22 Commits

Author SHA1 Message Date
b9f2e936de Use correct serial console ttyS1 in fox
All checks were successful
CI / build:cross (pull_request) Successful in 8s
CI / build:all (pull_request) Successful in 2m53s
2026-01-21 13:19:34 +01:00
52caa714d4 Upgrade bench6 and enable MKL by default 2026-01-21 13:15:47 +01:00
9e8c610d8c Add old kernel specialisation in fox
Allow booting an old LTS kernel to compare performance.
2026-01-21 11:04:47 +01:00
dda6a66782 Fix gitea user to allow sending email
All checks were successful
CI / build:cross (pull_request) Successful in 8s
CI / build:all (pull_request) Successful in 16s
CI / build:all (push) Successful in 4s
CI / build:cross (push) Successful in 8s
In order to send email, the gitea user needs to be in the mail-robot
group.

Fixes: #220
Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-20 12:18:52 +01:00
22420e6ac8 Remove unneeded perf package from eudy
It is already included in the base list of packages, which is now only
"perf" and doesn't depend on the kernel version.

Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-20 12:18:49 +01:00
a71cd78b4c Fix infiniband interface names
Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-20 12:18:46 +01:00
e84a2cadbb flake.lock: Update
Flake lock file updates:

• Updated input 'nixpkgs':
    'github:NixOS/nixpkgs/8bb5646e0bed5dbd3ab08c7a7cc15b75ab4e1d0f?narHash=sha256-SqUuBFjhl/kpDiVaKLQBoD8TLD%2B/cTUzzgVFoaHrkqY%3D' (2025-11-30)
  → 'github:NixOS/nixpkgs/3c9db02515ef1d9b6b709fc60ba9a540957f661c?narHash=sha256-2GffSfQxe3sedHzK%2BsTKlYo/NTIAGzbFCIsNMUPAAnk%3D' (2026-01-05)

Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-20 12:18:41 +01:00
d3e43eb651 Remove conflicting definitions in amd-uprof-driver
See: https://lkml.org/lkml/2025/4/9/1709

Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:15:18 +01:00
a491546ffb Mark mcxx as broken and remove from package list
Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:15:14 +01:00
933c78a80b Fix moved package linuxPackages.perf is now perf
Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:15:10 +01:00
150969be9b Fix replaced nixseparatedebuginfod
nixseparatedebuginfod has been replaced by nixseparatedebuginfod2

Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:15:06 +01:00
9097729759 Use standard gcc for intel packages
This reverts 26f52aa27d

Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:15:02 +01:00
779449f1db Fix renamed option watchdog.runtimeTime
The option 'systemd.watchdog.runtimeTime' has been renamed to
'systemd.settings.Manager.RuntimeWatchdogSec'.

Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:14:59 +01:00
6cbe33bd80 Replace wrapGAppsHook with wrapGAppsHook3
Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:14:56 +01:00
3f1f5ae8f2 Fix changed cudaPackages.cuda_cudart output
See: https://github.com/NixOS/nixpkgs/pull/437723
Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:14:49 +01:00
fe8586e780 Set pyproject=true in buildPythonApplication
The buildPythonPackage and buildPythonApplication functions now
  require an explicit format attribute. Previously the default format
  used setuptools and called setup.py from the source tree, which is
  deprecated. The modern alternative is to configure pyproject = true
  with build-system = [ setuptools ].

Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:14:31 +01:00
8677adba27 Fix renamed llvm bintools
Moved from llvmPackages_latest.tools.bintools to
llvmPackages_latest.bintools

Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:14:27 +01:00
f614149edf Upgrade nixpkgs to 25.11
Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-20 12:14:11 +01:00
859eebda98 Change varcila shell to zsh
All checks were successful
CI / build:all (push) Successful in 59m37s
CI / build:cross (push) Successful in 1h27m33s
CI / build:cross (pull_request) Successful in 1h29m20s
CI / build:all (pull_request) Successful in 1h29m22s
Reviewed-by: Rodrigo Arias Mallo <rodrigo.arias@bsc.es>
2026-01-07 13:22:17 +01:00
c2a201b085 Increase fail2ban ban time on each attempt
Some checks failed
CI / build:all (push) Has been cancelled
CI / build:cross (push) Has been cancelled
CI / build:all (pull_request) Successful in 1h38m5s
CI / build:cross (pull_request) Successful in 1h38m3s
Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-07 13:14:34 +01:00
f921f0a4bd Disable password login via SSH in apex
Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-07 13:14:30 +01:00
aa16bfc0bc Enable fail2ban in apex login node
We are seeing a lot of failed attempts from the same IPs:

    apex% sudo journalctl -u sshd -b0 | grep 'Failed password' | wc -l
    2441

Reviewed-by: Aleix Boné <abonerib@bsc.es>
2026-01-07 13:14:22 +01:00
18 changed files with 96 additions and 27 deletions

6
flake.lock generated
View File

@@ -2,11 +2,11 @@
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1764522689,
"narHash": "sha256-SqUuBFjhl/kpDiVaKLQBoD8TLD+/cTUzzgVFoaHrkqY=",
"lastModified": 1767634882,
"narHash": "sha256-2GffSfQxe3sedHzK+sTKlYo/NTIAGzbFCIsNMUPAAnk=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "8bb5646e0bed5dbd3ab08c7a7cc15b75ab4e1d0f",
"rev": "3c9db02515ef1d9b6b709fc60ba9a540957f661c",
"type": "github"
},
"original": {

View File

@@ -57,6 +57,18 @@
};
};
services.fail2ban = {
enable = true;
maxretry = 5;
bantime-increment = {
enable = true; # Double ban time on each attack
maxtime = "7d"; # Ban up to a week
};
};
# Disable SSH login with password, allow only keypair
services.openssh.settings.PasswordAuthentication = false;
networking.firewall = {
extraCommands = ''
# Blackhole BSC vulnerability scanner (OpenVAS) as it is spamming our

View File

@@ -24,7 +24,7 @@
address = "10.0.40.40";
prefixLength = 24;
} ];
interfaces.ibp5s0.ipv4.addresses = [ {
interfaces.ibs785.ipv4.addresses = [ {
address = "10.0.42.40";
prefixLength = 24;
} ];

View File

@@ -4,13 +4,6 @@
# Use the GRUB 2 boot loader.
boot.loader.grub.enable = true;
# Enable GRUB2 serial console
boot.loader.grub.extraConfig = ''
serial --unit=0 --speed=115200 --word=8 --parity=no --stop=1
terminal_input --append serial
terminal_output --append serial
'';
boot.kernel.sysctl = {
"kernel.perf_event_paranoid" = lib.mkDefault "-1";

View File

@@ -139,6 +139,7 @@
openssh.authorizedKeys.keys = [
"ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIKGt0ESYxekBiHJQowmKpfdouw0hVm3N7tUMtAaeLejK vincent@varch"
];
shell = pkgs.zsh;
};
pmartin1 = {

View File

@@ -11,4 +11,11 @@
"console=tty1"
"console=ttyS0,115200"
];
# Enable GRUB2 serial console
boot.loader.grub.extraConfig = ''
serial --unit=0 --speed=115200 --word=8 --parity=no --stop=1
terminal_input --append serial
terminal_output --append serial
'';
}

View File

@@ -1,11 +1,6 @@
{ config, pkgs, lib, ... }:
{ pkgs, lib, ... }:
{
# add the perf tool
environment.systemPackages = with pkgs; [
config.boot.kernelPackages.perf
];
# allow non-root users to read tracing data from the kernel
boot.kernel.sysctl."kernel.perf_event_paranoid" = -2;
boot.kernel.sysctl."kernel.kptr_restrict" = 0;

View File

@@ -3,13 +3,13 @@
{
imports = [
../common/base.nix
../common/xeon/console.nix
../module/amd-uprof.nix
../module/emulation.nix
../module/nvidia.nix
../module/slurm-client.nix
../module/hut-substituter.nix
./wireguard.nix
./serial-console.nix
];
# Don't turn off on August as UPC has different dates.
@@ -19,6 +19,9 @@
# Select the this using the ID to avoid mismatches
boot.loader.grub.device = "/dev/disk/by-id/wwn-0x500a07514b0c1103";
# Increase time so we can boot other entries
boot.loader.timeout = 60;
# No swap, there is plenty of RAM
swapDevices = lib.mkForce [];
@@ -31,7 +34,10 @@
# Use performance for benchmarks
powerManagement.cpuFreqGovernor = "performance";
services.amd-uprof.enable = true;
# Enable amd-uprof in >= 6.15 kernels only
services.amd-uprof.enable =
let ver = config.boot.kernelPackages.kernel.version;
in (lib.strings.compareVersions ver "6.15") >= 0;
# Disable NUMA balancing
boot.kernel.sysctl."kernel.numa_balancing" = 0;
@@ -42,6 +48,11 @@
# Disable NMI watchdog to save one hw counter (for AMD uProf)
boot.kernel.sysctl."kernel.nmi_watchdog" = 0;
specialisation.oldKernel.configuration = {
system.nixos.tags = [ "old-kernel" ];
boot.kernelPackages = lib.mkForce pkgs.linuxPackages_6_12;
};
services.openssh.settings.X11Forwarding = true;
services.fail2ban.enable = true;

21
m/fox/serial-console.nix Normal file
View File

@@ -0,0 +1,21 @@
{
# Restart the serial console
systemd.services."serial-getty@ttyS1" = {
enable = true;
wantedBy = [ "getty.target" ];
serviceConfig.Restart = "always";
};
# Enable serial console
boot.kernelParams = [
"console=tty1"
"console=ttyS1,115200"
];
# Enable GRUB2 serial console
boot.loader.grub.extraConfig = ''
serial --unit=1 --speed=115200 --word=8 --parity=no --stop=1
terminal_input --append serial
terminal_output --append serial
'';
}

View File

@@ -45,7 +45,7 @@
address = "10.0.40.7";
prefixLength = 24;
} ];
interfaces.ibp5s0.ipv4.addresses = [ {
interfaces.ibs785.ipv4.addresses = [ {
address = "10.0.42.7";
prefixLength = 24;
} ];

View File

@@ -46,7 +46,7 @@
address = "10.0.40.42";
prefixLength = 24;
} ];
interfaces.ibp5s0.ipv4.addresses = [ {
interfaces.ibs785.ipv4.addresses = [ {
address = "10.0.42.42";
prefixLength = 24;
} ];

View File

@@ -20,7 +20,7 @@
address = "10.0.40.1";
prefixLength = 24;
} ];
interfaces.ibp5s0.ipv4.addresses = [ {
interfaces.ibs785.ipv4.addresses = [ {
address = "10.0.42.1";
prefixLength = 24;
} ];

View File

@@ -21,7 +21,7 @@
prefixLength = 24;
} ];
# Watch out! The OmniPath device is not in the same place here:
interfaces.ibp129s0.ipv4.addresses = [ {
interfaces.ibs801.ipv4.addresses = [ {
address = "10.0.42.2";
prefixLength = 24;
} ];

View File

@@ -4,6 +4,7 @@
imports = [
../common/base.nix
../common/ssf/hosts.nix
../common/xeon/console.nix
../module/emulation.nix
../module/debuginfod.nix
../module/nvidia.nix

View File

@@ -27,4 +27,7 @@
};
};
};
# Allow gitea user to send mail
users.users.gitea.extraGroups = [ "mail-robot" ];
}

View File

@@ -19,7 +19,7 @@ in stdenv.mkDerivation {
'';
hardeningDisable = [ "pic" "format" ];
nativeBuildInputs = kernel.moduleBuildDependencies;
patches = [ ./makefile.patch ./hrtimer.patch ];
patches = [ ./makefile.patch ./hrtimer.patch ./remove-wr-rdmsrq.patch ];
makeFlags = [
"KERNEL_VERSION=${kernel.modDirVersion}"
"KERNEL_DIR=${kernel.dev}/lib/modules/${kernel.modDirVersion}/build"

View File

@@ -0,0 +1,20 @@
diff --git a/inc/PwrProfAsm.h b/inc/PwrProfAsm.h
index d77770a..c93a0e9 100644
--- a/inc/PwrProfAsm.h
+++ b/inc/PwrProfAsm.h
@@ -347,6 +347,7 @@
#endif
+/*
#define rdmsrq(msr,val1,val2,val3,val4) ({ \
__asm__ __volatile__( \
"rdmsr\n" \
@@ -362,6 +363,7 @@
:"c"(msr), "a"(val1), "d"(val2), "S"(val3), "D"(val4) \
); \
})
+*/
#define rdmsrpw(msr,val1,val2,val3,val4) ({ \
__asm__ __volatile__( \

View File

@@ -9,13 +9,15 @@
, nanos6
, nodes
, nosv
, mkl
, mpi
, tampi
, openblas
, ovni
, gitBranch ? "master"
, gitURL ? "ssh://git@bscpm04.bsc.es/rarias/bench6.git"
, gitCommit ? "bf29a53113737c3aa74d2fe3d55f59868faea7b4"
, gitCommit ? "fe30c2cfe36b535ef26a0054e010bc005e88ba04"
, useMkl ? true
}:
stdenv.mkDerivation rec {
@@ -42,10 +44,13 @@ stdenv.mkDerivation rec {
nosv
mpi
tampi
ovni
] ++ (if (useMkl) then [
mkl
] else [
openblas
openblas.dev
ovni
];
]);
env = {
NANOS6_HOME = nanos6;