Acked-by: Myles Watson mylesgw@gmail.com
Remaining problem: in src/arch/i386/Kconfig, I define AP_IN_SIPI_WAIT. But in .config, I get: #CONFIG_AP_IN_SIPI_WAIT is not set Clearly I have more to learn about Kconfig. I want it so that if the mainboard Kconfig does not select it, it comes up as 'n'.
I think it's working like you want it to. If it was not defined it wouldn't appear in .config.
$ grep CONFIG_AP .config # CONFIG_AP_IN_SIPI_WAIT is not set
$ grep -Ir SIPI build/ build/coreboot.map:00000000 A CONFIG_AP_IN_SIPI_WAIT build/ldoptions:CONFIG_AP_IN_SIPI_WAIT = 0; build/auto.conf:CONFIG_AP_IN_SIPI_WAIT=n build/config.h:#define CONFIG_AP_IN_SIPI_WAIT 0 build/coreboot_ram.map:00000000 A CONFIG_AP_IN_SIPI_WAIT
Thanks, Myles
On Fri, Aug 28, 2009 at 07:51:42AM -0700, ron minnich wrote:
Index: src/cpu/x86/tsc/Makefile.inc
--- src/cpu/x86/tsc/Makefile.inc (revision 4607) +++ src/cpu/x86/tsc/Makefile.inc (working copy) @@ -1,7 +1,2 @@ obj-y += delay_tsc.o
-# default CONFIG_TSC_X86RDTSC_CALIBRATE_WITH_TIMER2=0 -# if CONFIG_UDELAY_TSC -# default CONFIG_HAVE_INIT_TIMER=1 -# object delay_tsc.o -# end
This looks strange. You use "select UDELAY_TSC" below, but that is never defined anywhere, and also not used here?
Index: src/mainboard/amd/rumba/Kconfig
--- src/mainboard/amd/rumba/Kconfig (revision 0) +++ src/mainboard/amd/rumba/Kconfig (revision 0) @@ -0,0 +1,62 @@
[...]
+config BOARD_AMD_RUMBA
- bool "Rumba"
- select ARCH_X86
- select CPU_AMD_GX2
- select NORTHBRIDGE_AMD_GX2
- select SOUTHBRIDGE_AMD_CS5536
- select UDELAY_TSC
- select HAVE_PIRQ_TABLE
- help
AMD Rumba mainboard.
[...]
+config TSC_X86RDTSC_CALIBRATE_WITH_TIMER2
- int
- default 0
- depends on BOARD_AMD_RUMBA
Same as UDELAY_TSC, never defined?
Index: src/mainboard/amd/rumba/Makefile.inc
--- src/mainboard/amd/rumba/Makefile.inc (revision 0) +++ src/mainboard/amd/rumba/Makefile.inc (revision 0) @@ -0,0 +1 @@ +include $(src)/mainboard/Makefile.romccboard.inc
We'll have to check if this works. From a quick glance the Rumba does not have the mmx related lines (which _are_ in Makefile.romccboard.inc, though):
crt0-y += ../../../../src/cpu/x86/fpu/enable_fpu.inc crt0-y += ../../../../src/cpu/x86/mmx/enable_mmx.inc crt0-y += auto.inc crt0-y += ../../../../src/cpu/x86/mmx/disable_mmx.inc
Rumba only has:
mainboardinit cpu/x86/fpu/enable_fpu.inc mainboardinit ./auto.inc
It's not a big deal to adapt Makefile.romccboard.inc to handle both cases, but it would be good to know if those mmx lines are actually needed in general (and on GX2 specifically).
Uwe.
On Fri, Aug 28, 2009 at 10:09 AM, Uwe Hermannuwe@hermann-uwe.de wrote:
On Fri, Aug 28, 2009 at 07:51:42AM -0700, ron minnich wrote:
Index: src/cpu/x86/tsc/Makefile.inc
--- src/cpu/x86/tsc/Makefile.inc (revision 4607) +++ src/cpu/x86/tsc/Makefile.inc (working copy) @@ -1,7 +1,2 @@ obj-y += delay_tsc.o
-# default CONFIG_TSC_X86RDTSC_CALIBRATE_WITH_TIMER2=0 -# if CONFIG_UDELAY_TSC -# default CONFIG_HAVE_INIT_TIMER=1 -# object delay_tsc.o -# end
This looks strange. You use "select UDELAY_TSC" below, but that is never defined anywhere, and also not used here?
We should define this in either src/cpu/x86/Kconfig or src/cpu/x86/tsc/Kconfig.
I vote for src/cpu/x86/Kconfig because the src/cpu/x86 directory is really way too fragmented already.
Index: src/mainboard/amd/rumba/Kconfig
--- src/mainboard/amd/rumba/Kconfig (revision 0) +++ src/mainboard/amd/rumba/Kconfig (revision 0) @@ -0,0 +1,62 @@
[...]
+config BOARD_AMD_RUMBA
- bool "Rumba"
- select ARCH_X86
- select CPU_AMD_GX2
- select NORTHBRIDGE_AMD_GX2
- select SOUTHBRIDGE_AMD_CS5536
- select UDELAY_TSC
- select HAVE_PIRQ_TABLE
- help
- AMD Rumba mainboard.
[...]
+config TSC_X86RDTSC_CALIBRATE_WITH_TIMER2
- int
- default 0
- depends on BOARD_AMD_RUMBA
Same as UDELAY_TSC, never defined?
Defined here, right? But it really ought to be in the src/cpu/x86 tree somewhere.
Index: src/mainboard/amd/rumba/Makefile.inc
--- src/mainboard/amd/rumba/Makefile.inc (revision 0) +++ src/mainboard/amd/rumba/Makefile.inc (revision 0) @@ -0,0 +1 @@ +include $(src)/mainboard/Makefile.romccboard.inc
We'll have to check if this works. From a quick glance the Rumba does not have the mmx related lines (which _are_ in Makefile.romccboard.inc, though):
I speculate that rumba ought to just have its own?
crt0-y += ../../../../src/cpu/x86/fpu/enable_fpu.inc crt0-y += ../../../../src/cpu/x86/mmx/enable_mmx.inc
Or we need to make these conditional includes. Your call, I think you know better than I do.
crt0-y += auto.inc crt0-y += ../../../../src/cpu/x86/mmx/disable_mmx.inc
Rumba only has:
mainboardinit cpu/x86/fpu/enable_fpu.inc mainboardinit ./auto.inc
It's not a big deal to adapt Makefile.romccboard.inc to handle both cases, but it would be good to know if those mmx lines are actually needed in general (and on GX2 specifically).
I will hold off on committing until we decide these things, or I get back to this and this patch has hung ... it's already acked :-)
ron
I went ahead and fixed it, the questions were not that hard.
On Fri, Aug 28, 2009 at 7:59 PM, ron minnichrminnich@gmail.com wrote:
I went ahead and fixed it, the questions were not that hard.
Except for the Makefile question.
Committed revision 4618.
ron
Rumba now builds. I can't say if it will work, but this is a big step.
ron
On 8/29/09 5:03 AM, ron minnich wrote:
Rumba now builds. I can't say if it will work, but this is a big step.
Do you get VSA from CBFS already?
On Sat, Aug 29, 2009 at 1:10 PM, Stefan Reinauerstepan@coresystems.de wrote:
On 8/29/09 5:03 AM, ron minnich wrote:
Rumba now builds. I can't say if it will work, but this is a big step.
Do you get VSA from CBFS already?
That's next. I was going to put it in / and call it VSA.
ron
-
Myles Watson -
ron minnich -
Stefan Reinauer -
Uwe Hermann