Hello again,
two years have passed since I developed the support for MS-6147. Now I return to slowly activate myself again. In a first run I would like to report the following:
1. Coreboot-v4.0, svn rev 5740, works well on MSI MS-6147.
2. In the overview of supported mainboards, there ought to be an entry "Y" for MSI MS-6147, stipulating that the DIL32-ROM is set in a socket.
3. SeaBIOS works on MS-6147, but I had to strip down most functionality in order to fit the combined image into the 64 kB available (120 byte to spare!) since the standard fallback duplication was left intact. No VGA BIOS was included.
4. From this minimal SeaBIOS payload, I did successfully boot into a fully functional OpenBSD 4.4, which I as always use with a serial console. The disk is a 1GB Compact Flash card.
Best regards,
Mats Erik Andersson, fil. dr mats.andersson@gisladisker.se
Abonnerar på: debian-mentors, debian-devel-games, debian-perl, debian-ipv6, debian-qa
On Fri, Aug 27, 2010 at 11:21:06AM +0200, Mats Erik Andersson wrote:
Hello again,
two years have passed since I developed the support for MS-6147. Now I return to slowly activate myself again. In a first run I would like to report the following:
Coreboot-v4.0, svn rev 5740, works well on MSI MS-6147.
In the overview of supported mainboards, there ought to be an entry "Y" for MSI MS-6147, stipulating that the DIL32-ROM is set in a socket.
SeaBIOS works on MS-6147, but I had to strip down most functionality in order to fit the combined image into the 64 kB available (120 byte to spare!) since the standard fallback duplication was left intact. No VGA BIOS was included.
From this minimal SeaBIOS payload, I did successfully boot into a fully functional OpenBSD 4.4, which I as always use with a serial console. The disk is a 1GB Compact Flash card.
Wow! Coreboot and seabios in a 64KiB flash - that's pretty impressive. Were you using lzma compression with this?
-Kevin
I have worked on variations on the previous setup.
lördag den 4 september 2010 klockan 01:52 skrev Peter Stuge detta:
Mats Erik Andersson wrote:
- In the overview of supported mainboards, there ought to be an entry "Y" for MSI MS-6147, stipulating that the DIL32-ROM is set in a socket.
Done. Flashrom too?
Yes, and in fact the untested erasure of EON EN29F002NT can now be marked as working for Flashrom, besides the already known read/write.
fully functional OpenBSDNice!
Now, when using the full 256kB capacity, some quirks surface in practice.
I have extracted the VGA-BIOS from the original Award-BIOS, and I have incorporated it with Coreboot. The graphics chip is ATI 3D Rage Pro AGP, with pci-id 1002:4742. Thus I do arrive at a serial console, as well as the usual text console. This is a good thing, if not for other reasons that I managed to amalgamate CBFS/VGA-BIOS/SeaBIOS!
When booting Debian GNU/Linux via Coreboot/SeaBIOS, the kernel complains
[0.204012] weird, boot CPU(#0) not listed by BIOS.
Harmless, but annoying. Is CBFS saving somespace?
The booting from SeaBIOS is troublesome. Switching floppy and CD-ROM booting off, did not help. I had to switch PNP, DMA and 32bit-PIO off too, before I got bootable systems.
It is annoying that the OS commences only after
"WARNING - Timeout at await_ide:39!".
This message originates in SeaBIOS, but would CBFS or SeaBIOS be the one to blame the most for this lost time?
The uhci_hcd circuitry is detected, but it is not assigned an IRQ, so Linux as well as OpenBSD considers it to be broken. SeaBIOS reports an io-port 0x20c0, though.
Neither Linux, nor OpenBSD, are able to accomplish a complete power down, as power supply is still applied after shutdown.
Moving the single DIMM to another socket, makes the nortbridge initialisation fail immediately after printing
"Northbridge following SDRAM init".
I thought I had solved that issue two years ago. Did I not, or have other mechanisms been implemented since then? This must be investigated again. In fact, booting is successful with only DIMM2 being populated, not when DIMM1 is used in its stead.
Best regards for now,
Mats E A
Am 08.09.2010 12:55, schrieb Mats Erik Andersson:
with pci-id 1002:4742. Thus I do arrive at a serial console, as well as the usual text console. This is a good thing, if not for other reasons that I managed to amalgamate CBFS/VGA-BIOS/SeaBIOS!
I guess you meant "coreboot"? CBFS is just the filesystem.
When booting Debian GNU/Linux via Coreboot/SeaBIOS, the kernel complains
[0.204012] weird, boot CPU(#0) not listed by BIOS.
Some of the tables (mptable and the like) are probably incomplete or missing. The completeness of tables and ACPI varies a lot between boards. We work on unifying the code, to improve matters for existing boards and simplify support for new ones, but it's far from completed.
"WARNING - Timeout at await_ide:39!".
Might be incomplete IDE init. coreboot usually (depends on chipset) does no init at all. Some chipset docs actually require some (eg. figure out if 80-pin cables are used, and set a bit if so)
The uhci_hcd circuitry is detected, but it is not assigned an IRQ, so Linux as well as OpenBSD considers it to be broken. SeaBIOS reports an io-port 0x20c0, though.
IO Ports are assigned properly in generic code, but assigning IRQ is mainboard/chipset specific code.
Neither Linux, nor OpenBSD, are able to accomplish a complete power down, as power supply is still applied after shutdown.
The board has no ACPI support at all (and we don't do APM either). That might be the reason why this isn't working.
Moving the single DIMM to another socket, makes the nortbridge initialisation fail immediately after printing
"Northbridge following SDRAM init".
Again, chipset specific code.
Regards, Patrick
onsdag den 8 september 2010 klockan 13:08 skrev Patrick Georgi detta:
Am 08.09.2010 12:55, schrieb Mats Erik Andersson:
with pci-id 1002:4742. Thus I do arrive at a serial console, as well as the usual text console. This is a good thing, if not for other reasons that I managed to amalgamate CBFS/VGA-BIOS/SeaBIOS!
I guess you meant "coreboot"? CBFS is just the filesystem.
Yes!
Neither Linux, nor OpenBSD, are able to accomplish a complete power down, as power supply is still applied after shutdown.
The board has no ACPI support at all (and we don't do APM either). That might be the reason why this isn't working.
Does this fit with the fact that with the original Award-BIOS, Linux as well as OpenBSD are able to cut the power supply?
Regards, Patrick
Thank you for your previous explanations.
Regards, Mats E A
Am 08.09.2010 13:40, schrieb Mats Erik Andersson:
Does this fit with the fact that with the original Award-BIOS, Linux as well as OpenBSD are able to cut the power supply?
Sorry, "board" == "board support code in coreboot"
Patrick
-
Kevin O'Connor -
Mats Erik Andersson -
Patrick Georgi -
Peter Stuge