linux: Add version 6.12.XXX

[donald]

Tested Linux 6.12.27-482 on maleficent (AMD graphics).

[pmenzel]

Yout beat me to it. ;-) I already checked out the branch this morning.

Now that pahole is in MarIuX, it’d be great, if you could built the BTF (BPF Type Format) encoded debug information, so that we can use ptcpdump.

[donald]

With CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y, CONFIG_DEBUG_INFO_BTF=y:

 LD      vmlinux.o
  OBJCOPY modules.builtin.modinfo
  GEN     modules.builtin
  GEN     .vmlinux.objs
  MODPOST Module.symvers
  CC      .vmlinux.export.o
  UPD     include/generated/utsversion.h
  CC      init/version-timestamp.o
  KSYMS   .tmp_vmlinux0.kallsyms.S
  AS      .tmp_vmlinux0.kallsyms.o
  LD      .tmp_vmlinux1
  BTF     .tmp_vmlinux1.btf.o
die__process: DW_TAG_compile_unit, DW_TAG_type_unit, DW_TAG_partial_unit or DW_TAG_skeleton_unit expected got INVALID (0x0) @ 115a59b!
error decoding cu 
pahole: .tmp_vmlinux1: Invalid argument
  NM      .tmp_vmlinux1.syms
  KSYMS   .tmp_vmlinux1.kallsyms.S
  AS      .tmp_vmlinux1.kallsyms.o
  LD      .tmp_vmlinux2
  NM      .tmp_vmlinux2.syms
  KSYMS   .tmp_vmlinux2.kallsyms.S
  AS      .tmp_vmlinux2.kallsyms.o
  LD      vmlinux
  BTFIDS  vmlinux
libbpf: failed to find '.BTF' ELF section in vmlinux
FAILED: load BTF from vmlinux: No data available
make[2]: *** [scripts/Makefile.vmlinux:34: vmlinux] Error 255
make[2]: *** Deleting file 'vmlinux'
make[1]: *** [/scratch/local/linux/Makefile:1179: vmlinux] Error 2
make: *** [Makefile:224: __sub-make] Error 2

Just tried, no time for debugging right now. Any idea?

[pmenzel]

Just tried, no time for debugging right now. Any idea?

Unfortunately, I do not. Sorry for wasting your time. I am going to do more tests next week.

[pmenzel]

For next week from Reddit:

You need BTF debug info enabled for BTF to be found correctly but debug needs to be setup correctly.
Debug compressed needs to be disabled, split debugging disabled, struct randomization for gcc disabled ect.
There are other options that will block or interfere with debugging for btf.

[donald]

What exactly do we need this for?

[pmenzel]

What exactly do we need this for?

To fully use eBPF. For example, ptcpdump is able to also show the process ids. I do not know the details, but it needs the BTF symbols.

[donald]

Shoud we build a nvidia driver for this (if so, what version) ? @thomas

[thomas]

An Nvidia update would be truly useful, I think the 'Latest Production Branch Version' 570.144 makes the most sense. [archive] https://www.nvidia.com/en-us/drivers/unix/ [570.144] https://www.nvidia.com/en-us/drivers/details/243330/ Ahij, Thomas

…

-- Thomas Kreitler - Information Retrieval ***@***.*** 49/30/8413 1702

[pmenzel]

Could you please also build with CONFIG_NET_SCH_INGRESS=m for https://github.molgen.mpg.de/mariux64/mariux64-issues/issues/157#issuecomment-20474? (You could also do this in 6.12.25.)

[pmenzel]

For testing:

@sigcont:~$ sudo bee install -f linux-6.12.23-481 nvidia_current-570.144-0 nvidia_linux-6.12.23-481-570.144-0

[pmenzel]

Size increase of 52 MB for package nvidia_current:

[…]
-rw-r--r-- 1 pmenzel pmenzel 184M Jan 24 15:40 /src/mariux/beeroot/packages/nvidia_current-535.230.02-0.x86_64.bee.tar.bz2
-rw-r--r-- 1 pmenzel pmenzel 236M May  5 10:24 /src/mariux/beeroot/packages/nvidia_current-570.144-0.x86_64.bee.tar.bz2

[pmenzel]

Loading Nvidia 570.144 module fails with

[   15.527570] nvidia: loading out-of-tree module taints kernel.
[   15.533684] nvidia: module license 'NVIDIA' taints kernel.
[   15.539506] Disabling lock debugging due to kernel taint
[   15.545212] nvidia: module license taints kernel.
[   15.836685] i915 0000:00:02.0: enabling device (0000 -> 0003)
[   15.843416] i915 0000:00:02.0: [drm] Found COFFEELAKE (device ID 3e98) display version 9.00 stepping N/A
[   15.854129] i915 0000:00:02.0: [drm] VT-d active for gfx access
[   15.861218] i915 0000:00:02.0: [drm] Using Transparent Hugepages
[   15.872476] i915 0000:00:02.0: [drm] Finished loading DMC firmware i915/kbl_dmc_ver1_04.bin (v1.4)
[   15.924521] nvidia-nvlink: Nvlink Core is being initialized, major device number 238

[   15.934738] nvidia 0000:01:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=io+mem
[   16.218135] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  570.144  Thu Apr 10 20:33:29 UTC 2025
[   16.285645] nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
[   16.354056] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  570.144  Thu Apr 10 20:03:03 UTC 2025
[   16.370767] nvidia_drm: Unknown symbol drm_fbdev_ttm_driver_fbdev_probe (err -2)

[pmenzel]

Works with the old Nvidia driver:

sudo bee install -f nvidia_linux-6.12.23-481-535.230.02-0.x86_64

[pmenzel]

[   16.354056] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  570.144  Thu Apr 10 20:03:03 UTC 2025
[   16.370767] nvidia_drm: Unknown symbol drm_fbdev_ttm_driver_fbdev_probe (err -2)

For the Nvidia module 570.144, manually inserting/loading the module drm_ttm_helper with sudo modprobe drm_ttm_helper, fixes the error. I guess /sbin/nvidiactl needs to be adapted.

[donald]

For the Nvidia module 570.144, manually inserting/loading the module drm_ttm_helper with sudo modprobe drm_ttm_helper, fixes the error. I guess /sbin/nvidiactl needs to be adapted.

We can do that but I wonder why the module isn't inserted automagically if it is needed?