The attached patch includes preliminary support for the Intel 3100 integrated northbridge/southbridge/superio (http://www.intel.com/design/chipsets/embedded/3100.htm). So far the code has been tested only on the Mt. Arvon development board (http://www.intel.com/design/intarch/devkits/pm3100.htm), which includes a 1.8GHz Pentium M and DDR2-400 ECC DRAM.
The code works well enough to boot Linux 2.6.23.1 with serial console (on the 3100's UART), IOAPIC, working USB, and working DRAM (memtest86 passes). The kernel reports the CPU is running at 600MHz; haven't figured out how to crank it up to full speed. A number of features are incomplete or untested, including PCI Express and IDE/SATA.
The southbridge code was adapted from 6300esb, the northbridge from e7520, and I forget where the mainboard code came from. Kudos to whoever implemented those.
I'm working with a fairly old version of LinuxBIOSv2 so the patch probably breaks the latest svn. Nevertheless:
Signed-off-by: Ed Swierk eswierk@arastra.com
--Ed
Hi Ed,
On 27.12.2007 08:18, Ed Swierk wrote:
The attached patch includes preliminary support for the Intel 3100 integrated northbridge/southbridge/superio (http://www.intel.com/design/chipsets/embedded/3100.htm). So far the code has been tested only on the Mt. Arvon development board (http://www.intel.com/design/intarch/devkits/pm3100.htm), which includes a 1.8GHz Pentium M and DDR2-400 ECC DRAM.
Great!
The code works well enough to boot Linux 2.6.23.1 with serial console (on the 3100's UART), IOAPIC, working USB, and working DRAM (memtest86 passes). The kernel reports the CPU is running at 600MHz; haven't figured out how to crank it up to full speed. A number of features are incomplete or untested, including PCI Express and IDE/SATA.
This is a good start.
The southbridge code was adapted from 6300esb, the northbridge from e7520, and I forget where the mainboard code came from. Kudos to whoever implemented those.
I'm working with a fairly old version of LinuxBIOSv2 so the patch probably breaks the latest svn.
Could you try to merge up to current svn and retest? If it still works, merging will be easier. Then again, all files except src/include/device/pci_ids.h are newly created and shouldn't conflict.
It seems that your code still depends on ROMCC. Do you see any chance to switch to CAR for initialization? I have some CAR fixes pending for Intel chips which should fix some problems of that code.
One final question: Did you use the public data sheets or are you under NDA with intel?
Regards, Carl-Daniel
On 12/27/07, Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net wrote:
It seems that your code still depends on ROMCC. Do you see any chance to switch to CAR for initialization? I have some CAR fixes pending for Intel chips which should fix some problems of that code.
The CAR stuff is still a bit of a mystery to me--is there a working Intel mainboard using CAR that you can point me to?
One final question: Did you use the public data sheets or are you under NDA with intel?
I used only publicly-available datasheets when developing this code.
--Ed
On 27.12.2007 16:16, Ed Swierk wrote:
On 12/27/07, Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net wrote:
It seems that your code still depends on ROMCC. Do you see any chance to switch to CAR for initialization? I have some CAR fixes pending for Intel chips which should fix some problems of that code.
The CAR stuff is still a bit of a mystery to me--is there a working Intel mainboard using CAR that you can point me to?
No idea if it works, but tyan/s2735 can either use CAR or ROMCC. CAR is the default, that can be changed by setting USE_DCACHE_RAM=0.
One final question: Did you use the public data sheets or are you under NDA with intel?
I used only publicly-available datasheets when developing this code.
Good to know. That way, someone else has a way to verify your code against the docs.
Regards, Carl-Daniel
On 27.12.2007 16:16, Ed Swierk wrote:
The CAR stuff is still a bit of a mystery to me--is there a working Intel mainboard using CAR that you can point me to?
The original CAR work was done by an Intel employee on leave at UNM, and tested first on Intel hardware.
I just can't remember which board it was on!
And it was utterly open -- all the details we needed were in the very useful Vol III.
ron
-
Carl-Daniel Hailfinger -
Ed Swierk -
ron minnich