[PATCH]kconfig and romcc

[Patrick Georgi]

Hi,

attached patches add support for tinybootblock to the generic
Makefile.romccboard.inc, and move dell/s1850 (non-tinybootblock) and
emulation/qemu-x86 (tinybootblock) to use that instead of their own set
of rules.

emulation/qemu-x86 is boot tested.

All romcc-boards use Makefile.romccboard.inc after these patches.

Signed-off-by: Patrick Georgi

[Peter Stuge]

Patrick Georgi wrote:
> attached patches add support for tinybootblock to the generic
> Makefile.romccboard.inc, and move dell/s1850 (non-tinybootblock) and
> emulation/qemu-x86 (tinybootblock) to use that instead of their own set
> of rules.

Why is s1850 not using TINY_BOOTBLOCK? Is it a technical problem or a
matter of not-done-yet?
> emulation/qemu-x86 is boot tested.
>
> All romcc-boards use Makefile.romccboard.inc after these patches.
>
> Signed-off-by: Patrick Georgi

Acked-by: Peter Stuge

[Patrick Georgi]

Am 08.01.2010 12:06, schrieb Peter Stuge:
> Why is s1850 not using TINY_BOOTBLOCK? Is it a technical problem or a
> matter of not-done-yet?

not-done-yet. I don't know if it requires a rom enable sequence. If yes,
it's not implemented yet (which might be considered a technical problem).
> Acked-by: Peter Stuge

Thanks,
Patrick

[ron minnich]

For s1850, if I end up going back to it, I'm probably going to want to
put serialice in the top 64k, and make it possible for serialice to
flash. That gives me a test and recovery path for failed BIOSes that's
almost as good as fallback. It's just too difficult to blow a machine
up each time I have a bios problem.

ron

[Kevin O'Connor]

On Fri, Jan 08, 2010 at 09:15:05AM +0100, Patrick Georgi wrote:
> Hi,
>
> attached patches add support for tinybootblock to the generic
> Makefile.romccboard.inc, and move dell/s1850 (non-tinybootblock) and
> emulation/qemu-x86 (tinybootblock) to use that instead of their own set
> of rules.
>
> emulation/qemu-x86 is boot tested.
>
> All romcc-boards use Makefile.romccboard.inc after these patches.

Thanks!

I've boot tested epia-cn with a Kconfig build.  I did need to add:

select TSC_X86RDTSC_CALIBRATE_WITH_TIMER2

to src/mainboard/via/epia-cn/Kconfig.  I also added:

obj-$(CONFIG_GENERATE_ACPI_TABLES) += fadt.o
obj-$(CONFIG_GENERATE_ACPI_TABLES) += dsdt.o
obj-$(CONFIG_GENERATE_ACPI_TABLES) += acpi_tables.o

to src/mainboard/Makefile.romccboard.inc.

Also, I noted that epia-m is also not setup to use romcc.

-Kevin

[Kevin O'Connor]

On Fri, Jan 08, 2010 at 07:52:57PM -0500, Kevin O'Connor wrote:
> On Fri, Jan 08, 2010 at 09:15:05AM +0100, Patrick Georgi wrote:
> > attached patches add support for tinybootblock to the generic
> > Makefile.romccboard.inc, and move dell/s1850 (non-tinybootblock) and
> > emulation/qemu-x86 (tinybootblock) to use that instead of their own set
> > of rules.

I just booted my epia-cn with Kbuild and tinybootblock.

I enabled tinybootblock by adding a "select TINY_BOOTBLOCK" to the
epia-cn/Kconfig file.  This did boot, but it took a very long time,
because the romstage XIP code wasn't cached (running lzma
uncompression without cache is crazy).  I was able to fix it by
adding:

config XIP_ROM_BASE
        hex
        default 0xfff00000

config XIP_ROM_SIZE
        hex
        default 0x100000

to the Kbuild file to force code caching.  (Though, I'm not sure how
much of this is because of other changes in my tree to enable
caching.)

-Kevin

[Patrick Georgi]

Am 09.01.2010 02:34, schrieb Kevin O'Connor:
> config XIP_ROM_BASE
>         hex
>         default 0xfff00000
>
> config XIP_ROM_SIZE
>         hex
>         default 0x100000
>
> to the Kbuild file to force code caching.  (Though, I'm not sure how
> much of this is because of other changes in my tree to enable
> caching.)

Intel and AMD CAR use an automatically adapted variable to retarget the
XIP area (without changing to change the size):
#if defined(CONFIG_TINY_BOOTBLOCK) && CONFIG_TINY_BOOTBLOCK
#define REAL_XIP_ROM_BASE AUTO_XIP_ROM_BASE
#else
#define REAL_XIP_ROM_BASE CONFIG_XIP_ROM_BASE
#endif
        movl    $REAL_XIP_ROM_BASE, %eax
        orl     $MTRR_TYPE_WRBACK, %eax
        wrmsr

instead of
    movl $(CONFIG_XIP_ROM_BASE | MTRR_TYPE_WRBACK), %eax

I'm not sure at which point XIP is configured on your system, maybe this
helps.

Patrick

[Kevin O'Connor]

On Sat, Jan 09, 2010 at 04:27:00PM +0100, Patrick Georgi wrote:
> Intel and AMD CAR use an automatically adapted variable to retarget the
> XIP area (without changing to change the size):

[...]
> I'm not sure at which point XIP is configured on your system, maybe this
> helps.

Thanks.  Looks like the equivalent should be added to
do_early_mtrr_init() in src/cpu/x86/mtrr/earlymtrr.c.

I had hacked rom caching into entry16.inc (which tinybootblock still
uses), but I no longer think doing that is a good idea.

I think having some documentation on what each coreboot phase does
would help here.  In particular, documenting what gets cached in each
phase would be useful.  I'll send a separate email.

-Kevin