I want to throw these arguments into the debate:
I vote for the ELF payload. I would say that linking it in would be just as good, except that if EFI ever does really become common, a lot of these payloads would become useful to a larger user base. I do agree that there would be some redundancy in code, but I also believe the build process would be simplified (rom-o-matic certainly seems more feasible without so many options) and those new to LinuxBIOS would be able to grasp how it all links together easier.
Right now, the split in idealogy has put us in quite the pickle. Linuxbios has two ways to tftp boot, three ways to polled ide boot, two ways to get a vga rom up and running.. If things are still even linked in to freebios2, there's a lot of confusion. Do different boot methods use elf payloads or are they compiled in? Well that would depend on which one. I too would like the Linux kernel to solve all of these problems, I hope I've never created the impression that I wish otherwise. But the chunk of mailing list traffic from embedded enthusiasts is getting still larger, nothing's really changed in ROM size since the Great Payload Debates of 2001, and Linux has only gotten bigger (though there's hope..).
- Adam Agnew
On Fri, 8 Aug 2003, ron minnich wrote:
folks, any opinions expressed on this VGA BIOS discussion would be welcome.
VGA BIOS in linuxbios, or in an ELF payload? What's your bet?
ron
Greetings,
At the risk of getting too complex, we are already generating a map for LinuxBIOS, what about placing it into an in RAM table so payloads can call the functions themselves rather than duplicate them?
That would certainly help out things like ADLO that need to do something chipset specific in an otherwise generic function (such as control shadow RAM).
The problem with BIOS and DOS wasn't that there were int calls, it was just that they were such an ad-hoc mismash and poorly defined.
Ultimately, I do want Linux as the bootloader, but form the looks of things, 2.6 is still too big unless the network stack is compiled out.
Perhaps 2.7 will see a continuation of the embedded configuration so we can have a very simple stack and polled I/O driver set, but it's clearly not ready yet.
Meanwhile even as bigger flash chips become available, mainboard designers almost seem determined to do the wrong thing.
G'day, sjames
On Fri, 8 Aug 2003, Adam Agnew wrote:
I want to throw these arguments into the debate:
I vote for the ELF payload. I would say that linking it in would be just as good, except that if EFI ever does really become common, a lot of these payloads would become useful to a larger user base. I do agree that there would be some redundancy in code, but I also believe the build process would be simplified (rom-o-matic certainly seems more feasible without so many options) and those new to LinuxBIOS would be able to grasp how it all links together easier.
Right now, the split in idealogy has put us in quite the pickle. Linuxbios has two ways to tftp boot, three ways to polled ide boot, two ways to get a vga rom up and running.. If things are still even linked in to freebios2, there's a lot of confusion. Do different boot methods use elf payloads or are they compiled in? Well that would depend on which one. I too would like the Linux kernel to solve all of these problems, I hope I've never created the impression that I wish otherwise. But the chunk of mailing list traffic from embedded enthusiasts is getting still larger, nothing's really changed in ROM size since the Great Payload Debates of 2001, and Linux has only gotten bigger (though there's hope..).
- Adam Agnew
On Fri, 8 Aug 2003, ron minnich wrote:
folks, any opinions expressed on this VGA BIOS discussion would be welcome.
VGA BIOS in linuxbios, or in an ELF payload? What's your bet?
ron
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
-
Adam Agnew -
steven james