20 Sep 2024 |
Tristan Ross | I think I forgot to change the device tree because lsblk doesn't seem to mention my NVME | 19:29:24 |
Alex | The physical board has the version number written on it.
Only early buyers have a 1.2a board. | 19:33:37 |
Tristan Ross | Oh, I have a case on it so getting it open requires me to shut it down and disconnect everything | 19:34:11 |
Tristan Ross | It's a 1.3b | 19:36:07 |
Alex | Which kernel are you using? | 19:36:24 |
Tristan Ross | 6.10.10 | 19:36:39 |
Alex | Is that mainline? | 19:36:56 |
Tristan Ross | Idk | 19:37:01 |
Tristan Ross | It's whatever Nick's flake uses | 19:37:08 |
Alex | Looks like mainline then.
You'll need to update to 6.11, which has PCIe merged. | 19:37:57 |
Tristan Ross | Alright | 19:38:09 |
Alex | 6.10 can't see the NVMe disk because it doesn't have PCIe support (NVMe uses PCIe). | 19:38:22 |
Tristan Ross | Fun | 19:38:32 |
Alex | Alternatively, you can find the patches on the LKML and apply them onto 6.10. | 19:38:44 |
Tristan Ross | Hopefully github:NickCao/nixpkgs/f0e01e7ec90964913845b818ebe3daf0ca11fc19 has that kernel | 19:39:07 |
Tristan Ross | Yep, it's 6.11 | 19:39:35 |
Tristan Ross | Simple nix flake update | 19:39:56 |
Tristan Ross | Oh, I don't see it still | 20:26:55 |
Tristan Ross | $ lsmod | grep pci
xhci_pci 102400 0
xhci_pci_renesas 65536 1 xhci_pci
pcie_starfive 49152 0
phy_jh7110_pcie 32768 2
$ lsmod | grep nvme
| 20:30:40 |
Tristan Ross | $ ls /sys/class/pci_bus/
0000:00 0000:01 0001:00 0001:01
| 20:34:48 |
Alex | To double check, are you sure you have an NVMe M.2 SSD and not a SATA M.2 SSD? | 20:36:02 |
Tristan Ross | Idk | 20:40:05 |
Alex | On my system (vendor kernel, not 6.11), it shows up under 0001:00. | 20:40:07 |
Tristan Ross | It's a Samsung 970 Evo Plus | 20:40:24 |
Tristan Ross | Says V-NAND SSD | 20:40:34 |
Tristan Ross | In reply to @alex:tunstall.xyz On my system (vendor kernel, not 6.11), it shows up under 0001:00. What branch of the vendor kernel are you using? I'm building the kernel right now and I have this:
boot.kernelPackages = let
starfive_pkg = { lib, fetchurl, buildLinux, ... }@args:
buildLinux (args // rec {
version = "6.10.0";
modDirVersion = version;
src = fetchurl {
url = "https://github.com/starfive-tech/linux/archive/37087c6cd7f818990c3c37d2e3cb01b05fc8bb69.tar.gz";
hash = "sha256-n3By/goN6+SEjd7a8zirhqTZyb6qvqZPDb9TYd4riQI=";
};
kernelPatches = [];
structuredExtraConfig = with lib.kernel; {
PINCTRL_STARFIVE_JH7110_SYS = yes;
SERIAL_8250_DW = yes;
};
extraMeta.branch = "esmil_starlight";
} // (args.argsOverride or {}));
in pkgs.recurseIntoAttrs (pkgs.linuxPackagesFor (pkgs.callPackage starfive_pkg {}));
I went based on activity on GitHub.
| 21:32:50 |
Alex | The upstream branch.
https://github.com/AlexandreTunstall/nixos-riscv/blob/df3a550f99766ae7bfe88449494648cdb5f65e53/flake.nix#L5 | 21:34:09 |
Tristan Ross | Alright | 21:34:38 |
Tristan Ross | [ 431.435107] rcu: rcu_sched kthread timer wakeup didn't happen for 100268 jiffies! g4449 f0x2 RCU_GP_WAIT_FQS(5) ->state=0x200
[ 431.450921] rcu: Possible timer handling issue on cpu=0 timer-softirq=1268
[ 431.462411] rcu: rcu_sched kthread starved for 100276 jiffies! g4449 f0x2 RCU_GP_WAIT_FQS(5) ->state=0x200 ->cpu=0
[ 431.477377] rcu: Unless rcu_sched kthread gets sufficient CPU time, OOM is now expected behavior.
[ 431.490929] rcu: RCU grace-period kthread stack dump:
[ 431.500539] task:rcu_sched state:R stack:0 pid:17 ppid:2 flags:0x00000000
[ 431.513528] Call Trace:
[ 431.520563] [<ffffffff80e9db9e>] __schedule+0x37c/0xafc
[ 431.530394] [<ffffffff80e9e376>] schedule+0x58/0xf2
[ 431.539818] [<ffffffff80ea4d2e>] schedule_timeout+0x92/0x18e
[ 431.549966] [<ffffffff800bacc2>] rcu_gp_fqs_loop+0x2f8/0x3fa
[ 431.560048] [<ffffffff800bde90>] rcu_gp_kthread+0x132/0x178
[ 431.569996] [<ffffffff800422be>] kthread+0xe8/0x108
[ 431.579180] [<ffffffff8000411a>] ret_from_fork+0xe/0x20
This isn't good, serial is filled with this.
| 22:26:05 |
Tristan Ross | Alex: I see your repo has jh7110-crypto blacklisted, could that cause this? I see there's console outputs mentioning the module. | 22:43:25 |