[Letux-kernel] SMP issue between LX16 and LX20 found...

[Paul Boddie]

On Sunday, 15 June 2025 12:49:05 CEST H. Nikolaus Schaller wrote:
> 
> [    4.063505] Reserved instruction in kernel code[#1]:

So, I configured the kernel to use SMP and experienced something similar to 
this. The call stack was this:

[   18.028821] [<801699a8>] vm_normal_page+0x50/0x108
[   18.028128] [<8016ce08>] do_wp_page+0x68/0xe14
[   18.029818] [<8016faac>] handle_mm_fault+0x788/0x1070
[   18.029381] [<8003448c>] __do_page_fault+0x10c/0x570
[   18.028861] [<80039470>] tlb_do_page_fault_1+0x114/0x124

Admittedly, I did switch on virtualisation features, but this could just be a 
bad instruction appearing in an instruction stream due to cache/concurrency 
issues.

I also got a crash that seemed to involve a bad virtual address of zero with 
this trace:

[  783.542595] [<80704fbc>] sdhci_prepare_dma+0x61c/0x6e0
[  783.544980] [<8070571c>] sdhci_send_command+0x2f4/0x66c
[  783.544717] [<80705ae0>] sdhci_send_command_retry+0x4c/0x190
[  783.544889] [<80705ca4>] sdhci_request+0x80/0x104

This seems like a concurrency issue, too, at first glance.

Without SMP enabled, the system seems fairly robust. I installed QEMU and then 
realised that I would never have enough memory to run QEMU+KVM, anyway. But 
those virtualisation features might still be usable in L4Re, so they might not 
go to waste!

Paul