Re: NetBSD/hpcmips userland crashes (Re: IBM Workpad z50 dmesg collection)

On Mon, 27 Mar 2023 at 23:34, Alexander Bochmann <[email protected]> wrote:
>
> Hi David,
>
> ...on 2023-03-26 11:31:07, David Brownlee wrote:
>
>  > - Download and test boot 2016-07-11 kernel from
>  > http://sync.absd.org/hpcmips/ (with the existing netbsd-7 install)
>
> 2016-07-11, of all things, is missing ;)
>
> I can boot 2016-07-10, but 2016-07-12 and a couple of those after
> that I tried from the new set won't even boot to multiuser on my
> machine - they will stop with:
>
> pcic0:  WARNING:  powerhook_establish is deprecated
>
> On a working kernel, ne0 would be detected right after that.

OK, I'm running a final build of
388e93b5d10e1d069771de62b3f9bf4227dedb34, which is the commit directly
before this one:

    commit f694fbd0571344422f046e53019ab383c37b8816
    Author: matt <[email protected]>
    Date:   Mon Jul 11 16:15:35 2016 +0000
        Change MIPS to use the common pmap code.
        Switch to 8KB pages on CPUs with a R4K MMU.
        Simplify cache code.
        Merge in most of changes from matt-mips64 branch

I would expect this to work just fine. The above commit was quite
large, and there were a bunch of follow up commits to the above,
tweaking things for different mips platforms after it, so it is
*possible* that  the initial vr41 issues from the above change were
resolved, and the breakage comes from something else, but Occam
suggests that its most likely the above.

Warner mentioned somethings about needing to disable interrupts in the
kernel, which makes the following very suggestive...

    commit dd0f28c401d101194103bc28f693177bb614e1d5
    Author: skrll <[email protected]>
    Date:   Tue Jul 26 05:52:55 2016 +0000
        Set the cause register to zero after disabling interrupts now that spl0
        doesn't do it.
        My cobalt now boots (again again)

How hard would it be to put in a hack to disable interrupts in a
current kernel to test...

David