Boot failed using the latest branch of LiteX with Vexiiriscv

[looongbin]

Hi,

When using the latest branches of LiteX and VexiiRiscv to boot Linux, the kernel gets stuck during the boot process. The log is shown below:

[    0.017769] Console: colour dummy device 80x25
[    0.021388] Calibrating delay loop (skipped), value calculated using timer frequency.. 200.00 BogoMIPS (lpj=400000)
[    0.031399] pid_max: default: 32768 minimum: 301
[    0.040704] Mount-cache hash table entries: 4096 (order: 3, 32768 bytes, linear)
[    0.047326] Mountpoint-cache hash table entries: 4096 (order: 3, 32768 bytes, linear)
[    0.095378] riscv: ELF compat mode unsupported
[    0.095537] ASID allocator disabled (0 bits)
[    0.105637] rcu: Hierarchical SRCU implementation.
[    0.109150] rcu:     Max phase no-delay instances is 1000.
[    0.123364] EFI services will not be available.
[    0.129121] smp: Bringing up secondary CPUs ...
[    0.132499] smp: Brought up 1 node, 1 CPU
[    0.137727] Memory: 1976548K/2097152K available (11862K kernel code, 5902K rwdata, 6144K rodata, 2435K init, 388K bss, 117356K reserved, 0K cma-reserved)
[    0.159327] devtmpfs: initialized
[    0.185774] clocksource: jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 7645041785100000 ns
[    0.194504] posixtimers hash table entries: 512 (order: 1, 8192 bytes, linear)
[    0.201841] futex hash table entries: 256 (16384 bytes on 1 NUMA nodes, total 16 KiB, linear).
[    0.221635] DMI not present or invalid.
[    0.233095] NET: Registered PF_NETLINK/PF_ROUTE protocol family
[    0.242689] DMA: preallocated 256 KiB GFP_KERNEL|GFP_DMA32 pool for atomic allocations
[    0.249808] audit: initializing netlink subsys (disabled)
[    0.263654] thermal_sys: Registered thermal governor 'step_wise'
[    0.264423] audit: type=2000 audit(0.000:1): state=initialized audit_enabled=0 res=1
[    0.277350] cpuidle: using governor menu
[    0.284440] SBI misaligned access exception delegation ok

**The system stops at this point and no further output is produced.**

My build command is:

MyBoard.py --build --update-repo latest --cpu-type vexiiriscv --cpu-variant debian --with-coherent-dma --no-netlist-cache --with-sdcard --libc-mode full

Has anyone encountered the same issue or knows how to resolve it?

[chillakalyan]

Hi,

I attempted to reproduce this issue while investigating the LiteX + VexiiRiscv Linux boot flow.

Environment I used:

• LiteX (latest master)
• VexiiRiscv (latest)
• riscv64-unknown-elf toolchain
• Linux build targeting VexiiRiscv
• Host: WSL Ubuntu

While setting up the environment I noticed that several build scripts and generated paths may still reference Windows-mounted paths (e.g. /mnt/c/...) when building under WSL. This can sometimes lead to inconsistent execution of scripts during the software build phase (picolibc / BIOS generation).

Because of this, I am currently rebuilding the environment fully inside the Linux filesystem (~/litex) to eliminate potential WSL filesystem issues.

So far I have observed that the boot log proceeds normally through:

• SMP initialization
• memory initialization
• devtmpfs setup
• NETLINK registration
• SBI misaligned access delegation

and then stops after:

SBI misaligned access exception delegation ok

This suggests the system may already be transitioning into later kernel initialization stages when the hang occurs.

I am still working on reproducing the issue consistently and will share additional observations once I can run the full Linux boot with the same configuration (--cpu-type vexiiriscv --cpu-variant debian).

If anyone has additional debug logs (especially with earlycon or additional kernel debug flags), that might help narrow down where the kernel stops progressing.

Thanks.

[KunoTV]

I had the exactly the same problem myself.
The issue occurs in the Linux Kernel arch -> riscv -> kernel -> unaligned_access.speed.c within the function check_unaligned_access(). There is a while loop that pauses as long as the jiffies remain the same: while ((now = jiffies) == start_jiffies), and the problem is that it gets stuck there.
My solution was to add the following lines to linux.config (VexiiRiscv->doc->litex->buildroot->board->litex_soc):

# Disable Misaligned
CONFIG_RISCV_MISALIGNED=n
CONFIG_RISCV_PROBE_UNALIGNED_ACCESS=n

Hope this helps.
Best regards, Alex

[looongbin]

@KunoTV
It works, thanks for your help!!!!

However, it still gets stuck there for about 10 seconds. Previously, when I booted Linux, I didn’t need to set this option, and there was no such delay.
Maybe there is a new issue with the hardware?