FPU disabled on VCU-118

NOEL-V
1

Greetings. I have been trying to synthesize NOEL-V in the Xilinx VCU-118. I have reached a functional implementation, with GRMON successfully reporting the system information.

Unfortunately, I have not been able to get Linux to boot. In some attempts at debugging, I found it might be cause of the FPU being missing. This agrees with GRMON’s “float” command reporting “FPU is disabled!” and “reg” reporting the same.

I have seen that nanofpunv should be instantiated, so I don’t understand why this might be a problem.
One last resort would be trying to synthesize GRFPU instead.

I would appreciate any suggestions.
Daniel.

2

You can use GRMON command ‘info sys’ to print a detailed report, check if D & F extension is supported by hardware.

  dm0        Frontgrade Gaisler  RISC-V Debug Module
             AHB Master 9
             AHB: f6000000 - f7000000
             hart0: DXLEN 64, MXLEN 64, SXLEN 64, UXLEN 64
                    ISA A C D F H I M X,  Modes M S U
                    Stack pointer 0xbffffff0
                    icache 4 * 4 kB, 32 B/line, dir
                    dcache 4 * 4 kB, 32 B/line, dir
                    5 triggers,
                    itrace 64 lines
3

The ‘float’ and ‘reg’ command can not show the FPU registers, maybe this is a GRMON bug, I will feedback this issue in another post.

But it doesn’t affect Linux boot.

4

Thank you for your reply. In fact, this seems to be the case, as info sys does adequately report that the target supports floating point extensions:

dm0        Frontgrade Gaisler  RISC-V Debug Module
             AHB: fe000000 - ff000000
             hart0: DXLEN 64, MXLEN 64, SXLEN 64, UXLEN 64
                    ISA A C D F H I M P,  Modes M S U
                    Stack pointer 0x7ffffff0
                    icache 4 * 4 kB, 32 B/line, dir
                    dcache 4 * 4 kB, 32 B/line, dir
                    3 triggers,
                    itrace 64 lines

Unfortunately, then it seems that other problem is causing Linux not to boot, as I am receiving a signal 4 (SIGILL), which I thought was caused by this.
Do you have any pointer to where I should be looking? Maybe I got the device tree wrong? How can I debug further?
Thank you again for your response.

Kernel panic log:

[   24.722096] debug_vm_pgtable: [debug_vm_pgtable         ]: Validating architecture
[   24.880624] Freeing unused kernel image (initmem) memory: 9584K
[   24.917968] Run /init as init process
[   24.959391] init[1]: unhandled signal 4 code 0x1 at 0x0000003f99feece0 in ld-linux-
[   25.028080] CPU: 0 PID: 1 Comm: init Not tainted 6.3.0 #2
[   25.064813] Hardware name: noel-min64-sc (DT)
[   25.092591] epc : 0000003f99feece0 ra : 0000003f99fe88e8 sp : 0000003ffeced2e0
[   25.138698]  gp : ffffffff81b00400 tp : 0000000000000000 t0 : 0000000000000018
[   25.183207]  t1 : 0000003f99fd7dac t2 : 0000000000000008 s0 : 0000003f99ff9208
[   25.228968]  s1 : 0000003f99ff92b0 a0 : 0000003ffeced318 a1 : 0000000000000000
[   25.274545]  a2 : 0000003ffeced518 a3 : 2424242424242424 a4 : 0000000000000024
[   25.321273]  a5 : 0000003ffeced308 a6 : 870303010103030f a7 : 24160a4b570a5248
[   25.367980]  s2 : 0000003f99ff92b0 s3 : 0000003ffeced480 s4 : 0000000000000000
[   25.412377]  s5 : 0000000000000000 s6 : 0000000000000001 s7 : 0000002ae16ebe49
[   25.458122]  s8 : 0000003ffeced480 s9 : 0000002ae17d7de0 s10: 0000003ffeced480
[   25.504187]  s11: 0000003ffeced480 t3 : 0000003f99fe8870 t4 : 00000000ffffffff
[   25.548393]  t5 : 0000003ffeced818 t6 : 0000003ffeced814
[   25.581627] status: 0000000200000020 badaddr: 0000000006853827 cause: 0000000000000002
[   25.634056] Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004
[   25.681189] CPU: 0 PID: 1 Comm: init Not tainted 6.3.0 #2
[   25.715322] Hardware name: noel-min64-sc (DT)
[   25.742104] Call Trace:
[   25.757154] [<ffffffff800053be>] dump_backtrace+0x1c/0x24
[   25.791105] [<ffffffff8086b450>] show_stack+0x2c/0x38
[   25.822299] [<ffffffff808766ca>] dump_stack_lvl+0x3c/0x54
[   25.855395] [<ffffffff808766f6>] dump_stack+0x14/0x1c
[   25.886001] [<ffffffff8086b7ea>] panic+0x106/0x29e
[   25.915481] [<ffffffff800133f0>] do_exit+0x73c/0x742
[   25.945857] [<ffffffff8001356a>] do_group_exit+0x24/0x70
[   25.978072] [<ffffffff8001df9c>] get_signal+0x694/0x6e6
[   26.009607] [<ffffffff800046ba>] do_work_pending+0xfc/0x3e2
[   26.058108] [<ffffffff8000378a>] resume_userspace_slow+0x8/0xa
[   26.096256] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004 ]---
5

I guess it’s a configuration mismatch between Buildroot and Linux.

Buildroot: Target options > Target Architecture Variant > [*] General purpose (G)
Buildroot: Target options > Target ABI   > [*] lp64d

Linux: General setup > Platform type > -*- Emit compressed instructions when building Linux
Linux: General setup > Platform type > [*] FPU support
6

Hello. Those options are already present in the buildroot configuration. In fact, even the prebuilt noel64 images panic with the same exact error. I am unsure what could be causing this panic, as it does not seem to depend on the buildroot/linux kernel configuration.

7

How did your DTS describe it.

riscv,isa = "rv64imafdc";
8

Hello, it describes it as “g”, general purpose, although I have also tried using your extensions with the same (un)success.