ad@NetBSD.org wrote:

> I just committed a bunch of changes from the vmlocking branch to the kernel.
> Most of them are fairly mechanical changes after modifications to the kernel
> APIs for callouts, kthreads and device access. I don't expect major upheaval
> but undoubtedly there will be some fallout as the changes were fairly
> extensive.

Previously the softclock_si handler is established in initclocks()
after cpu_configure(9), but now it's done in callout_startup2(),
which is called before cpu_configure(9).

On some (most m68k) ports softintr is initialized in cpu_configure(9)
so callout/tsleep (like "waiting 2 seconds" on scsi probe) won't work
properly after this change.

Should we move callout_startup2() after cpu_configure(9), or
initialize MD softintr stuff earlier (in cpu_startup(9) or so)?
---
Izumi Tsutsui