Panic on Rpi3 at r358976

[Kyle Evans]

On Sun, Mar 15, 2020 at 11:11 PM Mark Millard <marklmi at yahoo.com> wrote:
>
>
>
> On 2020-Mar-15, at 19:03, Kyle Evans <kevans at freebsd.org> wrote:
>
> > On Sun, Mar 15, 2020 at 8:54 PM Mark Millard <marklmi at yahoo.com> wrote:
> >>> On 2020-Mar-15, at 17:33, Kyle Evans <kevans at freebsd.org> wrote:
> >>
> >>> On Sat, Mar 14, 2020 at 11:12 PM bob prohaska <fbsd at www.zefox.net> wrote:
> >>>>
> >>>> Tried to boot a kernel built from r358976 on a Pi3 and got a panic:
> >>>>
> >>>> [... snip ...]
> >>>
> >>> Hi,
> >>>
> >>> I've got a patch against sysutils/u-boot-rpi{3,4} based on what I've
> >>> submitted upstream that I'm test-building again and will soon be
> >>> submitting to Phabricator; please give it a shot and confirm if it
> >>> makes life happier or not:
> >>> https://people.freebsd.org/~kevans/rpi-psci.diff
> >>
> >> I grep'd in the area that holds where I did the
> >> investigative patch that enabled the RPi4 to boot
> >> and such without the armstub8-gic.bin memory being
> >> slamed. (I've not done any clean-out of the materials
> >> in that area.)
> >>
> >> The result is not suggestive of CONFIG_RPI_EFI_NR_SPIN_PAGES
> >> making a difference:
> >>
> >> [... snip ...]
> >
> > Indeed; note these lines in my patch:
> >
> > PATCHFILES+= 1245351/raw 1245352/raw
> >
> > These pull in the patches I submitted upstream that introduces
> > CONFIG_RPI_EFI_NR_SPIN_PAGES so that they don't have to accept an
> > arbitrary bump of the reserved page count, since it's just our PSCI
> > stub that's larger.
>
> Sorry. The day has gone as one where I need to
> separately validate that I've not omitted something
> for pretty much whatever I was trying to do.
>
> I applied the patch and rebuilt and substituted
> the new u-boot.bin for my hacked one. The context
> is head -r358510 that was booting with my hack
> okay. (I'll soon be updating to -r358966 .)
>
> The result did not go well. Using boot -v
> indicates that the 2nd page is not protected on
> the RPi4:
>
> [... snip ...]

Indeed- my apologies. In the config fragment,  it should read
CONFIG_RPI_EFI_NR_SPIN_PAGES=2 rather than
CONFIG_RPI_EFI_NR_SPIN_PAGES="2" (note the lack of quotes in the
correct version) -- apparently I hadn't transcribed that properly when
I was porting it from my initially tested patch to u-boot.

Thanks,

Kyle Evans