Fix building on 6.15

[Lawstorant]

6.15 introduces breaking changes in timers: https://lkml.org/lkml/2025/4/6/101

After fix (no errors on 6.15 and previous versions):

(2/3) Install DKMS modules
==> dkms install --no-depmod xone/0.3.3 -k 6.14.1-arch1-1
==> dkms install --no-depmod xone/0.3.3 -k 6.12.22-1-lts
==> dkms install --no-depmod xone/0.3.3 -k 6.15.0-rc1-00060-ga24588245776
==> depmod 6.12.22-1-lts
==> depmod 6.15.0-rc1-00060-ga24588245776
==> depmod 6.14.1-arch1-1

Fixes #44

[tskaar]

FYI:
The "new" API for timer_*_sync() was introduced in kernel v6.2 [1], so you could instead do the conditional:

#if LINUX_VERSION_CODE >= KERNEL_VERSION(6,2,0)

And the hrtimer_setup() was introduced in kernel v6.13 [2], so you could do the conditional:

#if LINUX_VERSION_CODE >= KERNEL_VERSION(6,13,0)

This would make it easier to increase the "minimum kernel version" if that were to happen down the line.

Sources:
[1] - timer_delete_sync() - kernel v6.2
[2] - hrtimer_setup() - kernel v6.13

Other than that, thanks for the fix! 🚀

[Lawstorant]

I know, but on the other hand you normally want to limit possible blast radius. If, for some reason, my PR actually would break the driver, this way it will only break it for 6.15 which already is broken. This won't break it for already known-working versions.

In software dev and all things IT better is the enemy of good.

[dlundqvist]

I know, but on the other hand you normally want to limit possible blast radius. If, for some reason, my PR actually would break the driver, this way it will only break it for 6.15 which already is broken. This won't break it for already known-working versions.

From kernel's point of view a NULL function is a bug since there's a WARN_ON_ONCE for that condition. So please pass in correct callback in hrtimer_setup.

[Lawstorant]

Good catch, I now checked how to correctly handle this. I'll update in a moment

[dlundqvist]

Thanks.