On Sat, Sep 6, 2025 at 7:21 AM Cy Schubert <[email protected]> wrote: > > Hi, > > I'd like to open up discussion about a weird bug I stumbled across on > Thursday. You can get the gory details, including backtrace in PR/289320. > > Long story short, after the FreeBSD version bump from 15 to 16 amdgpu > panics on a kernel page fault. For fun I reverted, locally, the commit that > updated the version strings in FreeBSD-CURRENT from 15 to 16, reverting > 8b4e4c2737305df8807abc6cd054a32586085c93 on that particular machine, a > Framework 13 laptop. The panic was solved. > > My other machines don't use amdgpu, they either don't use X or in the case > of my HP 840, it uses the i915 driver (also part of drm-66-kmod), which > works after the 15 to 16 version bump. But the 15 to 16 version bump causes > amdgpu also from drm-66-kmod to reference a NULL pointer. > > What in drm-66-kmod would be sensitive to a FreeBSD version bump? Blame it on quantum entanglement..
One a slightly more serious note, I'd grep for uses of MODULE_KERNEL_MAXVER in the sources as a starting point. rick > > > -- > Cheers, > Cy Schubert <[email protected]> > FreeBSD UNIX: <[email protected]> Web: https://FreeBSD.org > NTP: <[email protected]> Web: https://nwtime.org > > e**(i*pi)+1=0 > > >
