I got an idea today.
We could create a wiki page where we document how LinuxBIOS (porting, and otherwise) is affected by hardware design decisions.
This could be a very helpful resource for vendors that want to leverage LinuxBIOS to the fullest in new designs.
Some or most points will have tradeoffs, but the point is to document them since some decisions that a hardware engineer or even manager or exec may consider unimportant can actually make a big difference for the software developers.
Am I being too naive? Maybe it would be old news?
Here are some discussion points to start with. I'll add a page with content after some feedback. There are many more points, I've only come up with those that affect LinuxBIOS porting. I imagine a lot can be learned from OLPC that would go on this page.
* SuperIO serial port rs232 tx/rx on header or pads, or TTL levels, or EHCI debug port on one of the onboard USB ports.
Pros: Reliable debugging, faster porting Cons: Board land, BOM cost, less so with only TTL and pads, possibly more complex routing for the USB debug port
* Avoid or document flash chip write protection
Pros: Less hard/software requirements for porting, faster porting Cons: Risk of malware or user error bricking board
* Avoid SMM
Pros: Faster porting Cons: Board land, BOM cost
* Use already supported chips
Pros: Faster, if not instant, porting Cons: Limited BOM suppliers, may lack state-of-the-art features
* Avoid embedded controllers (for things that concern LinuxBIOS)
Pros: Faster porting Cons: EC may be required for key features
//Peter
On 3/30/07, Peter Stuge stuge-linuxbios@cdy.org wrote:
I got an idea today.
We could create a wiki page where we document how LinuxBIOS (porting, and otherwise) is affected by hardware design decisions.
This could be a very helpful resource for vendors that want to leverage LinuxBIOS to the fullest in new designs.
This is a fantastic idea. I think you should do it. We can all contribute our horror stories too :-)
thanks
ron
On Sat, Mar 31, 2007 at 06:19:52AM +0200, Peter Stuge wrote:
I got an idea today.
We could create a wiki page where we document how LinuxBIOS (porting, and otherwise) is affected by hardware design decisions.
Great! Go ahead :)
What should the page name be? 'Designing Hardware for LinuxBIOS'? 'LinuxBIOS Hardware Design'? Others?
[...]
- Avoid embedded controllers (for things that concern LinuxBIOS)
Pros: Faster porting Cons: EC may be required for key features
* Only use "open" hardware components (public datasheets, Free Software drivers available, maybe even open hardware description/schematics etc).
* Put a JTAG header on the hardware. Provide a BDSL file or source code patches for (open source) JTAG programmers.
* Most important: Use a _socketed_ BIOS chip.
Uwe.
* Uwe Hermann uwe@hermann-uwe.de [070331 14:56]:
Only use "open" hardware components (public datasheets, Free Software drivers available, maybe even open hardware description/schematics etc).
Put a JTAG header on the hardware. Provide a BDSL file or source code patches for (open source) JTAG programmers.
Most important: Use a _socketed_ BIOS chip.
* or an LPC header