Arch Linux

The error from output you posted is due to [1][2] not implementing a fix similar to [3]. Please post the full journal for a boot with the issue and one without for comparison.

[1] https://aur.archlinux.org/packages/v4l2loopback-dc-dkms
[2] https://github.com/dev47apps/droidcam
[3] https://github.com/umlaeute/v4l2loopback/commit/ddce39b5eb667d3a5db0009f96f6050496bd58ad

For the record, I am experiencing the same issue on Parabola GNU/Linux-libre.

Have you been able to bisect [1] which commit introduced the issue? Is there reason to believe this is a packaging issue? The release announcement for kmod 31 [2] from the same mailing list you can use to report the issue upstream.

[1] https://wiki.archlinux.org/title/Bisecting_bugs_with_Git
[2] https://lore.kernel.org/linux-modules/lnbkvt5hhvgksgjqko7t6niw6uzr5ewjp32wyy2s26rzwdgh2y%40l775iv6f6tkz/T/#u

Maybe related [1]? But not sure why only a small minority would be affected...

[1] https://lore.kernel.org/linux-modules/acf37f00-cea3-4f77-9c05-3c0301221d2a%40moonlit-rail.com/T/

> Maybe related [1]? But not sure why only a small minority would be affected...
Seems fairly easy to test by reverting [2]. See attached diff.

[1] https://lore.kernel.org/linux-modules/acf37f00-cea3-4f77-9c05-3c0301221d2a%40moonlit-rail.com/T/
[2] https://git.kernel.org/pub/scm/utils/kernel/kmod/kmod.git/patch/?id=09c9f8c5df0475d2627d9498e26250592dbfebd6

On Parabola GNU/Linux-libre I experience the issue with linux-libre kernel but it works with linux-xanmod. I don't know if it is relevant, but I use Booster for initramfs generation. I have yet to try the patch provided by loqs.

@dario does the initramfs contain any .xz files? Are there any such files in /usr/lib/modules `find /usr/lib/modules/ -name '*.ko.xz'`?

«find /usr/lib/modules/ -name '*.ko.xz'» yelds nothing with linux-libre kernel.

@dario would it help if I built the patched kmod for you to test? I could also help you with the bisection if needed.

Thank you, loqs, I can build it myself. I will get back to you in a few hours and let you know.

I confirm the patched version is working on Parabola GNU/Linux-libre with linux-libre kernel.

Does the output of `lsinitcpio /path/to/initrd` show any compressed kernel modules?
Does linux-libre enable CONFIG_MODULE_DECOMPRESS CONFIG_MODULE_COMPRESS_GZIP CONFIG_MODULE_COMPRESS_XZ CONFIG_MODULE_COMPRESS_ZSTD?

Of these modules, two are enabled in linux-libre: https://git.parabola.nu/abslibre.git/tree/libre/linux-libre/config.x86_64

# lsinitcpio /boot/booster-linux-libre.img
usr
usr/lib
lib
usr/local
usr/local/lib
usr/bin
usr/sbin
bin
sbin
usr/local/bin
usr/local/sbin
run
var
var/run
usr/lib64
lib64
init
usr/lib/modules
usr/lib/modules/libps2.ko
usr/lib/modules/crc16.ko
usr/lib/modules/vivaldi_fmap.ko
usr/lib/modules/xor.ko
usr/lib/modules/xhci_pci.ko
usr/lib/modules/dm_mod.ko
usr/lib/modules/mbcache.ko
usr/lib/modules/cdrom.ko
usr/lib/modules/i8042.ko
usr/lib/modules/crc32c_intel.ko
usr/lib/modules/xhci_pci_renesas.ko
usr/lib/modules/vfat.ko
usr/lib/modules/cryptd.ko
usr/lib/modules/ext4.ko
usr/lib/modules/isofs.ko
usr/lib/modules/fuse.ko
usr/lib/modules/btrfs.ko
usr/lib/modules/squashfs.ko
usr/lib/modules/dm_snapshot.ko
usr/lib/modules/crypto_user.ko
usr/lib/modules/uas.ko
usr/lib/modules/sr_mod.ko
usr/lib/modules/usb_storage.ko
usr/lib/modules/serio.ko
usr/lib/modules/usbhid.ko
usr/lib/modules/atkbd.ko
usr/lib/modules/crypto_simd.ko
usr/lib/modules/jbd2.ko
usr/lib/modules/crct10dif_pclmul.ko
usr/lib/modules/serio_raw.ko
usr/lib/modules/raid6_pq.ko
usr/lib/modules/libcrc32c.ko
usr/lib/modules/aesni_intel.ko
usr/lib/modules/blake2b_generic.ko
usr/lib/modules/crc32c_generic.ko
usr/lib/modules/ghash_clmulni_intel.ko
usr/lib/modules/fat.ko
usr/lib/modules/dm_bufio.ko
usr/lib/modules/crc32_pclmul.ko
usr/lib/modules/booster.alias
etc
etc/booster.init.yaml
etc/initrd-release

To confirm you are on linux-libre 6.3.3-1 or newer as that added CONFIG_MODULE_DECOMPRESS=y? If so I do not know why that kernel is having issues.

[1] https://git.parabola.nu/abslibre.git/commit/libre/linux-libre/config.x86_64?id=f530eea22924e327de176787a29854fabb8691fd

I am on linux-libre 6.3.3-1.

If you use the config from linux-libre 6.3.3-1 to build linux 6.3.3 does that kernel have the issue? If so does updating the kernel to 6.5.7 keeping the same config still produce the issue?

This kernel built from source works for me: https://git.parabola.nu/abslibre.git/tree/libre/linux-libre. Unfortunately at the moment there is no 6.5.7 version of linux-libre packaged for my distribution to test out.

As the issue appears to be limited to the linux-libre kernel have you raised the issue with the linux-libre maintainers? While there may be an issue with mkinitcpio and xz compressed modules due to the missing --lzma2=dict=1MiB option [2] you are using booster and there are no compressed modules in the initrd.

[1] https://gitlab.archlinux.org/archlinux/mkinitcpio/mkinitcpio/-/blob/v36/mkinitcpio?ref_type=tags#L233
[2] https://lore.kernel.org/all/1bf10ec1-c154-52f3-79dd-7ec5380aa159%40tweek.dk/

I did not notice before but now there is an open ticket for Parabola GNU/Linux-libre: https://labs.parabola.nu/issues/3539. They are aware of the issue. If it turns out to be a problem with linux-libre, the maintainers of the distribution will eventually raise the issue to the Linux-libre maintainers.

The OP didn't specify the kernel though.

According to https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1050582#30, it sounds like kmod should not be trying to use the kernel to decompress XZ modules until it can handle CRC64
and dictionaries larger than 1 MiB.
I was able to resolve this issue by compiling my custom kernel using either the patch in the debian link (needed to edit it for the 6.1 LTS kernel) or using gzip compression (tested both).

Has the issue been fixed for linux-libre [1]?

[1] https://labs.parabola.nu/issues/3539#note-9

Yeah. Now it is using linux-libre-6.5.8-1 vanilla variant as main kernel which is working fine.

Seeing as Arch uses zstd compressed kernel modules by default, the workaround for anyone affected who is using xz compression is straightforward.