Re: [Qemu-devel] [PATCH v3 7/7] raspi: add raspberry pi 2 machine

[Andrew Baumann]

Hi Peter,

> From: Peter Crosthwaite [mailto:address@hidden
> Sent: Tuesday, 12 January 2016 16:44
> On Tue, Jan 12, 2016 at 3:53 PM, Andrew Baumann
> <address@hidden> wrote:
> >> From: Peter Crosthwaite [mailto:address@hidden
> >> Sent: Monday, 11 January 2016 19:58
> > [...]
> >> > +static void write_board_setup(ARMCPU *cpu, const struct
> arm_boot_info
> >> *info)
> >>
> >> This is almost identical to Highbank, I'm guessing you are stubbing monitor
> >> firmware where you get away with nopping all the SMCs? Maybe we
> should
> >> split
> >> Highbanks version off to common code, and parameterise the few
> >> differences.
> >>
> >> write_board_setup_dummy_monitior(ARMCPU *cpu ..., uint32_t
> scr_flags);
> >>
> >> or something like it. Makes sense to be in arm/boot.c .
> >
> > Actually, I added this only to make Linux happy (and yes, it was derived
> from highbank). Without it, I was seeing complaints about:
> > Ignoring attempt to switch CPSR_A flag from non-secure world with
> SCR.AW bit clear
> > Ignoring attempt to switch CPSR_F flag from non-secure world with SCR.FW
> bit clear
> >
> > I don't believe anything actually uses the SMC handler after boot. I think 
> > it's
> just an architectural requirement to flush the change to non-secure mode.
> >
> > I would prefer not to touch highbank, because I don't know how to test it.
> Is it better to submit this patch without the board setup?
> >
> 
> I can test Highbank for you, or you can use this project:
> 
> https://lists.gnu.org/archive/html/qemu-devel/2015-10/msg00755.html
> 
> The build takes a while and costs about 50GB of disk space, but the
> amount of setup needed should be pretty low.
> 
> FWIW, that was the test system that found the need for this FW in HB.

If it's easy for you, I'd prefer if you could test with Highbank. You should be 
able to apply patch 7/8 from v4 of the series (sent just now) independently. 
The changes are minor but significant: mostly avoiding armv7 instructions like 
movw, and loading the vectors and setup code as separate ROM blobs. Overall, 
I'm still a little worried about the long-term maintainability of this approach 
versus separate copies...

Thanks,
Andrew