an offset of 64 bytes.
>> 
>> Ok, thanks.
>> 
>> I have got further by using -T kernel for the u-boot image with it
>> linked at the second L1 boundary.
>> 
>Your uboot doesn’t understand kernel_noload?

I seems to treat kernel_noload as an instruction to not copy the kernel
to the load address that is in the header.

I tried explicitly loading netbsd.ub at 0x0 but that caused the panic
from arm_bad_vtopdiff.

>> Maybe we will need a PARALLELLA kernel as well as ZYNQ or GENERIC with
>> them built from the same sources but linked at different addresses.
>
>I’ll look to see if there’s a way to get a kernel_noload and a kernel
>uboot image with a best fit load address.

Which comes back to my question on what is supposed to happen at boot
time.

The kernel that I have now is booting fine.