RE: When iso install fails to a memory-map modified virt machine, how can I debug it..

[ckim]

Hi, all,

 

In my virtual machine, which is a ‘address map modified’ and a ‘peripheral added’ version of aarch64 virt machine, when I start ubuntu install, I see this message below.

 

Booting a command list

 

EFI stub: Booting Linux Kernel...

EFI stub: EFI_RNG_PROTOCOL unavailable, no randomness supplied

EFI stub: Using DTB from configuration table

EFI stub: Exiting boot services and installing virtual address map...

 

Because it’s just an iso image, I can’t see the kernel source in debug mode. (I’ll try with deboostrap and chroot, hoping to be able to see )

In what condition can the machine be?

Any suggestion will be really appreciated.

 

Chan Kim

 

From: ckim@etri.re.kr <ckim@etri.re.kr>
Sent: Thursday, June 17, 2021 12:13 PM
To: 'qemu-discuss' <qemu-discuss@nongnu.org>
Subject: When iso install fails to a memory-map modified virt machine, how can I debug it..

 

Hello list members,

 

I need a help from experts here again. Thank you for the supports.

I can install an ubuntu iso file on qemu arm “virt” machine. (following https://futurewei-cloud.github.io/ARM-Datacenter/qemu/how-to-launch-aarch64-vm/)

But when I tried to install it on a modified version of “virt”, it didn’t work.

So I tried two simple tests just by modifying the address map.

 

( Test 1 )

First I changed part of the address map slightly,

(I reduced the gic redistributor memory to a smaller value, I don’t need so many number of cores. And I moved uart to another empty location)

 

[VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },

[VIRT_UART] =               { 0x09000000, 0x00001000 },

è

[VIRT_GIC_REDIST] =         { 0x080A0000, 0x00800000 },

[VIRT_UART] =               { 0x088B0000, 0x00001000 },

 

With this changes, the ubuntu seemed to be being installed (I didn’t check until the end)

 

( Test 2 )

I further changed the address map. (adding to test 1)

(the PCI was moved from (256MB ~ 1GB) to (768MB ~ 1536MB). And now there is a gap from 256MB to 768MB))

[VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },

[VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },

[VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },

/* Actual RAM size depends on initial RAM and device memory settings */

[VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },

è

[VIRT_PCIE_MMIO] =          { 0x30000000, 0x2eff0000 },

[VIRT_PCIE_PIO] =           { 0x5eff0000, 0x00010000 },

[VIRT_PCIE_ECAM] =          { 0x5f000000, 0x01000000 },

/* Actual RAM size depends on initial RAM and device memory settings */

[VIRT_MEM] =                { 0x60000000, LEGACY_RAMLIMIT_BYTES },

 

But this time, the install hangs somewhere in the early stage. It enters qemu_main_loop.

 

I think I have two questions here.

1. If I change the address map of “virt” machine, should it install ubuntu with no problem? (I understand the “virt” machine takes care of fdt and acpi)
2. If I have this kind of problem (hang during the installing with iso file), is there an effective way of debuggig? I found from an answer in this question (https://askubuntu.com/questions/161194/how-to-build-an-orignal-ubuntu-installer-live-iso-image) I can use debootstrap and chroot to make the iso image. I’m not sure if I can use debugger using the generated files in the iso. I’ll look into modifying iso file as advised in the answer.

 

Thanks for any help!

 

Chan Kim