Re: RPI4: fail too boot with an initrd

[LABBE Corentin <clabbe@baylibre.com>]

On Mon, Feb 17, 2020 at 11:50:04AM +0100, Matthias Brugger wrote:
> 
> 
> On 17/02/2020 11:37, LABBE Corentin wrote:
> > On Fri, Feb 14, 2020 at 06:15:27PM +0000, James Morse wrote:
> >> Hi Corentin,
> >>
> >> On 14/02/2020 13:27, LABBE Corentin wrote:
> >>> Since the inclusion of the "enable network support in RPi4 config" serie on uboot, I
> >>> have started to work on adding the rpi4 in kernelCI.
> >>> But I fail to succeed in using a kernel/dtb/ramdisk downloaded via tftp.
> >>>
> >>> Using booti I hit:
> >>> [    0.000000] Linux version 5.6.0-rc1-next-20200212 (clabbe@build2-bionic-1804) (gcc version 7.4.1 20181213 [linaro-7.4-2019.02 revision 56ec6f6b99cc167ff0c2f8e1a2eed33b1edc85d4] (Linaro    GCC 7.4-2019.02)) #66 SMP PREEMPT Wed Feb 12 10:14:20 UTC 2020
> >>> [    0.000000] Machine model: Raspberry Pi 4 Model B
> >>> [    0.000000] earlycon: uart0 at MMIO32 0x00000000fe215040 (options '')
> >>> [    0.000000] printk: bootconsole [uart0] enabled
> >>> [    0.000000] efi: Getting EFI parameters from FDT:
> >>> [    0.000000] efi: UEFI not found.
> >>
> >> So no EFI,
> >>
> >>> [    0.000000] OF: reserved mem: failed to allocate memory for node 'linux,cma'
> >>
> >> Out of memory.
> >>
> >>> [    0.000000] cma: Failed to reserve 32 MiB
> >>> [    0.000000] Kernel panic - not syncing: Failed to allocate page table page
> >>
> >> Out of memory...
> >>
> >>> [    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 5.6.0-rc1-next-20200212 #66
> >>> [    0.000000] Hardware name: Raspberry Pi 4 Model B (DT)
> >>> [    0.000000] Call trace:
> >>> [    0.000000]  dump_backtrace+0x0/0x1a0
> >>> [    0.000000]  show_stack+0x14/0x20
> >>> [    0.000000]  dump_stack+0xbc/0x104
> >>> [    0.000000]  panic+0x16c/0x37c
> >>> [    0.000000]  early_pgtable_alloc+0x30/0xa0
> >>
> >> ... really early!
> >>
> >>> [    0.000000]  __create_pgd_mapping+0x36c/0x588
> >>> [    0.000000]  map_kernel_segment+0x70/0xa4
> >>> [    0.000000]  paging_init+0xf4/0x528
> >>> [    0.000000]  setup_arch+0x250/0x5d8
> >>> [    0.000000]  start_kernel+0x90/0x6d8
> >>>
> >>>  
> >>> Since the same kernel boot with bootefi and that bootefi lack ramdisk address,
> >>
> >> Booting with EFI will cause linux to use the EFI memory map.
> >>
> >> Does your DT have a memory node? (or does it expect EFI to provide the information)
> >>
> >>
> >>> I tried to add the address in the dtb via:
> >>> fdt addr 0x02400000; fdt resize; fdt set /chosen linux,initrd-start 0x02700000; fdt set /chosen linux,initrd-end 0x10000000; bootefi 0x00080000 0x02400000
> >>> But with that, I get:
> >>> initrd not fully accessible via the linear mapping -- please check your bootloader ...
> >>
> >> So this one is an EFI boot, but you can't find where to put the initramfs such that the
> >> kernel agrees its in memory.
> >>
> >> If you boot with 'efi=debug', linux will print the EFI memory map. Could you compare that
> >> to where U-Boot thinks memory is?
> >>
> >> (it sounds like your DT memory node is missing, and your EFI memory map is surprisingly small)
> > 
> > Hello
> > 
> > Thanks for your advices.
> > 
> > In the dtb of mainline linux:
> > 	/* Will be filled by the bootloader */
> > 	memory@0 {
> > 		device_type = "memory";
> > 		reg = <0 0 0>;
> > 	};
> > 
> > In uboot I have:
> > static struct mm_region bcm2711_mem_map[] = {
> >         {
> >                 .virt = 0x00000000UL,
> >                 .phys = 0x00000000UL,
> >                 .size = 0xfe000000UL,
> >                 .attrs = PTE_BLOCK_MEMTYPE(MT_NORMAL) |
> >                          PTE_BLOCK_INNER_SHARE
> >         }, {
> >                 .virt = 0xfc000000UL,
> >                 .phys = 0xfc000000UL,
> >                 .size = 0x03800000UL,
> >                 .attrs = PTE_BLOCK_MEMTYPE(MT_DEVICE_NGNRNE) |
> >                          PTE_BLOCK_NON_SHARE |
> >                          PTE_BLOCK_PXN | PTE_BLOCK_UXN
> >         }, {
> >                 /* List terminator */
> >                 0,
> >         }
> > };
> > But I dont know if uboot use that for filling the memory node.
> 
> No it doesn't. U-Boot uses the DT from the firmware and passes this to the
> kernel. But it seems you pass instead your own device-tree to the kernel, so you
> will need to update the memory node to show the available memory on you board.
> 

I dont understand, in the Linux commit "ARM: dts: Add minimal Raspberry Pi 4 support" I read:
The RPi 4 is available in 3 different variants (1, 2 and 4 GB RAM), so leave the memory size to zero and let the bootloader take care of it.
But if uboot dont fill that...
So the DTB in mainline is wrong, right ?

Regards

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel