Gentoo installation issue

[bernie]

I have an NVIDIA GT710 that I'd like to get working with my Blackbird. After getting past the boot issue by disabling the nouveau module in Petitboot, I still can't get the card to work in the OS. I tried Fedora, Debian and Trisquel. Then, realizing that maybe the 4KB page size issue was affecting me, I decided to try Gentoo. I tried the Distribution (pre-built) kernel, but this also has a 64KB page size, and failed with the same errors as the others. So I compiled a kernel with the options that I thought best, but it won't boot with that kernel. The console output shows:

Code: [Select]

Run /init as init process
init[1]: illegal instruction (4) at 3fff894c8fe0 nip 3fff894c8fe0 lr 3fff894bcbdc code 1 in ld64.so.2[3fff89488000+4e000]
init[1]: code: 7ca32a14 7ca92850 78bfd183 41820084 73ea0001 7c070166 7d2a4b78 38c00010
init[1]: code: 39600020 39800030 381fffff 7fe8fb78 <f0000050> 41820020 2c200000 7c004f98
Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004
CPU: 10 PID: 1 Comm: init Not tainted 6.6.47-gentoo #6
Hardware name: C1P9S01 REV 1.02 POWER9 (raw) 0x4e1203 opal:skiboot-ecb1dc7 PowerNV
Call Trace:
[c0000000021a7a50] [c000000000d6ee58] dump_stack_lvl+0x6c/0x9c (unreliable)
[c0000000021a7a80] [c0000000000da2e4] panic+0x170/0x3ec
[c0000000021a7b20] [c0000000000e3680] do_exit+0xa70/0xa80
[c0000000021a7bf0] [c0000000000e38e4] do_group_exit+0x44/0xc0
[c0000000021a7c30] [c0000000000f7d30] get_signal+0xc50/0xc80
[c0000000021a7d20] [c00000000001c7f0] do_notify_resume+0xf0/0x420
[c0000000021a7dd0] [c000000000028ad8] interrupt_exit_user_prepare_main+0x158/0x1f0
[c0000000021a7e20] [c000000000028d7c] interrupt_exit_user_prepare+0x4c/0x70
[c0000000021a7e50] [c00000000000d444] interrupt_return_srr_user+0x8/0x12c
--- interrupt: f40 at 0x3fff894c8fe0
NIP:  00003fff894c8fe0 LR: 00003fff894bcbdc CTR: 0000000000000000
REGS: c0000000021a7e80 TRAP: 0f40   Not tainted  (6.6.47-gentoo)
MSR:  900000000200f033 <SF,HV,VEC,EE,PR,FP,ME,IR,DR,RI,LE>  CR: 24000420  XER: 00000000
CFAR: c00000000000d55c IRQMASK: 0
GPR00: 0000000000000005 00003fffd6440800 00003fff894eff00 00003fffd6440820
GPR04: 0000000000000000 00000000000001a0 0000000000000010 0000000000000000
GPR08: 0000000000000006 00003fffd6440820 00003fffd6440820 0000000000000020
GPR12: 0000000000000030 0000000000000000 00003fff894e7f10 00003fffd6440fa0
GPR16: 000000000000fff1 0000000000000000 00003fff89488350 00003fff89488000
GPR20: 0000000000000001 0000000000000001 00003fffd6440fa0 0000000080001000
GPR24: 000000007fff9000 0000000000010000 00003fffd6440ac0 00003fffd64410c0
GPR28: 00000000ffffffff 00003fffd6440a20 00003fffd64410f0 0000000000000006
NIP [00003fff894c8fe0] 0x3fff894c8fe0
LR [00003fff894bcbdc] 0x3fff894bcbdc
--- interrupt: f40
Reboot[  148.135496306,5] OPAL: Reboot request...
It appears that something is wrong with ld64, but I have no idea what. How should I proceed from here? I wondered about posting in the Gentoo forums, but decided to try here first.

[witsu]

I think I've seen a similar error before and the issue was not being able to mount the root filesystem
I'd check if you are missing file system or disk controller drivers.

For my Talos I used genkernel to configure and build the kernel and create an initrd, and that worked for me.

Not sure if this is still a problem, but if you are using btrfs, at one point you couldn't use a file system created on a 64k system on a 4k one.

[draconx]

Since it's been some time maybe you've figured all this out already, but...

So I compiled a kernel with the options that I thought best, but it won't boot with that kernel. The console output shows:

Code: [Select]

Run /init as init process
init[1]: illegal instruction (4) at 3fff894c8fe0 nip 3fff894c8fe0 lr 3fff894bcbdc code 1 in ld64.so.2[3fff89488000+4e000]
init[1]: code: 7ca32a14 7ca92850 78bfd183 41820084 73ea0001 7c070166 7d2a4b78 38c00010
init[1]: code: 39600020 39800030 381fffff 7fe8fb78 <f0000050> 41820020 2c200000 7c004f98
Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004
<snip>


The kernel booted just fine and successfully loaded the first userspace application (/init).
This program then crashed when it executed an illegal instruction in ld64.so.2.

The output above shows that the faulting instruction was <f0000050>.  This is a VSX instruction (xxpermdi) which should normally be supported on POWER9, so my guess is that you have booted a kernel with VSX support disabled (CONFIG_VSX=n) but your userspace is compiled to use VSX unconditionally (e.g., built with gcc -mvsx).

[bernie]

Thank you. I had not figured it out. In the meantime, I had taken the config from the dist kernel, changed it to Power9 and 4K page size and have been running with that. I had not disabled VSX myself. However, looking back, I think my error was that I installed the kernel with the instructions in the PPC handbook, not the PPC64 handbook. That included make-ing the pmac32_defconfig target, which I did think suspicious at the time, but went ahead and did anyway, not realising that I was reading the wrong handbook. I'm going to assume for now that that's how VSX became disabled.

Thank you for your help.